One of the hottest topics in storage is automatic storage tiering. Rick Vanover explains what it is and why it will be a big splash in many SANs.

———————————————————————————————————

Storage tiering is a practice that has existed in many ways for storage administrators for a very long time. In a static disk assignment, you can provision a number of storage tiers by manipulating these key design elements:

• Storage protocol: Ethernet-based (iSCSI/NFS), fibre channel, or direct-attached storage. Each connectivity media has its own throughput that will affect the overall experience of the storage.
• Drive speed: The revolutions per minute (RPMs) of the drives in an array is a factor in the overall performance of the storage design.
• Drive interface: The most popular disk interfaces in use today are Ultra320 SCSI, SATA, SAS, fibre channel, and solid state drives. The throughput and I/O operations per second (IOPS) for each of these drive types is a factor in determining the behavior of the storage array.
• RAID level in use: RAID 1, 4, 5, 6, 0+1, 5+0, and other proprietary levels can make significant differences in throughput. Check this AC&NC RAID.EDU resource for information on standard RAID levels and the NetApp page on RAID-DP.
• Quantity of drives: Generally speaking, if your array hits on more drives, you can access less surface area from more drives to enhance the performance of the array.
• Disk size: Drives that are very large (potentially with many arrays striped across them) can bog down the overall throughput of the array. While the 2 or 4 Terabyte drives are attractive for SATA storage, the throughput and interface rate for the drives are the same as 1 TB or smaller drives.

For most storage systems, aligning these design elements to craft the best performing storage system with the resources available or what can be purchased is about as much storage tiering as can be done. If you’ve never built up a few designs and put a performance benchmark on the storage design, you really should — there can be an incredible variance in the performance results.

As you can see, this can be very tedious in the flat storage arena without advanced management. A new set of features are showing up on some of the more full-featured storage processors that provide automated storage tiering. Automated storage tiering will allow the storage processor to put the segments of data on the level of disk that it needs, when it needs it. Examples of this technology are 3PAR’s Adaptive Optimization, Compellent’s Data Progression feature, IBM’s Easy Tier, and EMC’s fully automated storage tiering (FAST). This video shows a preview of EMC’s FAST:

The automated storage tiering technologies allow the storage administrator to permit a volume that may reside on a lesser-performing disk set to be dynamically moved to a higher-performing tier automatically. One of the best use cases is to put the bulk of a SAN’s storage requirements on less expensive SATA storage and use automated storage tiering to move those volumes or sub volumes to higher performing SAS or solid state disk drives.

The ability to automate storage tiers is quite attractive, primarily because the right resources will get the right disk when it is needed; however, some administrators may express concern about the data blocks or volumes being moved dynamically around a SAN. While each product will implement automated tiered storage differently, it is not far from standard volume migration technologies that exist on most storage processors, operating systems, or hypervisors. I can see a great reduction in the amount of tier-1 (SAS) or tier-0 (solid state disk) storage zones that many administrators may need to provision in order to keep the hot spots on the premium disk.

Where does automated storage tiering fit into your storage roadmap? Let us know in the comments.

TechRepublic’s Servers and Storage newsletter, delivered on Monday and Wednesday, offers tips that will help you manage and optimize your data center. Automatically sign up today!

For a high-level overview of the health of your virtual environment, check out VKernel’s free Capacity View tool.

—————————————————————————————————-

Back in the days of the physical server, an administrator purchasing a server would size a server with plenty of RAM, disk, and processor for a server’s long life. In many cases, computing resources were over-provisioned in an attempt to make sure that potentially growing resource needs would not max out the resources of the new server. While over-provisioning resources cost a little extra money on the front end, upgrading a server mid-cycle tended to be pretty expensive.

How times have changed.

The era of virtualization is in full swing, and over-provisioning servers — now in the form of virtual machines (VMs) — is not only unnecessary since resources can be added on-the-fly, but it also has a significant cost. For example, every unnecessary gigabyte of RAM added to a VM is a gigabyte of RAM taken from the central resource pool that can’t be added to some other needy VM. The same goes for disk space and processor. We’re in an era of “right-sizing” our individual workloads. By right-sizing VMs, companies avoid prematurely buying additional virtual hosts.

Although VMware and Hyper-V provide performance-viewing capabilities, both tools are far from complete and simply provide you with a point-in-time look at where things stand at any particular point. From that aspect, the built-in tools don’t really provide much in the way of analysis. VKernel aims to fix that shortcoming by offering tools that take performance analysis to the next level. Some of VKernel’s tools, including Capacity View, are free, although there are some limitations, which I discuss below.

(Note: For this article, I’m using Capacity View with VMware vSphere and vCenter.)

The free Capacity View tool provides you with a very high-level overview of the health of your virtual environment. Beyond telling you what resources are currently being consumed by individual VMs, Capacity View gives you insight into how appropriately your VMs are sized, but only at an aggregate level. For example, Capacity View won’t tell you that the VM named vm-print is over- or under-subscribed, but it will tell you that you have 10 VMs using too much RAM or 12 VMs that haven’t been assigned enough disk resources.

You can download Capacity View from VKernel’s Web site; it’s quick and easy to install. Once installed and upon first execution, you’re prompted to answer a few questions (Figure A). Specifically, you need to point Capacity View at your vCenter server or, if you don’t have a vCenter server, at an individual ESX host. The only other parameters that you need to specify are the Username and Password parameters that you see in Figure A.

Figure A

The Capacity View configuration screen. (Click the image to enlarge.)

Capacity View makes heavy use of vSphere’s built-in API to gather performance statistics that are then matches against the limited parameters that you specified on the configuration page.  Capacity View then displays a dashboard-style screen with an overview of your virtual environment.  One such screen is shown in Figure B and gives you an overview of a live virtual environment consisting of six hosts, thirty-two powered on virtual machines and eleven data stores.

Figure B

The high-level environment overview. (Click the image to enlarge.)

In the center of the screen, you’ll see an overview of the physical and virtual resources available in this environment. Below the overview, you’ll find these three columns of information:

• Performance Problems. How many VMs are insufficiently configured or suffering from high I/O latency problems?
• Available Capacity. Given current trends, how many more VMs can be safely added to this environment before resources are taxed beyond a reasonable level?
• Over-Allocated Resources. How many VMs have been assigned too many resources, including too much CPU, over-allocated RAM, or an overly generous amount of disk space? Correcting these issues helps to increase the available capacity, which will lead to higher numbers in the Available Capacity column.

Because it’s a free tool, Capacity View does not notify you about specific actions to take to free up capacity or tell you which VMs you need to add RAM to in order to fix performance problems; VKernel’s other tools, including Capacity Analyzer and the Optimization Pack, provide those details.

Summary

Capacity View identifies where you might be experiencing problems in your virtual environment; this free tool is intended to entice you to purchase the more advanced VKernel tools: Capacity Analyzer and Optimization Pack. Capacity View does a good job of telling both that you might be giving away the farm to some VMs and that you might be being stingy with others. This kind of detail is the first step in your VM right-sizing effort.

Want to keep up with Scott Lowe’s posts on TechRepublic?

• Automatically sign up to the Servers and Storage newsletter
• Subscribe to the Servers and Storage RSS feed
• Follow Scott Lowe on Twitter

The Oracle Application Users Group released the results of a study last week that revealed 87% of the respondents blame data growth for their performance issues. As I opined recently, I think it could be very good for IT if users culled data that they don’t need in order to reduce stress on storage and backup infrastructures. A recent PC World article suggests that users who don’t do such cleaning regularly may be costing their companies a lot of money.

On the infrastructure side of things, there are a lot of technologies aimed at trying to help organizations spend less on storage. Deduplication technologies remove redundant information on storage devices so that things like operating system files or presentations that exist on many users’ home drives will only take up space once. Multi-tiered storage allows the most critical data to be stored on high-speed, expensive hardware while less crucial or less frequently used data resides on slower, cheaper hardware.

