EMC PowerPath/VE 5.4 SP2 + vSphere Enterprise

It’s here… Power/VE 5.4 SP2.

What’s new:

• No more RPQs, and no ESXi 4.1 gap - support for VMware ESX and ESXi 4.1 on PowerPath/VE 5.4 SP2.
• Support for VMware Update Manager (VUM) installation using both offline bundle and online depot format for importing updates.
• Support for Windows 2008 and 2008 R2 as guests
• Support for latest EMC stuff (FLARE 30, DART 6.0, Invista 2.3/2.4, Enginuity 5874) and more third-party arrays (Hitachi AMS 2300 storage array, IBM DS8x00 storage array were added)

As always, simplicity built in, along with enterprise-grade multipathing - important.   All reasons why PowerPath/VE is selling like hotcakes.

Couple of other things I want to make sure people know:

1. **NEW** PowerPath/VE is supported on vSphere Enterprise Edition as of vSphere 4.1 (not limited to vSphere Enterprise Plus only)
2. EMC absolutely supports vSphere NMP (Native Multipathing), and NMP Round Robin is our recommended default on all modern EMC storage environments (as shown here, we will even be automating the setup of this).   Consider PowerPath/VE “a more than” (and isn’t free) not “in substitution for”.   As an example that we’re not the only ones doing this – note that Dell/EqualLogic has introduced their own MPP (Multipathing Plugin).
3. There is a new “unlimited vSphere cluster” licensing model.
4. PowerPath/VE technically supports (as in is on the EMC Support Matrix) MSCS/WSFC in the Guest OS as of 5.4 SP1.  This is important as vSphere NMP doesn’t support Round Robin in MSCS/WSFC configs. 
5. Very important IF YOU ARE USING A SYMMETRIX AND SRDF:

ETA emc236463: Symmetrix DMX-3, DMX-4 and VMAX: Enginuity 5773 and 5874 while using PowerPath/VE: Virtual machines periodically hang, become unresponsive and disappear from the VI Client after installing PowerPath/VE

EMC Technical Advisory

ETA Title: ETA emc236463: Symmetrix DMX-3, DMX-4 and V-Max: Enginuity 5773 and 5874 while using PowerPath/VE: Virtual machines periodically hang, become unresponsive and disappear from the VI Client after installing PowerPath/VE

Comment:   This ETA is being redistributed because the Enginuity fixes have been tested and are now available. This ETA is also now being distributed to customers.

Web Link: http://csexplorer.isus.emc.com/eservice/iviewcs/ui/eserver.asp?id=emc236463

BTW – I know that the licensing model leaves something to be desired.  Am working this with the PowerPath team – IMO, ideally would just use vSphere’s new licensing model.

Here’s a quick demo of how simple it is to install:

Enjoy!

Comments

[Jason Boche]

From a technology perspetive, PP/VE is great. 5-10 years ago, the licensing model would have been acceptable for non ESX environments. My only issue is, outside of the "unlimited model", the current PP/VE licensing scheme essentially enacts a snail mail version of activation each time a (what should be stateless) host is rapdily redeployed. This removes the rapid attribute from the redeploy process. Thank you for taking customer feedback and making products better.

[Tomi Hakala]

Chad, in your video you say that with PowerPath/VE you have full support for HA/DRS with MSCS enabled guests, but "Setup for Failover Clustering and Microsoft Cluster Service" document for vSphere 4.1 says that vMotion of clustered virtual machines is not supported? I don't see how MP plugin could change that. This is an very important issue for us, so can you be more specific on this?

[Chad Sakac]

Tomi - I know about that disconnect. The VMware support position for clustering is fundamentally gated by QA/Test resources. the eLab folks tested it, and it passed all the qual work we've done for EONS for MSCS/WSFC clustering testing. If you look at the old WHCL (when cluster configs needed to be explicitly submitted), EMC was the storage behind HALF of all the cluster configs.

The test harness deals with all the conditions that a storage device needs to handle during all cluster operations (all the various reservation states) and is a very, VERY mature test harness.

So - EMC will support that config, today, period.

I'm trying to see if I can get VMware to update their doc to reflect that (and I'm also proposing that EMC eLab test basic NMP RR - also not supported in that doc, and iSCSI and FCoE - also not supported in that doc).

A little blurry, but is important to understand that it's fundamentally not a technical issue, but a question of what officially is qualled.

Working to eliminate any confusion (by trying to get VMware to update their VMware's docs).

[Chad Sakac]

@Jason - I totally agree that the licensing scheme is archaic. Amusingly, spent my morning with the licensing team trying to get them to adapt EMC's licensing models (broadly) to the virtual world, and in the VMware-specific context around PP/VE, make it link into vSphere's native licensing model.

[JeffS]

Chad,

I'll toss in another vote for improving the PP\VE licensing model. It should, as you comment, be part of vSphere's native licensing model. While we wait for that to happen, one step in the right direction would be to provide a PP\VE licensing VM template/appliance (such as is done for the solutions enabler). It would save having to build one from scratch.

Also on PP\VE install - The documentation could use a bit of help. It's not that it's missing information, but rather, one must jump between several sections and even different documentation-sets just to get it running.

[Derek]

Can you publish a list of all currently supported arrays? For example, the EMC site generically lists IBM and HP, but I doubt the full line of each vendor is supported. If this list was publically available, it might help people decide whether they can even use it in their environment. For example, does it support HP P4000, EVA and 3PAR? How about mid-range IBM arrays?

[Matt]

is 5.4 SP2 supported on ESXi 4.1 Update 1?

[Bryan Rood]

I've read this and the LONG licensing doc and I still don't know *what* EMC's licensing model is for PPVE. Can someone comment on that and just mention what the model is? I.e. is it per host, per HBA, per array, per socket, per core?

Thank you.