Microsoft and Intel Pushing iSCSI Performance Limits

Posted on Jan 07, 2010 by Stephen

How fast can iSCSI get?

“Maximizing Hyper-V iSCSI Performance with Microsoft and Intel” might sound like another “blah blah” marketing piece, but a little birdy tells me that this webcast will drop a bombshell about iSCSI performance.

Lots of storage and networking folks don’t give iSCSI and Microsoft the credit they deserve. “iSCSI is cheap and easy,” they say, “but real performance requires Fibre Channel.” Those of us who have an open mind about such things know that this is simply not the case. The fastest SAN I ever saw was based on iSCSI, and Microsoft demonstrated wire-speed iSCSI over 10 Gb Ethernet in March. I never saw a Fibre Channel SAN (even an 8 Gb/s one) push over a gigabyte per second over a single link!

Still, ask the average sysadmin and they will tell you that iSCSI isn’t for high performance applications. That’s why folks should tune in to this webcast, as Microsoft and Intel knock down another iSCSI performance myth. Note that even though Hyper-V is called out in the title and description, this discussion is really about Windows Server 2008 R2 and applies equally regardless of whether or not you use Microsoft’s hypervisor.

Watch this space for a summary of the news immediately following the announcement.

What: Maximizing Hyper-V iSCSI Performance with Microsoft and Intel webcast
When: Thursday, January 14, 2010 8:00 AM Pacific Time
Where: MSEvents.Microsoft.com
Who: Anyone interested in high-performance storage and server I/O

You might also want to read these other posts...

Tags: 10 GbE, 10 gigabit Ethernet, Fibre Channel, Hyper-V, Intel, iSCSI, Microsoft, performance, webcast, Windows Server 2008 R2

My Event Schedule

No events

http://thestorageanarchist.com the storage anarchist

For the record, both Symm and CLARiiON routinely deliver > 1Gbps over single FC links. Server-side configuration is usually the bottleneck (queue depth, etc.)…iSCSI can be limited by these same settings – would be great to see apples-apples comparisons with equivalent configurations.

Oh, and Symmetrix is still the only enterprise-class storage array with native iSCSI support. It's great to see the performance advances being made by Microsoft and Intel- hopefully will drive up adoption across the board.
http://blog.fosketts.net sfoskett

I've seen greater than 1 gigabit per second on 4 Gb Fibre Channel links, too. But I've never seen anything like the more than 1.1 gigaBYTES per second Microsoft and Intel showed in March. It's simply beyond the capabilities of any Fibre Channel SAN mathematically, let alone practically.

How fast can the Symmetrix or Clariion push data out those iSCSI interfaces? I know they were recently upgraded from 1 Gb to 10 Gb, but how fast are they?
http://thestorageanarchist.com the storage anarchist

I've seen nearly .88 GigaBYTES/sec on 8Gb fibre as well.

My point is that while it is commendable that one (apparently homogeneous) iSCSI driver stack is now able to attain “link speed” over 10GbE, that does not necessarily mean that FC cannot also deliver “link speed.” True, the max link speed of Ethernet is today faster than Fibre Channel, but the FC protocol is not necessarily any more limiting than iSCSI.

And all that said, did MS/Intel report on small-block latencies? I suspect there may be more separation if the metric is small-block I/O (say, 512 byte reads/writes).
http://blog.fosketts.net sfoskett

What storage system pushed that kind of throughput over 8 Gb FC? And what HBA was used? That's much faster than anything I've seen, and I'm truly impressed that FC could deliver that kind of link efficiency!

As for MS/Intel, I agree. Throughput is great, but what about latency? How many IOPS can be pushed through a 10 Gb iSCSI link? I can't wait to hear an answer to that question!

For what it's worth, how many IOPS have you seen over an FC link?
http://thestorageanarchist.com the storage anarchist

We push such single-port rates out of both Symmetrix and CLARiiON. It's easiest to get this using multiple servers to generate the I/O load; I don't have the specifics about the HBAs/hosts used.

II'm not sure what the raw IOP numbers are for each platform…
http://gestaltit.com/all/tech/storage/stephen/microsoft-and-intel-push-one-million-iscsi-iops/ Microsoft and Intel Push One Million iSCSI IOPS – Gestalt IT

[...] Microsoft and Intel Pushing iSCSI Performance Limits [...]

blog comments powered by Disqus

Stephen Foskett, Pack Rat