Unfortunately, the IT-centered solutions leave us with the same problem: Data growth is explosive and nearly unchecked in all industries. Granted, there are many good reasons to keep a lot of this data, including regulatory requirements, files that need to be quickly accessible, and files that are accessed frequently. However there are many files, particularly on users’ desktop PCs, that are simply irrelevant, old garbage that should be treated as such. In the business world, there is too much work to do for us to spend the necessary time cleaning our data.

In order to be a help, I am posting my strategy for getting rid of unnecessary files in bulk without the danger of losing any data. This is what I educate my users to do, and we have been able to forestall increasing server space simply through reducing file volume on our servers.

1. Get a CD or DVD burner for your PC.
2. Burn all of your data files to the burner you acquired.
3. Go through the media you just burned to make sure the burn was successful.
4. Delete everything you just burned from your PC.
5. Browse through the media you just burned and copy back to your PC only the files you absolutely know you will need in the next day or two.
6. Keep the media in your drive for a few weeks and, when you need a file that is on the media, copy it back to your PC.
7. Label the disc with the date (I do this yearly, so I label mine “Clean - 2009″).

For Outlook, the process is a little different. I keep a year of email in a PST on my desktop. Every year, I take the previous year’s PST, archive it to CD or DVD, and close the PST in Outlook. So, at the end of 2009, I archived all of my 2009 email, created a new PST for 2010, burned my 2008 email, and then deleted the 2008 PST.

Using this strategy, I still have access to my old email and files if I need them, but they aren’t taking up space on my hard drive, a network drive, or any backup medium. As a result, my inbox is a little over 13 MB, last year’s PST is just under 500 MB, and My Documents is under 50 MB. I am also secure in the knowledge that if someone shows up needing an email or file from two years ago (when I started my current job), I have access to it.

What data spring cleaning best practices do you recommend? Do you think expecting users to clean up their own data is realistic? Share your tips and your thoughts in the discussion.

TechRepublic’s Servers and Storage newsletter, delivered on Monday and Wednesday, offers tips that will help you manage and optimize your data center. Automatically sign up today!

Version 12 of the Sysinternals Process Explorer diagnostic tool has new features that allow admins to have even more visibility into Windows Server processes.

—————————————————————————————————

Process Explorer, which is arguably one of the most frequently used Sysinternals tools, has been updated to version 12. The latest version includes a number of new features that allow server administrators to see what program, service, or system process is interacting with a file. According to this TechNet blog, the update also includes a number of other minor improvements and bug fixes.

The significant new features for the popular diagnostic tool include the following:

• Show Web hosted Internet Explorer 8 processes
• Display of svchosts’s service category
• TCP/IP information for properties of a process
• Service name mapping to running threads

Process Explorer has a lot of tabs and views, so it may take you a while to find the new features. Figure A shows the new feature that displays the TCP/IP information of a given process.

Figure A

Click the image to enlarge.

Figure B shows an example of the new feature that will display a Windows service name to a running thread. This feature detects if a thread has been started by a service.

Figure B

Click the image to enlarge.

In Figure B, the two processes on the left (vpxd and JettyService) were originated from a Windows service and have the Service tab visible. On the right, the user process (Firefox Web browser) does not have a Service tab.

You can install Process Explorer version 12 via a free download from the Microsoft site. This update of Process Explorer works for Windows Server 2003 and higher. It also works on Windows XP and higher. This tool continues to hold the crown of most useful in my book.

Process Explorer is the most useful tool that I use in my server admin work. What do you think of Process Explorer? How do you use it? Share your thoughts with the TechRepublic community.

Stay on top of the latest Windows Server 2003 and Windows Server 2008 tips and tricks with our free Windows Server newsletter, delivered each Wednesday.
Automatically sign up today!

Scott Lowe provides an overview of the Management Server installation process for Windows Essential Business Server 2008.

—————————————————————————————————-

Windows Essential Business Server 2008 (EBS) is Microsoft’s answer to midmarket technology needs. While overall management of EBS is simpler than it is for larger organizations that rely on vanilla Windows, the installation process consists of a number of steps.

In part one of this two-part series on the installation process for EBS, I explained how to run the Planning And Configuration wizard, which is the initial process necessary before moving on to the step that is discussed in this column: deploying the Management Server.

Note: This tutorial is also available as an image gallery.

Begin the installation

Although it has streamlined administration tools, EBS is really just Windows Server 2008 with a set of management tools designed to simplify common tasks; as such, the installation process for EBS begins like it does for any Windows Server 2008-based system: Insert the installation media into the optical media drive on the server and power on the system.

What you see at the very beginning of the installation process is identical to what you see when you install a plain vanilla Windows Server 2008 system, so I won’t go over that part here except to note that you simply need to choose an installation method, which should be Custom since this is likely to be a new server rather than an upgrade. Perhaps the most notable difference is the lack of an edition selection page during installation. Whereas a plain Windows 2008 installation asks you to choose the edition you’re installing — Web, Standard, Enterprise, Data Center — EBS does not ask you to choose.

Management Server Installation wizard

The real installation difference comes into play after the base Windows 2008 system is fully installed. Once that happens, the system starts up and the Management Server Installation wizard begins, starting with the Welcome screen (Figure A). Click the Next button.

Figure A

The Management Server Installation wizard’s Welcome screen. (Click the image to enlarge.)

With the preliminaries out of the way, the wizard asks you to decide which network adapter you want the Management Server to use for network communications. As you see in Figure B, my lab’s virtual machine has a single network adapter from which to choose. If your server has multiple network adapters, all of them would be listed here. After you choose a network adapter, click the Next button. Once you do, the wizard queries the selected network adapter to gather information; then, you’re moved to the next installation step.

Figure B

Choose the network adapter to use for communication. (Click the image to enlarge.)

The installation wizard needs periodic network access in order to achieve its goals. For example, during the process, the wizard will look for updates that improve the installation process. In order to facilitate this communication, the wizard needs to assign a temporary IP address to the network adapter; this address can be assigned automatically via DHCP, or you can opt to manually assign an IP address (Figure C).

Figure C

Choose your temporary IP address assignment method. (Click the image to enlarge.)

With IP addressing out of the way, the installer moves on to ask if you’d like to download updates during installation. In order to maintain a secure system, I recommend selecting the Download And Install Optional Microsoft Updates During Installation (Figure D); once you do, and after you click the Next button, the installation wizard looks for and downloads necessary updates from Microsoft. (This step can take a few minutes.) Once the installation wizard is updated, the system restarts and the installation continues.

Figure D

Decide whether to update Windows during installation. (Click the image to enlarge.)

Regardless of your selection on the Microsoft Update screen, the installer requires you to download critical updates that might affect the installation wizard (Figure E).

Figure E

Installation wizard critical updates are installed. (Click the image to enlarge.)

If you’re installing the management server for the first time, there is a high likelihood that you have yet to create your Active Directory domain.  On the next page of the wizard, you’re asked if you still need to create a new forest and domain or if you plan to join this server to an existing domain (Figure F). In my case, I needed to create the domain, so I selected the Create A New Forest And Domain option. Click the Next button.

Figure F

Choose whether to create a new domain or join an existing one. (Click the image to enlarge.)

If you’re wondering if all of the work you went through during the planning phase was for naught, you’re about to find out. When you get to the next page of the wizard, you’ll see the Load Planning Data button (Figure G). At this point, you need to make available whatever media on which you stored the planning data from the first phase of the installation. In my lab, I added a second virtual CD-ROM drive to my virtual machine and pointed that drive to the location where my planning data was stored. You can use a disc or a USB key to store the planning data, or you can store the information at a network location. The goal is to point the wizard at the planning file that you created in the first phase.

Figure G

Locate the planning data. (Click the image to enlarge.)

You’ve already decided that you need to create a new domain, so now you need to name it and define the password that will be used for the domain administrator’s account (Figure H). After you provide the requested information and click the Next button, the wizard scans the network to make sure that the domain name is not already in use.

Figure H

Provide a Domain Name and an administrator Password. (Click the image to enlarge.)

