Processor Technology:

AMD Opteron has played a big part in us being able to reduce power. There are several reasons for this. When Opteron came out, they decided to take a different approach and deviate from what was the standard architecture that Intel always used. They integrated the memory controller on the processor and were able to remove the Northbridge from the motherboard. Because of this, the memory was linked directly to the processor and eliminated the usual bottlenecks that the Intel Platforms experienced. This allowed them to stay on older technology memory instead of jumping to the next while maintaining competitive performance. Because of this, we’ve been able to maintain performance with the Intel Xeons while taking advantage of faster memory throughput, reduced power, and reduced cost. One great example of this is that the AMD Socket F platform uses DDR2 memory and has better performance in some cases against Intel Xeon platforms with Fully Buffered Dimms (which take about twice as much power per module).

Processors Type:

Today we’re heavily focused on purchasing the most efficient processors. This includes the High Efficiency (HE) models from AMD and the Low Voltage Xeon from Intel. We tend to stay away from the more power hungry series unless our customers require them.

Hard Drives:

With Small Form Factor 2.5″ drives becoming more mainstream there are a lot of power benefits from these. They run cooler and use less power than their larger 3.5″ counterparts. We’ve been working on transitioning to these as they provide better performance as well.

Power Supplies:

We’ve evaluated a lot of different power supplies to identify the most efficient power supplies to use. We’ve found the perfect mix of efficient power supplies that meet the 80Plus Standard and that are very reliable. In a datacenter environment it’s critical to ensure that everything will be reliable.

Fans:

We do a lot of testing to ensure that our servers receive the proper amount of cooling but we don’t over do it. In a datacenter environment, a single fan may not eat up much power, but with thousands of servers it can add up to a lot of wasted power.

Virtualization:

While Virtualization is not a component of the server, it does allow for our customers to reduce their datacenter for servers that aren’t completely utilized. This in turn saves them money if they are able to consolidate and use fewer servers to accomplish the same goals.

If you purchase servers today for your datacenter, I urge you to push your server manufacturers to make their servers more efficient. We’ve been working with our manufacturers for quite some time to ensure our equipment is energy efficient and “Green”. I think this is very valuable not only for the customer but also for the Manufacturer. Dell, one of our server vendors, has been very helpful and open to listening to what we wanted. With Dell being an Intel only shop for a long time, they listened to us when we requested AMD Technology. Because of this, we now deploy the PowerEdge 2970 as part of our Green Configurations.

What are you doing in your datacenter today to reduce power and be more energy efficient?

As some of you know, Rackspace just launched our new virtualization product. It’s been in the works for a while and we’re proud of what we’ve put together. I’m also sure a lot of you are familiar with virtualization technology and how it works. However, just as many of you may be wondering how you can benefit from it.

Reference.com defines virtualization as “the abstraction of computer resources” or “a technique for hiding the physical characteristics of computer resources from the way in which other systems, applications, or end users interact with those resources.” So what exactly does all of that mean?

Virtualization allows you to carve up a physical server into individual operating system (OS) instances or Virtual Machines (VMs). VMs can run a variety of different types of OSs on a physical server, all operating individually and securely from one another. How many VMs can you put on one server? That depends mostly on your hardware and the resources your VMs require. As servers become more powerful and cores per processor increase, a lot of applications just don’t make use of those extra resources. Virtualization can give you the ability to make better use of those resources and reduce the number of servers you deploy, which eases system administration and can be more cost effective.

Usually you’ll see two common types of virtualization implementations. The first one is called a Hypervisor. A Hypervisor is essentially the OS layer of the machine that enables virtualization. It’s typically a very small footprint, allowing for better performance and more available resources for the physical machine.

The other type of virtualization implementation is application based.  In this case, the virtualization software installs on top of a running OS. This usually reduces the amount of features available, in addition to reducing performance and allowing fewer resources dedicated to VMs.

A lot of the larger players in the game are developing their virtualization software to go the Hypervisor route.  VMware’s ESX, also known as VI3, is one of the dominant players  when it comes to Hypervisor based virtualization.  They’ve been one of the leaders in virtualization for a while and have a very large feature set when compared to Xen or Microsoft’s Hyper-V, which is unreleased at this time. 

Some key virtualization features:

• Legacy Operating System Support (varies by Hypervisor): Virtualization abstracts the hardware and usually uses very common virtual hardware; therefore, older legacy OSs are able to function without issue in a virtual environment. This is useful for companies with older applications that only support certain legacy operating systems.
• VM OS Provisioning: Since the environment is virtualized, deployment of new OS VMs happens quickly. Once a base Hypervisor is in place, VMs can be provisioned quickly since a new physical server does not have to be assembled, and prepped for the Datacenter.
• Templates and Clones: Templates allow you to capture the state of a VM so you can redeploy a new VM later, maintaining the same configuration and software. Clones allow you to make a complete copy of the VM in case you ever need to revert back to that original copy.
• Snapshots: Allow you to take a picture of a VM at a point in time and revert back to that time point if you chose. This is useful for testing/development, deploying patches, upgrading applications, and trying something new. If something goes wrong, you simply revert back to the previous snapshot and the machine is back to its original state.
• Live Migrations: With shared storage, many virtualization technologies allow you to live migrate a VM from one system to another while the VM is still running. This is useful for performing upgrades or repairs to physical servers without any downtime. For example, if you had two physical servers running three VMs per server, you would need to do maintenance on one of the servers without any downtime. A live migration (also known as Vmotion to VMware users) allows you to shift the VMs off the server that is going down for maintenance while they are still running and serving connections. Once your maintenance is complete, you can shift them back to the original physical server.

There are many other features out there, but these are the features that most virtualization vendors offer. Rackspace has partnered with VMware to provide its current, dedicated virtualization environments, but we’re also heavily investigating other virtualization technologies like Xen and Microsoft’s Hyper-V (still in beta). 

I encourage you to visit the link above.  It has a wealth of information on virtualization and many links to articles about virtualization to decide if it’s right for you. If you have any questions about virtualization feel free to contact us.