Expert video: Brien Posey explains ODX in Windows Server 2012Date: Feb 18, 2013
In the fifth and final video tip in this series, Microsoft MVP Brien Posey covers Offloaded Data Transfer (ODX) in Windows Server 2012. Read some of Posey's remarks below or view his presentation above to learn more about how this feature can enhance copy and move operation performance and how these enhancements can affect the rest of the data center.
One of the big new storage features in Windows Server 2012 is Offload Data Transfer support, or ODX support. The basic idea behind ODX is that it's designed to improve performance for both copy and move operations for Windows 2012 servers that are connected to SAN storage. The task of copying data or moving data ends up getting offloaded from the server itself to the SAN hardware.
ODX is designed to improve the efficiency of copy and move operations by offloading resources from the server and helping those copy or move operations to work a lot faster. ODX is a hardware standard, and the process works by offloading the process from the Windows 2012 server to the storage hardware.
View the rest of Brien Posey's WS 2012 tip series:
Video tip 1: Native iSCSI Target Software
Video tip 2: Native data deduplication capabilities
Video tip 3: Resilient File System (ReFS)
Video tip 4: Windows Storage Spaces
With a standard copy, we have read the data onto the Windows server, and then, once that data has been read, it's buffered. Then, it's written back out to the other volume. That's how copy operations normally work in a Windows Server environment.
In the case of ODX, the entire process gets offloaded to the storage hardware. We don't have this traffic going back and forth to the Windows server. Instead, we're performing a copy operation straight from one volume to another. Obviously, there is a little bit of traffic that has to occur because you have to be able to request that the copy operation even happened in the first place. That's done through tokenization. This is, in a nutshell, how the ODX copy process works.
One thing that you have to understand about the copy process is that there are certain system resources that get used anytime you do a standard copy. For example, CPU cycles get used, some system memory gets used, and you're also consuming network bandwidth consuming. You're still using some of these resources when you make an ODX copy, but the resource consumption is greatly reduced because you are offloading that copy process to the storage hardware rather than using the server's own internal resources to make that copy happen.
It's one thing to generalize and say that you're going to reduce resource consumption by using ODX, but what about some benchmarks? In one test, Dell copied 11 GB of data. They did it two different ways. They made an ODX-enabled copy, and then they created a registry key to keep ODX from being used. Both tests used exactly the same hardware.
When a standard copy was done, that 11 GB took 108 seconds to copy. That copy process was reduced all the way from 108 seconds down to 20 seconds, just by using ODX. That's over a 500% improvement in the copy speed.
What about network utilization? Network utilization doesn't normally happen at a constant rate. It varies throughout the copy process, depending on exactly what's happening at any given moment. With the standard copy in this particular test, network utilization was 37% -- but it did occasionally also spike to 100%. When ODX was enabled, network utilization ranged anywhere from 0% to 1% because the operation was offloaded from the server hardware. As far as CPU utilization goes (and this was on a server with a 2.3 GHz processor), the CPU utilization was 42% for a standard copy. When ODX was enabled, that utilization dropped from 42% to about 10%. You can get a basic idea of the types of performance improvements that you can expect by using ODX.
Remember to only use ODX-certified storage hardware. If you use storage hardware that claims to support ODX, but it hasn't officially been certified as ODX-compliant, then it might or it might not work with Windows Server 2012.
About the presenter:
Brien Posey is a regular SearchStorage.com contributor and a Microsoft MVP with two decades of IT experience. Before becoming a freelance technical writer, Brien was CIO for a national chain of hospitals and health care facilities. He also served as a network administrator for some of the nation's largest insurance companies and for the Department of Defense at Fort Knox.