The names you use for your EBSs are important and are defined on the next page of the installation wizard. I named my management server ebs-mgmt, my security server ebs-sec, and my messaging server ebs-mail (Figure I). After you click the Next button, the wizard verifies your server names.

Figure I

Provide the names for your EBSs. (Click the image to enlarge.)

IP addresses make network communication possible. When you get to the Assign IP Addresses screen, provide the IP addresses that will be used for each of your services (Figure J). Click the Next button.

Figure J

Provide the IP addresses for each of your EBSs. (Click the image to enlarge.)

The installation wizard knows what you want — it even presents a list of internal network IP addresses for your consideration (Figure K). You can choose to accept this list of IP addresses (which I opted to do) or make changes. Click the Next button.

Figure K

Choose how you want to define internal IP addresses. (Click the image to enlarge.)

The Management Server can also act as the DHCP server for your network, or you can choose to continue using your existing DHCP service, as you can see in Figure L. In my lab, I didn’t use EBS’s DHCP server since I already have one, but in a real-world scenario, I highly recommend that you allow EBS to handle DHCP duty in order to facilitate smoother network communication. If you were to choose the EBS-based DHCP service, you would be presented with two additional DHCP configuration screens (Figure M and Figure N) on which you would be asked to configure DHCP settings, including the name of a DHCP scope, the starting and the ending IP addresses, and the subnet mask and gateway addresses for the new scope.

Figure L

Decide how DHCP should be handled on your network. (Click the image to enlarge.)

Figure M

Provide details for a new DHCP scope. (Click the image to enlarge.)

Figure N

The new DHCP scope also needs DNS and address duration information. (Click the image to enlarge.)

File storage continues to be one of the most basic uses for servers; your new EBS needs a place to store its files, and it’s up to you to define a volume to use for this purpose. If you have only a single volume in your server, you won’t have the option to choose any volume other than the system volume (Figure O). If you know that you have other volumes in your server, click the Disk Management button to initialize and format the volume. Click the Next button.

Figure O

Choose the volume on which data should be stored. (Click the image to enlarge.)

EBS includes the Remote Web Workplace (RWW) feature, which provides users with a portal environment used to access resources remotely. The Company Profile page of the wizard (Figure P) asks you to provide some information about your company; this information will be used to generate an SSL certificate for use with RWW.

Figure P

Provide company profile information. (Click the image to enlarge.)

The wizard also includes questions about whether you’d like to participate in Microsoft’s error reporting service, which automatically sends crash reports to Microsoft (Figure Q), and in the company’s Customer Experience Improvement Program, which sends actual usage information to Microsoft to help it develop future versions of Windows (Figure R). Many company security policies prohibit the sending of this kind of information, so choose carefully.

Figure Q

Do you want to automatically send crash reports to Microsoft? (Click the image to enlarge.)

Figure R

Does your company want to participate in the Customer Experience Improvement Program? (Click the image to enlarge.)

After you review your selections, click the Install button (Figure S). You’ll see a progress screen (Figure T) that shows how far along the installation has progressed. The installation can take hours. Also, your server may reboot multiple times during the installation. After each reboot, the installer picks up where it left off.

Figure S

Review your selections and click the Install button.

Figure T

A progress bar lets you know where things stand. (Click the image to enlarge.)

At the end of the installation, you’ll see the Continue Installation screen (Figure U), which indicates that the Management Server has been installed but that you still need to deploy the Security and Messaging Servers.

Figure U

It’s time to move on to the next server installation. (Click the image to enlarge.) Want to keep up with Scott Lowe’s posts on TechRepublic?

• Automatically sign up to the Servers and Storage newsletter
• Subscribe to the Servers and Storage RSS feed
• Follow Scott Lowe on Twitter

Windows Server 2008’s DHCP engine has a filter that can reduce risk of rogue devices. IT pro Rick Vanover discusses this new feature.

———————————————————————————————–

One of the least glamorous parts of IT is tracking down a rouge system that received a DHCP address, especially if it is causing an issue on the network. IT pros have become good at looking at a MAC address and determining what type of system it may be. For instance, users may bring a wireless router from home into the office or install a virtual machine without any protections on the network.

Windows Server 2008 R2’s DHCP engine introduces a MAC address filter engine. The filter is pretty cool; it allows you to specify wildcard MAC address ranges to allow or deny address assignment on the network.

For example, take the requirement to prohibit Hyper-V, VirtualPC, or VirtualServer virtual machines from receiving IP addresses on your network. Figure A shows how you can do this with the filter.

Figure A

Click the image to enlarge.

For the specific case of virtual machines, I created a scorecard of the MAC address types and the associated hypervisors.

Clearly, the DHCP filters are not bulletproof. Most hypervisors and wireless devices let users spoof MAC addresses. But chances are, this will knock out most of the users who could potentially do the least desirable things on your network. In the case of wireless devices, determining the organizationally unique identifiers (OUIs) for Linksys, Netgear, D-Link, and other products may be a good idea if you have a problem with unauthorized devices showing up on your network. A more strict approach is to set a filter only for the devices you expect to use on your network.

I think this is pretty neat for an otherwise boring service. Do you see yourself using this new feature? Let us know in the comments.

Stay on top of the latest Windows Server 2003 and Windows Server 2008 tips and tricks with our free Windows Server newsletter, delivered each Wednesday.
Automatically sign up today!

Red Hat has officially entered the desktop virtualization market. Red Hat Enterprise Virtualization (RHEV) for Servers will allow customers to deploy Linux, Windows XP, or Windows 7 clients virtually. This takes the processing that is normally done by the client PC on the user’s desk and moves it to the data center, where high-powered servers can process much more quickly.

Desktop virtualization market

Desktop virtualization is still in its infancy as far as the market goes, but it is reminiscent of the mainframe days when the processing was done on a huge machine in a data center, while the users had “dumb terminals,” or screens that displayed the data that was served up by the mainframe. These days, most people are more comfortable with the computers on their desk doing the majority of the work, while enterprise applications are run off of servers in the data center.

Benefits
Some of desktop virtualization’s benefits include the following:

• Client machines can be extremely scaled down thin clients with only the processing power necessary to display the screenshots that are sent down from the server.
• Users generally see performance benefits, as the machines running their desktop OS are servers. If the planning process is done properly, those servers can blow away the performance of all but the highest end user workstations.
• Virtual desktop environments are far easier to keep patched and upgraded. The environments also make software deployments a breeze since the updates only have to be applied to the server.
• Remote users and travelers have the ability to access their desktop from any machine capable of running a remote desktop application, from a laptop, a client’s office, or an Internet kiosk in a library or an airport.

Major drawback
Desktop virtualization hasn’t received much traction because, if all a user has is their virtual desktop, a network or server outage can stop an entire enterprise in its tracks. However, it is getting to the point that desktop virtualization’s benefits seem to be outweighing the infrastructure concerns.

Red Hat vs. the competition
There are established players in the desktop virtualization space, and they might make it tough for the Linux giant to make progress.

For instance, Microsoft is in the virtual desktop market, having teamed with Citrix to form a roadshow, which they are hoping will convince customers to choose their offerings over market leader VMware, which also has a desktop virtualization initiative.

Red Hat’s offering has one compelling feature that its competitors’ products lack: the ability to import virtual machines from VMWare, Citrix, and Microsoft’s virtual platforms. Red Hat senior product marketing manager Andrew Cathrow claims that moving virtual machines from the other platforms to RHEV is a “very simple” process.

Red Hat has also increased the resources available to each virtual desktop in order to make large processor and memory intensive applications perform better for the users. In addition, the management suite includes a data warehouse that can be exported to most SQL-based databases. The icing on the cake is the pricing, which includes a free hypervisor for customers already running RHEL, as well as management tools that Red Hat claims are “dramatically lower-cost than our competitors.”

