Free ESG report : Seamless data management with Avere FXT

TechEd Australia Tens of thousands of words have already been written about Microsoft's purchase of Nokia, and as far as The Register can tell, none of them are getting read by the hordes of sysadmins and developers that descended on the Gold Coast for TechEd.

Mergers and acquisitions aren't the sort of thing that gets an audience to interrupt a speaker with applause twenty minutes or so into his slide deck. After all, finalising the financials of an acquisition doesn't happen overnight, and the business of digesting an acquisition can take years.

They have bigger fish to fry.

If you want to see a TechEd audience break into spontaneous applause – and here I am one-hundred-percent serious – give them something that they really care about. Like a shared clipboard. The people running virtual servers really did interrupt Benjamin Armstrong, Microsoft Hyper-V program manager, to applaud the simple act of being able to cut and paste text or files between VMs.

As one user explained to El Reg, something so simple and straightforward isn't trivial: “It means I don't have to FTP sh*t between instances. I mean, really, it makes life easier.” Well, there you go.

It's not only the trivia that gets a sysadmin's heart all a-flutter, though. Remote Direct Memory Access (RDMA) is also a big thing for people tasked with building a private cloud, or so the users say. In the coffee queue, another server punter put it this way: “getting stuff across the network faster than memory? That's a really big thing. Server 2008 was a big advance, and I think R2 is on the same order of importance.”

He may be right: after all, in the perfect world of the on-stage demo, a live migration that under R1 needed more than a minute to complete, with R2 and compression and RDMA was knocked over in 8 seconds.

El Reg later put this to Armstrong: since memory becomes so easily available with RDMA, wouldn't it make sense to let workloads be more flexible in how they grab memory available to Hyper-V?

“I do not talk about future directions. The one observation I will make is that I have actually seen this in the lab. I've had two boxes side-by-side where the speed of the RDMA connection between them was faster than the memory back-channel in the NUMA inside the box,” he said.

“So it's actually faster to reach memory on the other box than on a remote node. It makes you think!”

After suffering some criticism for going quiet on the service provider market, TechEd has also seen Microsoft putting SPs back into the slide decks. For example, Armstrong highlighted the importance of the VHDX virtual hard disk format to SPs.

“This is one thing that Microsoft has been watching very closely,” Armstrong said. “Most service providers don't offer clustered virtual machines, because for backups, users can see too far into the infrastructure. That's not the case with VHDX.”

More debatable on the floor was the R2 limit of 64 virtual processors per Hyper-V VM. As The Register asked, doesn't a system with four sockets, 12 cores per socket, and two threads per core already beat the 64 virtual processor limit?

“Honestly? The 64 virtual processor limit is there because no-one's asked for more, with a legitimate reason other than playing with numbers. The day I start hearing this being an issue … yes," Armstrong said.

“But let's be clear, we're beating VMware on scalability, we're beating VMware on performance.

“When I talk to CEOs and CIOs today numbers, they're telling me that the numbers are 'way beyond what we care about'.” ®

Bootnote: Microsoft customers may not be asking for more than 64 virtual processors, but VMware customers are: the recently-released vSphere 5.5 can handle 4,096 virtual CPUs, rather more than Hyper-V, no matter what Armstrong says about who's on top.

The author is attending TechEd Australia as a guest of Microsoft, which has shelled out for travel, accommodation, nourishment and Nokia.

5 ways to reduce advertising network latency