nzsteve
Enthusiast
Enthusiast

Hyper-V R2 "dynamic I/O redirection"

Hi,

Just throwing this out for discussion - anyone has had a chance to look at this at all?

Microsoft describe it as a way of detecting and recovering from network or SAN failure. The doc here talks about it Hyper-V R2 Reviewers Guide

Extracted from the doc:

"Because of the architecture of CSV, there is improved cluster node connectivity fault tolerance that directly affects VMs running on the cluster. The CSV architecture implements a mechanism, known as dynamic I/O redirection, where I/O can be rerouted within the failover cluster based on connection availability...if the SAN connection on Node 2 fails, the I/O operations are redirected over the network to Node 1. Node 1 then performs the I/O operation to the SAN. This allows you do a Live Migration of the VM running on Node 2 to Node 1...The next type of failure that can be redirected is the failure of network connectivity for a cluster node. As shown in the following figure, the primary network connection between Node 1 and Node 2 fails. Node 2 automatically reroutes network traffic over a redundant network connection and Node 1 performs the network I/O."

So...

1) Anyone know how well this works in practice?

2) Assuming it works, how does vSphere compare?

I've not had much chance to play with vSphere as yet so I'm unsure if theres anything new in this area, but am I correct in saying that theres no direct equivalent? For 3.5 at least I'm pretty certain that:

  • Unless the VM network vSwitch is also used with an SC port group acting as the only HA heartbeat network, there's no way to failover automatically if all network uplinks on the VM's portgroup are offline; And

  • if all paths to a SAN LUN drop to a single host in a cluster, the VMs running on that host go offline, but would be brought back up on a different host if virtual machine HA is turned on?

Anyone have thoughts on this?

Cheers!

0 Kudos
12 Replies
AntonVZhbankov
Immortal
Immortal

CSV is kind of SMB sharing with master-host to resolve reservation conflicts.

VMFS unlike CSV works without master host, so you can't redirect traffic to another host - there is no storage level communication between hosts.


---

VMware vExpert '2009

http://blog.vadmin.ru

EMCCAe, MCITP: SA+VA, VCP 3/4/5, VMware vExpert http://blog.vadmin.ru
0 Kudos
azn2kew
Champion
Champion

Can't seem to open the reviewer's guide posted above, correct the link by chance? want to see what M$ has to bloat this time and see how it works.

If you found this information useful, please consider awarding points for "Correct" or "Helpful". Thanks!!!

Regards,

Stefan Nguyen

VMware vExpert 2009

iGeek Systems Inc.

VMware, Citrix, Microsoft Consultant

If you found this information useful, please consider awarding points for "Correct" or "Helpful". Thanks!!! Regards, Stefan Nguyen VMware vExpert 2009 iGeek Systems Inc. VMware vExpert, VCP 3 & 4, VSP, VTSP, CCA, CCEA, CCNA, MCSA, EMCSE, EMCISA
0 Kudos
nzsteve
Enthusiast
Enthusiast

Sorry, not sure whats up with the link (brackets in the doc name maybe?).

Copying the below, including the (beta).doc part into browser should work:

(BETA).doc

Or search google for "Windows Server 2008 R2 Reviewers Guide (BETA).doc". Should be the first hit. Redirected I/O starts page 14.

I've found some other references to redirected I/O here and there, but nothing that really gives details on how it works exactly, so I've not been able to guage how useful it might be in comparison to what we can do with ESX.

Cheers

0 Kudos
azn2kew
Champion
Champion

Thanks for the links and will read up how things written.

If you found this information useful, please consider awarding points for "Correct" or "Helpful". Thanks!!!

Regards,

Stefan Nguyen

VMware vExpert 2009

iGeek Systems Inc.

VMware, Citrix, Microsoft Consultant

If you found this information useful, please consider awarding points for "Correct" or "Helpful". Thanks!!! Regards, Stefan Nguyen VMware vExpert 2009 iGeek Systems Inc. VMware vExpert, VCP 3 & 4, VSP, VTSP, CCA, CCEA, CCNA, MCSA, EMCSE, EMCISA
0 Kudos
nzsteve
Enthusiast
Enthusiast

Just adding an update to this.