My thoughts on desktop virtualization

I have long been a proponent of desktop virtualization. I remember pining for a technology that would allow me to patch or upgrade the operating system for my users without having to visit every machine. I vividly remember upgrading our campus to Microsoft Office 2003 with Group Policy and having to physically visit every single PC to install Windows XP and being supremely disappointed with the hoops I had to jump through to get things done. Desktop virtualization can ease many of those problems, but you must have solid infrastructure components in place because a network failure could result in your users being unable to get to their desktop.

Is your company experimenting with desktop virtualization? If not, does it plan to in the near future? Share your thoughts on desktop virtualization in the discussion.

TechRepublic’s Servers and Storage newsletter, delivered on Monday and Wednesday, offers tips that will help you manage and optimize your data center. Automatically sign up today!

Serving a dynamic campus community out of a repurposed classroom in a decades-old building has its challenges. Scott Lowe describes his team’s efforts to revitalize a college’s neglected data center.

—————————————————————————————————-

Westminster College’s IT infrastructure has grown throughout the years in a very organic way. In the beginning, it was only a single VAX terminal connected to the University of Missouri; now it’s a fiber-connected network with thousands of nodes, dozens of servers, and terabytes of storage.

I’m very happy with where Westminster College is headed in terms of infrastructure and architecture. However, one area that has not been up to par is the physical location where we house our infrastructure. We’ve spent time fixing up what used to be a basement classroom in the campus administration. I wish I’d had the forethought to take pictures of what the server room looked like a while back, but I didn’t; I did take a few “after” photos. In this post, I describe what improvements our IT team made and explain why we made those changes.

Details about the “before” server room

Our server room is a repurposed basement classroom that was granted to IT well before my arrival at the college. The room measures about 15 x 15, has a concrete floor, and has a drop ceiling that is a few inches below the floor structure supporting the floor above.

When I started at Westminster, the room looked like the diagram in Figure A. This isn’t a scale drawing, but you can see where the various components were placed.

Figure A

The old server room layout

There were three server racks (none of which were full) that housed the college’s couple dozen servers. The campus fiber optic network at the time was all multimode and terminated in the network core you see on the diagram. The blue box represents the cabling for the building housing the server room. All of that cabling is terminated into a wall mount rack and feed the systems and users that work in Westminster Hall, the building in which the data center resides.

As you can see, there were three lights in the room. The lights provided enough light to work by but left some darkish spots.

The two green boxes on opposite walls of the room were patch panels that were connected to one another across the room with the cabling running above the drop ceiling; this was the cabling used to connect the servers to the core switch. Without a raised floor, this was a decent alternative, but it didn’t provide for much growth and working on it was a pain.

The A/C unit in the back of the room is a standalone unit with four air vents blowing out the top of the unit and pointing in various directions to cover as much of the room as possible.

The room has no raised floor, and it never will — it’s a basement with a concrete basement floor. Although we could jackhammer out enough material to create a raised floor, it’s not really necessary anymore.

The room had a lot of problems, including:

• Inadequate air flow. The Dell servers we use blow air from front to back; there was about a foot of space between the rear of the racks and the wall, and it was hot back there! Plus, cold air wasn’t being delivered where it was needed most.
• Inadequate lighting. As I mentioned, some areas of the room simply were not lit well enough.
• Really difficult to work on systems. There was very little space between the rear of the racks and the walls, making it very difficult to work behind the systems.
• No flexibility to expand network core. Without additional racks, we didn’t have enough space to extend our fiber optic network. As a part of a new building project, we ran 288 strands of single mode fiber cable across campus and needed a place to terminate that new cabling.
• The “blue box” that was the building’s internal cabling was a mess. The word spaghetti doesn’t come close to describing the morass of 150 patch cables run from that blue box to switches in the “network core” area.
• UPSs that were almost at capacity.
• No ability to easily disconnect and move the UPSs.

Improvements to create the “after” server room

Throughout the past few months, as time has permitted, my data center guy and I made a bunch of changes designed to correct some of these problems. Figure B is a look at the new layout.

Figure B

A newly fixed up server room

Moved server racks after eliminating one
First, we eliminated one server rack and turned the remaining two server racks (Figure C and Figure D) so the rear air flow was not impeded by a wall. This gives us ample room to work behind servers without worrying about claustrophobia setting in.

We were able to eliminate a full rack by following these steps:

• Virtualizing older, larger, more power-hungry systems to fewer smaller blade-based systems with a SAN backing the entire thing. This has reduced the number of physical servers and reduced power consumption to a point where our UPSs are happy again.
• Not leaving 1, 2, or even 3U of space between our systems anymore. There’s no point in that for servers with a front to back air flow, so why waste the rack space?
• Moving our backup server and tape library to another secure campus location where it has a direct fiber connection back to the core.

As we walk by the sliding glass door (yeah, a sliding glass door — it’s not my favorite entryway, and it’s something we’ll change in the future), we can now look for bad lights at a glance.

Figure C

Figure D

Improved cold air distribution
In Figure B, you’ll see green lines and boxes connected to the A/C unit (you can see the unit in Figure E); these are duct extensions that we installed in order to drop cold air where it’s needed rather than just allowing cold air to blow randomly around the room. This move has already made an impact. Our Dell M1000e blade chassis fans used to run at a very high rate to keep the chassis cool. With this move, the fans run noticeably slower, meaning that the unit is having an easier time staying cool.

Figure E

Added two new network/cabling racks
In order to accommodate a large, new single mode fiber installation, we added a new rack (Figure F) to the left of the existing network rack and moved the core switch and all the network electronics to the new rack and left the older rack in place to house fiber optic terminations for the campus network. We’ve added quite a bit in the way of network gear, and we don’t need to worry about having enough space to house it now.

Behind the existing network rack, we’ve added a new rack (Figure G) intended to house all of the cabling for the building itself and get it away from the core network switch. The 150 patch cables that run into the core network rack seriously get in the way and are a pain in the neck to manage. We added the new rack when we renovated the IT office space across the hall and needed to re-run all of the network cabling into the space. This summer, we’ll extend the existing patch panels to new patch panels in this common building rack and move the switches that service the building to this rack in order to be able to more easily work on the core switch. Although we don’t have to do a lot with the core network rack, we want to make it easy work on it when the time comes.

Figure G

Added lighting
We’ve added a couple of lights to darkened areas in order to improve our ability to work. It’s amazing what the simple addition of a couple of lights can do.

Added a cable tray
If you look carefully at Figure B, you’ll see a sort of thatched area on the diagram; this denotes a new suspended cable tray (Figure H) that we added at about 7′ high in the room, or about 6″ below the ceiling. We’ve also removed the patch panels that previously served the servers and instead laid the server connecting patch cables into this easy-to-reach cable tray. This step has made it much easier to make changes and add new cables as necessary; it has also made tracing cables a lot easier.

Figure H

Cut the cord on the UPSs
In order to make it possible for us to more easily do routine maintenance on the UPSs and to move them if it became necessary, we had an electrician install massive power plugs in the end of the UPS cord. Prior to doing this, the UPSs were hardwired right into the electrical panel, so maintenance was more difficult.

Summary

We didn’t make radical changes to the server room, but by correcting what were some relatively serious problems (particularly with air flow), we did create a very functional space. Westminster College is looking at the possibility of a new academic building, which may afford us the opportunity to build a true data center down the line.

Want to keep up with Scott Lowe’s posts on TechRepublic?

• Automatically sign up to the Servers and Storage newsletter
• Subscribe to the Servers and Storage RSS feed
• Follow Scott Lowe on Twitter

For Windows Servers, it is sometimes necessary to configure the number of processors in use by the OS. Learn how and when to use the configuration tool.

—————————————————————————————————

Most versions of Windows Server have the System Configuration Tool, which shows a number of configuration items for the local system. I’ve used this utility to step-down processors on a Windows Server.

System Configuration (or msconfig.exe) has a Boot tab that allows an interactive edit of the Boot.INI file. By clicking the Advanced Options button, you edit the number of processors that are visible to the operating system on boot. In the following example, the server has one physical processor with four cores. Figure A shows the system running the System Configuration Tool.

