Microsoft has enlisted X-IO and others as its proxies alongside Hyper-V and SMB in a war against VMware and NFS. SMB v3, formerly known as CIFS, is Microsoft's preferred way of getting file-based networked storage data into servers running its Hyper-V hypervisor. The opposing camp is seen as VMware and NFS. X-IO's UK solutions …
Fight! Fight! Fight!
So Microsoft is taking on Apple (and Android) at tablets and VMWare at Virtualisation - and is probably too late in both arenas.
Will they be putting a 'soccer' side out to face Spain too?
Would the real performance metrics please stand up?
McLaughlin said this was "over 50 per cent higher than anything else possible on the market."
Steve Sicola, X-IO's chief technology officer, called it [...] "delivering in just one rack what our competitors can only achieve with many times the equipment and at much higher costs."
Make up your minds, boys. It's either "many" or "one and a half".
Re: Would the real performance metrics please stand up?
Ken - I think you're confusing the performance (15GBs) with the amount of kit (200 drives).
Some of our competitors are achieving 10GBs but by using many more disks
So to answer your quesiton - both ;-)
For what client?
For a Linux VM or Windows VM? My own experience shows NFS as significantly faster than CIFS for the same hardware/network system, so can we please have some like-for-like comparisons?
Methinks MS pushes CIFS because that is Windows default networking and not because it is actually any better...
Re: For what client?
This is Hyper-v running over CIFS as far as I'm aware there isn't a version of Hyper-v for Linux as yet!!
It is entirely irrelevant what the VM is running, this is talking about the underlying architecture.
Try to pay attention to the article eh...
Re: For what client?
NFS may well be faster than CIFS, but we're talking about SMB not CIFS, there is a difference. CIFS
was dropped at Win 2008, IIRC.
I'm curious how they intend to get the latency the same using an intermediary file server compared to connecting direct to the SAN using block protocols. As far as I'm concerned any extra hop is to be avoided, but then I know how to set up the storage so don't need to "remove complexity"
They're not talking about getting this to FC SAN level performance, but NFS is a very popular way of hooking up back end disk to a VMware farm, so MS want to show that their prefered stack of hyper-v and SMB can out perform it. Obviously anyone with the cash would use FC SAN for either system, but it is very expensive.
nope, a lot of shops are using netapp and avoiding fc san like the plague. It is not just a matter of being cheaper (netapp is not), it is very much a matter of ease of use, equal or better performance than FC, nice to have goodies (snapshots are wonderful in netapp, for instance) and coupling your backups to your storage.
If you need a block device (exchange bullsh*t), just use iscsi. You have 10Gb switches, so who cares about a little overhead of the protocol.
FC is old tech but it will still be around for a while because plenty of people are very heavily invested in it. I feel their pain :-)
Microsoft took over 5 years to get their OS working right, and it took them 10 years to truely understand how to network, no they think they can take on VMWare? No way. VMWare is way ahead of microsoft, just like Novell was way ahead of Microsoft in Networking. It was only the long painful road of advancement and superrior marketing that enabled them to capture the networking market, and it will be the same road in the virtual machine market as well.
It will happen. It will just take a long time.
Re: Right :-)
Yes, because when I think of Microsoft, I immediately think "networking." Why, I've got a Microsoft switch sitting right here . . . no, wait, I don't.
You seem to have mistaken "server" for "network." Sun made the same mistake, and now look where they are.
Wasn't MS touting their new NFSv4 client a while back? Now NFS is the devil again?
Did I miss something?
I read a lengthy panegyric to Microsoft CIFS and one of Microsoft partners.
However I almost lost my eyeys looking for a *comparison* against NFS that (supposedly) was beaten badly, into dust by author's words.
Please, give me comparison and not another Microsoft advertisement!
SMB 3 versus NFS
Guys, here's the deal: SMB3 can do RDMA. With the right gear, it can go faster than NFS...much faster.
The issue is that the gear is expensive. Quite expensive. Performance crown? Maybe. But we're comparing apples to oranges to grapes to a planter of petunias.
It isn't about IOPS. It's about IOPS per $ and IOPS per W. How many network ports on what kind of gear? What's the power consumption like? The CPU overhead?
Oh wait, the CERN announcement is on, gotta jet...