I still havent had the chance to try and implement this, but had a talk to our MS partner virtualisation specialist earlier in the week. The I/O redirection does ecactly what the doc says - redirection of network or storage traffic via a different cluster node if the path to network / SAN from a host has failed.There is a potential performance impact, but it does ensure that I/O doesnt drop if a connection is lost. The redirect kicks in automatically if there is a problem detected, but for now that is it - there is no automated live migrate to a 'healthy' host to return optimal performance. It's potentially possible to script SCOM to trigger a live migrate of VMs from one host to another if this occurs, but theres no built in management pack for this at the moment.

Assuming again that it does all work, anyone have comments on how this maps to vSphere capability?

I don't know when I'm next going to get some lab time, but will report back anything else I can find out.

Steve

0 Kudos
mreferre
Champion
Champion

I don't think vSphere has anything like this (i.e. if a host loses connectivity to the SAN/Storage.. it's gone).

However in a proper setup with redundant paths this shouldn't happen. Or put it in another way... what's the potential scenario this feature would cover?

Massimo.

Massimo Re Ferre' VMware vCloud Architect twitter.com/mreferre www.it20.info
RParker
Immortal
Immortal

Microsoft describe it as a way of detecting and recovering from network or SAN failure.

I hate to burst your bubble, but this is called FAIL OVER. If you have multi path to the switch, and your infrastructure has many paths it will do this via multi path. So ESX has this, but it's a different name. I don't see where this is new technology, Microsoft typically likes to rename things to make it sound new, but it's not. It's a fancy name for fail over.

That's all.

RParker
Immortal
Immortal

if a host loses connectivity to the SAN/Storage.. it's gone

If the host loses connectivity to the SAN, that's because the path can't route due to HBA failure. If you have more than 1 path it should not fail. If you do lose connectivity, the SAN is probably down in which case ALL your data is probably gone...

0 Kudos
mreferre
Champion
Champion

If you have more than 1 path it should not fail. If you do lose connectivity, the SAN is probably down in which case ALL your data is probably gone...

Isn't this exactly what I was trying to say?

BTW what MS does is not standard multipathing... what they do is (apparently) that they can use a host that has SAN connectivity as a proxy for a host that has no longer SAN connectivity. Based on the slides I have seen this is done via the network.

Massimo.

Massimo Re Ferre' VMware vCloud Architect twitter.com/mreferre www.it20.info
0 Kudos
RParker
Immortal
Immortal

Isn't this exactly what I was trying to say?

Sorry I was distracted during this email and it was early... Smiley Happy

Based on the slides I have seen this is done via the network.

Yes, and what the slides DON'T tell you is that NTFS is NOT a cluster file system, Even Microsoft can't share access to the same files on NTFS directly. So they came up with this system to be an end around for the limitation. It's almost like a new way to do DFS, they just renamed it to make it sound cool.

0 Kudos
nzsteve
Enthusiast
Enthusiast

Thanks for the comments guys.

We can configure multiple paths / redundancy to storage and networks in Hyper-V via drivers in the parent partition, so I'm not sure that comparing this directly to multipathing is 100% correct? It is a pain to set up and manage this in Hyper-V compared to ESX and vCenter, but it can be done, so I'm not sure this is exactly the same thing.

I honestly don't know how useful this is going to be when deployed, but when looking at the feature set for R2 this stuck out as something potentially different (or extra) from how ESX works. Prehaps it's going to be more relevant to the SMB space where people may not have the funds to put in completly redundant fabrics and networking infrastructure, but extra resiliance in any form has to be a bonus?

Really just trying to get my head around the differences at the moment. The $$ savings on Hyper-V licensing mean that every customer is going want us to justify the extra spend on vSphere, so I think its important to understand the little bits like this that MS may push and customers could be asking questions around.

Cheers

0 Kudos
Smeagol
Enthusiast
Enthusiast

More info here http://www.delltechcenter.com/page/Hyper-VR2CSV+FAQ

I agree, the significant cost savings will see vmware lose the entry level sales and we all know what that means in the long run!

0 Kudos