Figure A

Click the image to enlarge.

Here I can step the system down from the current inventory of visible processors (four) to any number less than that.

This is useful for one main reason: determining if lesser performance is acceptable on a system to conserve resources. In a virtualized environment, this is pretty straightforward, as you would provision processors to the virtual machine upward or downward. For physical systems, this can save unnecessary handling of the internal parts of the server and sensitive processor components. By using the System Configuration Tool, you can quickly step down to a fewer number of processors on the system via a soft configuration. The System Configuration Tool will not distinguish between sockets and cores, however; you may need to intervene to the server’s BIOS to limit cores or hyperthreading.

Note: In most situations, using the System Configuration Tool to specify a limited number of processors is not an acceptable licensing workaround for per-processor situations. Consult your licensing professional for guidance on how to downsize processors for a system if necessary.

Check out a few of the System Configuration Tool’s other options. While the processor limit is not very fancy, other options such as startup services and applications are available. This is a good first-line tool if the more robust Autoruns SysInternal tool is not installed on the server.

Stay on top of the latest Windows Server 2003 and Windows Server 2008 tips and tricks with our free Windows Server newsletter, delivered each Wednesday.
Automatically sign up today!

Most organizations use some form of tape media when it comes to data protection; however, some companies exclusively use data protection by storage systems. Now that the LTO-5 tape format is available, will that change how your organization views its data protection strategy?

—————————————————————————————-

This week, I am at the HP StorageWorks Tech Day event in Houston. The first topic we discussed is HP’s announcement that it is offering LTO-5 tape devices. LTO-5 is the natural progression of the popular tape format. LTO-5 has a native storage capacity of 1.5 TB and a compressed capacity of 3.0 TB. While it isn’t exactly a surprise that LTO-5 is now available, how does tape fit into most organizations?

There are arguments for and against tape. During our sessions, a number of IT professionals said they are performing data protection exclusively via tape. For example, storage professional Devang Panchigar of StorageNerve still doesn’t see the need for tape. Among the key arguments for tape is the ability to, in a way, remove the human element. One situation where storage-based data protection (such as volume replication) can be flawed compared to tape is deleting a file on one end and being immediately deleted on the other. Compounding features on storage can hedge down the “advantages” of tape compared to disk-based data protection. These features include volume snapshots on the storage system as a way to roll back to a point in time. Further, backup software such as Backup Exec, Veeam Backup, and many others can protect data by maintaining large archives of systems and data.

When it comes to architecting data protection without tape, IT professionals have options. My preference is to protect data without tape; this can be done using backup software that creates large archives of protected systems and protecting that on disk that may be replicated. The end result is that every organization needs to apply their data retention requirements to their protection strategy.

What is your take on tape-based data protection? Does LTO-5 open some doors for your options? Share your comments below.

Disclosure: The event organizer covered my tickets, meals, airfare, and accommodations. The opinions stated are the result of an in-person demonstration of the technologies discussed. Read my full blogger disclosure.

TechRepublic’s Servers and Storage newsletter, delivered on Monday and Wednesday, offers tips that will help you manage and optimize your data center. Automatically sign up today!

Windows Server 2008 R2 installations create a default partition that is not visible to the OS. Rick Vanover explains what this allocation does for Windows.

———————————————————————————————–

When it comes to disk provisioning, in a general sense, Windows administrators should always provision what is required and have the ability to expand if that need increases. To that end, you might wonder what the 100 MB of System Reserved Partition space is doing for your servers.

If you check the default installation of Windows Server 2008 R2, you will see a partition like the one in Figure A.

Figure A

Click the image to enlarge.

While 100 MB isn’t much, you still don’t know what the partition is simply by looking at it. What makes it somewhat more interesting is that the Share And Storage Management console (a new tool for Windows Server 2008) does not report this partition. Figure B shows the same disk in this other view.

Figure B

Click the image to enlarge.

I researched why this partition is there, and I learned that it is to prepare the server for BitLocker — if it is configured and it functions as a boot loader. (Note: If a drive will receive a new Windows Server 2008 install, yet there is already a boot loader in place for one or more operating systems, System Reserved Partition may not be created.)

On the System Reserved Partition, there is nothing much of interest. Figure C shows Windows Server 2008 R2’s reserved partition viewed through a recovery tool.

Figure C

Click the image to enlarge.

Don’t try to delete System Reserved Partition, as BitLocker won’t work if it is added at a later date. Allocating 100 MB of disk space is worth it if you decide to add this feature.

Stay on top of the latest Windows Server 2003 and Windows Server 2008 tips and tricks with our free Windows Server newsletter, delivered each Wednesday.
Automatically sign up today!

Every environment has the one-off system that is just different. Rick Vanover shares tips on how to support and administer specialized Windows Servers.

———————————————————————————————–

Many software and hardware providers are moving their proprietary technologies to Windows development platforms. The end result is that IT departments may see Windows Server systems that used to hold the platform of a “black box” system.

I don’t know if we should label it a “black box” Windows Server, but there is an entire series of specialized editions of Windows available to OEMs called Specialized Server Solutions. The most popular version of this technology is the storage server line of products. Other popular implementations of this specialized version of Windows include appliance-like functions such as document imaging systems, teleconferencing systems, medical imaging, and many others. I’ve even seen this at my dentist’s office and a parking ticket payment system. In the course of being a Windows Server administrator, I’ve come across a few new systems that have a Windows Server Appliance Edition installation. Just like it catches your eye when you see Web Edition or Storage Server on the Windows splash screen, this version makes you take a second look.

When dealing with these specialized versions of Windows, there are three primary issues to consider:

You may need to rethink your Windows Update and patch management policies. In most situations, these specialized versions of Windows are an original equipment manufacturer (OEM) build. In my Windows administration practice, patch management for these custom builds of Windows would need to be coordinated with the OEM. This can be a simple as an approved patch list or as complicated as a series of regression testing or as scary as you being on your own.

You will need to rethink your network security. There may be a special service running on a network port that is a required function of the application. This can affect how you go about centralized management policies such as Group Policy, Windows Firewall, antivirus, and malware protection.

You will need to be very careful when adding Windows Server roles or features during the installation process. If anything needs to be retrieved from the original installation CD, it is critical not to mix different versions of Windows from the OEM distribution. This can be a big problem if a standard issue version of Windows Server is provided to add a component (such as IIS or SNMP) to a customized build of Windows.

Are you lucky enough to have an Appliance Edition installation of Windows Server? If so, tell us about the changes you have made in supporting these servers?

Stay on top of the latest Windows Server 2003 and Windows Server 2008 tips and tricks with our free Windows Server newsletter, delivered each Wednesday.
Automatically sign up today!

How do consumer-grade solid state drives compare to enterprise-grade solid state drives when it comes to cost, performance, and reliability? Scott Lowe answers that question.

—————————————————————————————————-

I’ve written about how you can measure the overall performance of your storage array and how you can break down performance into a dollar amount — that is, a cost per Input/Output Operations Per Second (IOPS). In the cost per IOPS article, I used the example of a 400 GB solid state disk that cost a whopping $23,000 (that’s each). With a cost of $58,750 per total terabyte of solid state drive (SSD) storage, each SSD provides 6,000 IOPS of performance capability. From a pure IOPS/dollar perspective, the SSD was a clear winner in my calculations, but from a total cost perspective, it was very prohibitive.

Vendors have begun to release more affordable, but smaller, SSDs; Intel, Crucial/Micron, and OCX have released enterprise- and consumer-grade SSD products. Some of the performance data that is associated with these drives needs to be carefully analyzed to make sure that actual use cases are being tested against.  In the following video, Micron’s product was tested under stellar circumstances; namely, the tests used an empty SSD and random 4K reads and writes, which will always provide better performance. In fact, on a 6 Gb SATA channel, the tester achieved in excess of 60,000 IOPS of read performance and about 48,000 IOPS of write performance. In contrast, Intel’s X25-E Extreme enterprise-grade SSD gets 35,000 read IOPS and 3,300 write IOPS of performance according to the product’s technical data sheet.

In this high level overview, I look at four SSDs that are much more affordable than the EMC unit I presented in the previous article. In Table A, I list one enterprise-grade SSD and three consumer-grade (but “RAIDable”) SSDs. By RAIDable, I mean that, theoretically, you could lace these SSDs into a RAID 1/10, 5, or 6 array to gain the data protection benefits inherent in these configurations. Click each of the model numbers to get a look at the source information I’ve used for the Read and Write IOPS values and for the Mean Time Between Failure information.

Table A

Manufacturer

Intel

Crucial

Kingston

Kingston

Model

X25-E Extreme

RealSSD C300

SSDNow V+

SSDNow V+

Grade

Enterprise

Consumer

Consumer

Consumer

Capacity (GB)

64

256

256

64

Price/Unit

$750

$800

$700

$200

$/TB

$12,000

$3,200

$2,800

$3,200

Read IOPS

35,000

60,000

6,300

6,300

Write IOPS

3,300

48,000

84

291

MTBF (hrs)

2,000,000

1,500,000

1,000,000

1,000,000

Four affordable SSDs

The Intel X25-E 64GB SSD is quite expensive when compared to the three other selections. However, the Intel device is targeted at the data center, which is apparent when you look at the MTBF information. The Intel device has a 2 million hour MTBF value, while the other SSDs in the table have 1.5 million and 1 million hour MTBF values; so you can expect that the Intel drive will last longer, which partially justifies its higher price. Intel also has a consumer grade line — the X25-M and the X18-M — which carry 1.2 million hour MTBF values, as well as much lower price tags ($449 for a 160 GB SSD, or just under $2,900 per TB).

With the exception of the Intel enterprise-grade SSD, each of the consumer-grade SSDs runs about the equivalent of $3,000 per TB. That’s not a bad per TB cost if you’re willing to risk running a consumer-grade drive in production.

Frankly, the IOPS values outlined for some of the drives are out of sight. I don’t have one of each model drive to independently verify the values, so attempting to divine an IOPS per dollar value for these drives would be pretty ridiculous and pointless. That said, if you compare the values in my previous article in this series, you’ll note that the cost per IOPS would be much lower for all of the drives listed in this article, particularly for the Intel and the Crucial offerings, assuming those IOPS values hold up in real-world usage.

There is also the issue of the RAID controller itself. I assume that, if you’re looking for cheap storage, you’re probably not going to stick these disks into an expensive SAN; you might even be attempting to build your own storage device. Can you even find a reasonably priced RAID controller that can support a bunch of these disks? If your goal is to get massive IOPS on the cheap (read IOPS, anyway — the Kingston write IOPS are surprisingly low), you probably want to avoid dropping $20,000 on the controller, if you can even find one that supports massive throughput. Read an article that explains the situation. Once you go past a certain throughput level on individual drives, the PCI X8 slot into which a RAID controller is connected simply can’t pass more data, so there is little point to continuing to add drives. In these cases, consider multiple smaller arrays instead, with each one connected to a separate RAID controller.

I’m providing the information in this article primarily to satisfy a curiosity. If you have an application that truly needs the IOPS levels in Table A, you probably also have needs that go well beyond a desire to run consumer grade storage in a RAID array in your data center. That said, the Intel device presented is an enterprise grade device with a much lower price tag than the EMC 400 GB SSD discussed in my previous post.

What we’re seeing in the SSD market is encouraging. Hopefully, prices on enterprise-grade gear will continue to come down and eventually reach the consumer-grade levels I reference here.  When that happens, the days of storing data on magnetic platters spinning at insane speeds inside a tiny little box will come to an end, and we’ll look back on today’s hard drive and wonder how we ever thought that was a good idea.

Want to keep up with Scott Lowe’s posts on TechRepublic?

• Automatically sign up to the Servers and Storage newsletter
• Subscribe to the Servers and Storage RSS feed
• Follow Scott Lowe on Twitter

Our organization uses SharePoint 2007 in production for document storage and management, shared calendars and task lists, blogging, wiki libraries, and discussion boards. I’ve been keeping up with the news about SharePoint 2010 so I’ll know what to expect when we’re ready to upgrade.

The beta version of SharePoint 2010 is available for download. It’s rumored that Microsoft plans to release the newest version of the business productivity suite to manufacturing next month. The final release of SharePoint 2010 and Office 2010 for business customers will be May 12, with the wider retail release set for June.

SharePoint 2010 was put through the paces during the 2010 Winter Olympics and appears to be ready for prime time. Also, if you have Software Assurance, Microsoft claims that your upgrade may be paid for already.

Since not many technologies compete with SharePoint, the result is slower innovation and changes that are more cosmetic than substantive. For instance, SharePoint 2010 includes a ribbon much like the one in Office 2007 applications, which allows faster access to the most commonly used commands, but it doesn’t seem to have many additional features. I was hoping that the collaboration suite would have Silverlight and RemoteFX technologies to allow easier audio/video conferencing and whiteboarding (it feels cumbersome to have to open LiveMeeting for these functions), but I suppose we will have to wait for another version.

Although SharePoint 2010 looks quite useful, I would recommend to anyone who might ask to wait until SharePoint 2010 is out of beta for installation in a production environment.

Windows Server 2008 R2 SP1 may be a boon for server sales

The timing for SharePoint’s newest version works out well for Microsoft, which is rumored to be in the planning stages for its release of the first service pack to its latest version of its flagship server OS. Windows Server 2008 R2 SP1 will surely drive new server sales as a result of Microsoft abandoning 32-bit architecture altogether in favor of the 64-bit version (read 10 reasons to consider upgrading to Windows Server 2008 R2). Don’t underestimate the need for upgrades — the recent economic woes have led many technology executives to put off upgrades; for instance, I recently wrote about an Intel executive saying that a third of the processors in operation today are over four years old.

My gut tells me that server purchasing will be ramping up over the next few months since it looks like budgets will free up a bit as the economy continues to improve. Many companies may wait for Windows Server 2008 R2 to release before taking delivery, but the new offerings from Microsoft seem destined to increase demand for beefier hardware.

What are your organization’s plans?

Does your organization use enterprise content management software such as SharePoint? Will the new version drive hardware purchases for your business? Let us know in the discussion.

TechRepublic’s Servers and Storage newsletter, delivered on Monday and Wednesday, offers tips that will help you manage and optimize your data center. Automatically sign up today!

Data deduplication technologies have delivered on the promise of significant cost savings through backup data reduction and enlarged the scope of potential applications that can be protected effectively and affordably at central and at remote sites.

During the TechRepublic Webcast on Tuesday, April 6, 2010 at 2:00 PM ET / 11:00 AM PT / 6:00 PM GMT, Dell and EMC experts will review best practices for protecting data, including deduplication offerings from EMC’s Data Domain and Avamar product lines. They will also review some of the factors to be considered when assessing what data deduplication products might best fit your environment.

Register today for the Dell-sponsored TechRepublic Webcast, which will feature live Q&A.

Intel announced its new line of microprocessors on Tuesday, making claims of a five month return on investment and a 15:1 consolidation estimate. The new Xeon 5600 chip line is truly impressive and, according to the company’s analysis, a third of the servers currently in use are running processors made four or more years ago.

AMD is chasing the same upgrade cycle, but it’s going after it in a slightly different way than Intel. AMD’s upcoming strategy will be to change the dynamic somewhat by making more processors available by bringing down the price of four-processor servers. AMD’s bet is that it can get better performance per watt by increasing the number of processors in each server. If AMD is able to reduce the price jump between two- and four-processor servers, it may be able to increase the number of four-processor servers out there over the 5% where it is now.

Both companies are also adding features to their chips; Intel has added features that improve performance when dealing with encrypted files, and AMD is engineering a line that includes graphics processing for laptop and netbook markets.

AMD is hovering at about 9% of server chip sales and has hovered around 30% market share overall for the past few years; if AMD is ever going to seriously challenge Intel, it must find a way to tip the scales in its favor.

The last time I bought servers, I was very surprised at the big price and size difference between two- and four-processor machines; I remember the difference being substantial enough that a four-processor machine wasn’t even under serious consideration. If AMD can cram more processing power in a rack than Intel, its similar performance per watt numbers would give the company an edge it could use.

I have long been processor agnostic; the very first 486 I ever owned was AMD, and I would definitely consider AMD if I could get more processing into my data center or reduce the number of racks necessary to do the job.

Weigh in on whether you think AMD’s new strategy will pay off.

TechRepublic’s Servers and Storage newsletter, delivered on Monday and Wednesday, offers tips that will help you manage and optimize your data center. Automatically sign up today!

You don’t have to know everything about cloud computing, but a familiarity with the terminology will help you follow the trends and industry developments. This glossary offers a rundown of the terms you’re likely to come across.

Cloud computing is one of the hottest topics in IT these days, with Microsoft, Google, Amazon, and other big players joining in the fray. However, the technology brings with it new terminology that can be confusing. Here are some common cloud-related terms and their meanings.

Note: This glossary is also available as a PDF download.

Advertising-based pricing model

A pricing model whereby services are offered to customers at low or no cost, with the service provider being compensated by advertisers whose ads are delivered to the consumer along with the service.

Amazon EC2

Amazon’s Elastic Compute Cloud Web service, which provides resizable computing capacity in the cloud so developers can enjoy great scalability for building applications.

Amazon S3

Amazon Simple Storage Services — Amazon’s cloud storage service.

CDN

Content delivery network — A system consisting of multiple computers that contain copies of data, which are located in different places on the network so clients can access the copy closest to them.

Cloud

A metaphor for a global network, first used in reference to the telephone network and now commonly used to represent the Internet.

Cloud broker

An entity that creates and maintains relationships with multiple cloud service providers. It acts as a liaison between cloud services customers and cloud service providers, selecting the best provider for each customer and monitoring the services.

Cloud operating system

A computer operating system that is specially designed to run in a provider’s datacenter and be delivered to the user over the Internet or another network. Windows Azure is an example of a cloud operating system or “cloud layer” that runs on Windows Server 2008. The term is also sometimes used to refer to cloud-based client operating systems such as Google’s Chrome OS.

Cloud Oriented Architecture

A term coined by Jeff Barr at Amazon Web Services to describe an architecture where applications act as services in the cloud and serve other applications in the cloud environment.

Cloud portability

The ability to move applications and data from one cloud provider to another. See also Vendor lock-in.

Cloud provider

A company that provides cloud-based platform, infrastructure, application, or storage services to other organizations and/or individuals, usually for a fee.

Cloud storage

A service that allows customers to save data by transferring it over the Internet or another network to an offsite storage system maintained by a third party.

Cloudsourcing

Replacing traditional IT services with cloud services.

Cloudstorming

Connecting multiple cloud computing environments.

Cloudware

Software that enables creating, deploying, running, or managing applications in the cloud.

Cluster

A group of linked computers that work together as if they were a single computer, for high availability and/or load balancing.

Consumption-based pricing model

A pricing model whereby the service provider charges its customers based on the amount of the service the customer consumes, rather than a time-based fee. For example, a cloud storage provider might charge per gigabyte of information stored. See also Subscription-based pricing model.

Customer self-service

A feature that allows customers to provision, manage, and terminate services themselves, without involving the service provider, via a Web interface or programmatic calls to service APIs.

Disruptive technology

A term used in the business world to describe innovations that improve products or services in unexpected ways and change both the way things are done and the market. Cloud computing is often referred to as a disruptive technology because it has the potential to completely change the way IT services are procured, deployed, and maintained.

Elastic computing

The ability to dynamically provision and de-provision processing, memory, and storage resources to meet demands of peak usage without worrying about capacity planning and engineering for peak usage.

External cloud

Public or private cloud services that are provided by a third party outside the organization.

Google App Engine

A service that enables developers to create and run Web applications on Google’s infrastructure and share their applications via a pay-as-you-go, consumption-based plan with no setup costs or recurring fees.

Google Apps

Google’s SaaS offering that includes an office productivity suite, email, and document sharing, as well as Gmail, Google Talk for instant messaging, Google Calendar and Google Docs, spreadsheets, and presentations.

HaaS

Hardware as a service; see IaaS.

Hosted application

An Internet-based or Web-based application software program that runs on a remote server and can be accessed via an Internet-connected PC or thin client. See also SaaS.

Hybrid cloud

A networking environment that includes multiple integrated internal and/or external providers.

IaaS

Infrastructure as a service — Cloud infrastructure services, whereby a virtualized environment is delivered as a service over the Internet by the provider. The infrastructure can include servers, network equipment, and software.

IBM Smart Business

IBM’s cloud solutions, which include IBM Smart Business Test Cloud, IBM Smart Analytics Cloud, IBM Smart Business Storage Cloud, IBM Information Archive, IBM Lotus Live, and IBM LotusLive iNotes.

Internal cloud

A type of private cloud whose services are provided by an IT department to those in its own organization.

Mashup

A Web-based application that combines data and/or functionality from multiple sources.

Microsoft Azure

Microsoft cloud services that provide the platform as a service (see PaaS), allowing developers to create cloud applications and services.

Middleware

Software that sits between applications and operating systems, consisting of a set of services that enable interoperability in support of distributed architectures by passing data between applications. So, for example, the data in one database can be accessed through another database.

On-demand service

A model by which a customer can purchase cloud services as needed; for instance, if customers need to utilize additional servers for the duration of a project, they can do so and then drop back to the previous level after the project is completed.

PaaS

Platform as a service — Cloud platform services, whereby the computing platform (operating system and associated services) is delivered as a service over the Internet by the provider.

Pay as you go

A cost model for cloud services that encompasses both subscription-based and consumption-based models, in contrast to traditional IT cost model that requires up-front capital expenditures for hardware and software.

Private cloud

Services offered over the Internet or over a private internal network to only select users, not available to the general public.

Public cloud

Services offered over the public Internet and available to anyone who wants to purchase the service.

SaaS

Software as a service — Cloud application services, whereby applications are delivered over the Internet by the provider, so that the applications don’t have to be purchased, installed, and run on the customer’s computers. SaaS providers were previously referred to as ASP (application service providers).

Salesforce.com

An online SaaS company that is best known for delivering customer relationship management (CRM) software to companies over the Internet.

Service migration

The act of moving from one cloud service or vendor to another.

Service provider

The company or organization that provides a public or private cloud service.

SLA

Service level agreement — A contractual agreement by which a service provider defines the level of service, responsibilities, priorities, and guarantees regarding availability, performance, and other aspects of the service.

Subscription-based pricing model

A pricing model that lets customers pay a fee to use the service for a particular time period, often used for SaaS services. See also Consumption-based pricing model.

Utility computing

Online computing or storage sold as a metered commercial service in a way similar to a public utility

Vendor lock-in

Dependency on the particular cloud vendor and difficulty moving from one cloud vendor to another due to lack of standardized protocols, APIs, data structures (schema), and service models.

Vertical cloud

A cloud computing environment that is optimized for use in a particular industry, such as health care or financial services.

Virtual private data center

Resources grouped according to specific business objectives.

VPC

Virtual private cloud — A private cloud that exists within a shared or public cloud, e.g., the Amazon VPC that allows Amazon EC2 to connect to legacy infrastructure on an IPsec VPN.

Windows Live Services

Microsoft’s cloud-based consumer applications, which include Windows Live Mail, Windows Live Photo Gallery, Windows Live Calendar, Windows Live Events, Windows Live Skydrive, Windows Live Spaces, Windows Live Messenger, Windows Live Writer, and Windows Live for Mobile.

The DroboElite uses BeyondRAID to provide a high level of protection to the array and to simplify overall storage administration. Find out how BeyondRAID compares to other RAID levels, and learn about the technology’s features.

———————————————————————————————–

I recently published a Product Spotlight outlining my first impressions of the DroboElite, an iSCSI-based storage appliance aimed squarely at the small and medium business market. BeyondRAID is a technology intended to provide RAID-like data protection, while attempting to simplify it and remove some significant limitations. While Drobo’s BeyondRAID is a very cool technology that simplifies storage configuration for small and medium business, it’s not a panacea.

In this overview, I compare traditional RAID and BeyondRAID. I also look at RAID’s high-level capabilities rather than the nitty-gritty like IOPS and such.

Comparing RAID and BeyondRAID

Traditional RAID has a couple of major limitations:

• All drives in the RAID set must be the same capacity. If drives differ in size, space is wasted. For example, if you have four 300 GB drives and two 1 TB drives and you decide to create a five disk RAID set, your RAID set will consist of five drives each with 300 GB of usable capacity. Even if you have just one 300 GB drive and the rest are 1 TB drives, you’re still limited to creating that RAID set with that smallest disk (the 300 GB unit), dictating the usable capacity on the remaining disks.
• Desired protection level change requires a destroy-and-rebuild process. Let’s say you create a RAID 5 disk set and then decide you need an additional protection so you want to move to RAID 6. With a traditional RAID system, your only option is to back the data up, destroy the RAID 5 set, build a RAID 6 set, and restore the data. There are time and potential data integrity issues with this process.

Drobo’s BeyondRAID technology, which relies on similar data protection processes as traditional RAID systems, helps to minimize the limitations listed above.

• Use drives of any size (BeyondRAID’s Mixed Drive Size Utilization feature). BeyondRAID allows administrators to mix and match drives of different sizes. Through the creative use of this new technology, Drobo can make more effective use of drives of varying sizes than traditional RAID systems.
• Forget worrying about drive order (BeyondRAID’s Drive Reordering feature). In many traditional RAID systems, if you have to move a storage array, you need to be extremely careful to make sure that disks are removed and put back in the exact same order. With BeyondRAID, disk order makes no difference, so there is a bit less overall risk.
• Use on-the-fly methods to change protection levels (BeyondRAID’s Single Disk Redundancy and Dual Disk Redundancy features). With the click of a mouse, you can change your BeyondRAID protected system from single to dual-drive protection, assuming that you have enough free disk space to make the change.
• Swap disks any time, any way (BeyondRAID’s Instant Expansion feature). When a disk fails in a traditional RAID set, you replace it. If you need to expand the size of a traditional RAID set, you replace the disks one at a time, do a rebuild each time, and eventually expand the volume. With BeyondRAID, you can replace any disk at any time and even replace smaller disks with bigger ones. While this process will still initiate a rebuild, depending on what size disks you already have in your Drobo, you may immediately gain access to newly added storage. The difference is: An experienced storage administrator needs to work with the traditional RAID array, while any person with reasonable knowledge can handle the Drobo BeyondRAID technology.

There are, however, still some limitations with Drobo. The limitations include the following:

• Dual-drive systems gain no storage benefit. As you add drives to your Drobo system, you start to gain the benefit of Drobo’s ability to automatically resize and use new space. However, with just two drives, Drobo’s BeyondRAID is simply a mirrored drive set, regardless of the size of the disks. So, if you have a BeyondRAID set with one 500 GB drive and one 2 TB drive, your resulting usable capacity will be 500 GB, just as it would be under a RAID 1 scenario.
• Mixed-size drives are supported but still present real-world challenges. In order to gain the full protection offered by single or dual-drive redundancy, enough space needs to be available for protecting the array. In the case of systems with multiple drives of differing sizes, this means that the system has to be able to withstand the loss of the largest disk in the unit. So, if you have a fully populated DroboElite with seven 500 GB disks and one 2 TB disk, overall, the unit needs to hold back at least 2 TB in order to support the loss of that single 2 TB disk. Although drives of differing sizes are supported, you still have to face reality.

In Figure A, this particular DroboElite has three 2 TB disks and two 1 TB disks for a raw total of 8 TB of which 7.27 TB is actually available. Of that 7.27 TB, 1.86 TB is dedicated to data protection, leaving 5.41 TB available for data. (Note: Disk drive manufacturers continue to market disks with capacities that differ from the way that operating systems report available space, hence having only 7.27 TB of total available space as opposed to the advertised 8 TB of raw space. This same math is the reason that we require 1.86 TB of space for data protection overhead rather than 2 TB. A 2 TB disk actually formats to 1.86 TB of capacity, hence the discrepancy.)

Figure A

Overall DroboElite storage capacity. (Click the image to enlarge.)

For completeness, I thought it was important to show the overall disk layout on the DroboElite that I’ve been using. In Figure B, you can see that my unit holds three 2 TB drives and two 1 TB drives.

Figure B

The DroboElite’s disk configuration. (Click the image to enlarge.)

DroboElite’s BeyondRAID service also supports dual-disk redundancy so that the array can withstand the loss of multiple drives. As is the case with a single disk, you have to make sure there is adequate space in the array to withstand the loss of that second disk, so dual-disk redundancy means that you lose about the equivalent of the two largest disks in your array. In the case of the 3×2 TB disk and the 2×1 TB disk array, you’d lose enough capacity to withstand the loss of two of your 2 TB disks.

In Figure C, this DroboElite is in the process of converting to dual-disk redundancy, and only 3.61 TB of the original 7.27 TB will be available when the process is complete. It takes the unit a while to rebalance enough data in order to move to a dual-disk redundancy scheme. In order to make the change from single to dual-drive redundancy, you need to have enough disk space available on the Drobo to be able to lose that second disk’s worth of available capacity. By the way, enabling dual redundancy is very, very easy — simply open the DroboElite’s Settings page and select the check box next to Dual Disk Redundancy (Figure D). Enabling Dual Disk Redundancy might take a rather long time on a Drobo with a lot of data; I enabled it on my unit, which has no data, and it took quite a while to complete the process.

Figure C

With dual redundancy, the available capacity plummets. (Click the image to enlarge.)

Figure D

Enabling Dual Disk Redundancy. (Click the image to enlarge.)

Look at the bottom of this page to see a table on the Data Robotics Web site that shows a side-by-side comparison of how various features in BeyondRAID compare to other RAID technologies. I’ll explain the features that I haven’t already discussed in this column.

• Smart Volumes. When volumes are created, the space necessary is pulled on demand rather than simply being allocated immediately to the volume. This is a sort of thin provisioning technique that allows volumes to use as little space as possible.
• Automatic Healing. In order to provide a high level of protection, BeyondRAID alerts an administrator if a hard drive is going bad and then takes steps to attempt to work around potentially bad disk sectors. If a drive is starting to go bad, you want to replace it, but this interim repair step might just buy you some time.
• Data Aware. BeyondRAID technology keeps track of where actual data (as opposed to unused space) is located on each drive. By doing so, the unit can rebuild and recover more quickly after a drive failure since unused portions of the disk won’t be touched. If you’ve been managing storage for very long, you’ll probably see this as a benefit; after all, a shorter rebuild time leaves less time during which you might suffer the loss of an additional disk.
• Proportional Rebuild Times. The less data that is on a disk, the less time it takes to rebuild that data after a failure.
• Virtualization Engine. Storage virtualization has become a major trend in the storage industry, and BeyondRAID takes advantage of this technology in order to abstract a data block from its physical location. This is a major method by which BeyondRAID is able to move information around drives pretty easily. While a user thinks they’re accessing a file, they’re actually accessing a pointer to a file. When that data is physically moved, BeyondRAID simply updates the pointer, which is a seemless process to the user.
• Self-Managing. Once put in place and initially configured, BeyondRAID-based systems manage themselves for the most part. As you add and remove disks, the array does most of the work without asking an administrator to jump through hoops to get the job done.

Want to keep up with Scott Lowe’s posts on TechRepublic?

• Automatically sign up to the Servers and Storage newsletter
• Subscribe to the Servers and Storage RSS feed
• Follow Scott Lowe on Twitter