At VMworld? Try VPLEX. Like it? Take one home :-)

Customer interest in active/active datacenters is through the roof – it’s a compelling idea.

We’ve got a little pile of VPLEX/vTeleportation sessions at VMworld – I’m doing PC8051 with Beth Phalen, who is the VP who owns the product (which covers practical questions, how it works, but also shows where we are going).   Scott Lowe is doing session TA8101 (which is very focused on do’s, don’ts and best practices in these active/active datacenter use cases.

The VMware KB article that covers how EMC VPLEX supports VM HA stretched cluster also went up, you can read it here.

I remain convinced that we (VMware and the storage community) have more work to do (specifically around VM HA details discussed in our sessions, as well as partition handling at the VMware and storage layers) before I would personally do this myself.  But we decided that people were going to do stretched clusters, so it was more useful to be explicit about how to do it, pitfalls to be aware of.   The solution spot that is a rock right now is VMotion between clusters.

BTW – those areas of hardening the use case are things that we are absolutely working on together.   Come to the sessions – we’ll be joined by the VMware product teams for those areas, and let’s talk about it!

There were two shockers today in Pat’s keynote (well, not for me, we’ve been working on it for a while :-):

1) VMworld attendees can try VPLEX (it’s in the EMC hands-on-lab on the show floor), and can try it for 90 days.  Yup.  Go AT IT – and have some fun :-)   We’re confident (not arrogant, but confident).  I’ve used VPLEX and it is indeed cool, and simple.  All you need to do is go to the booth and talk to the VPLEX team there.   While you’re hanging out, you can play with it in the EMC HoL, where the Virtual Storage Appliance version is running parts of the lab.

2) we gave a preview of VPLEX Geo – aka vTeleportation across thousands of kilometers.  Please DO NOT TRY THIS AT HOME :-)  

This is a big deal, but I really want to be cautious that people don’t take it out of context.  The demo literally pulled a build (it’s software that runs on the existing shipping VPLEX hardware) right hot off the engineering presses into vSpecialist hands.   There’s also a LOT of things we need to work out with VMware to make the solution a rock (many of which I discuss in PC8051).  That’s going to take some time.  

BUT – it highlights one of the most cool things about VPLEX:

• Sure, it’s active/active – that’s unique in this category, and means that you can move a VM, instead of storage objects like datastores.
• Sure, it’s a scale-out model of commodity x86 engines – meaning loads of ports, brains, cache, and grow as you need – that’s unique in this category and scale
• Sure, it federates across distance.
• Sure, it can front EMC and non-EMC kit
• Sure, it has a weird unique capability of being read/write BEFORE all the data gets sucked over.
• But…
• IT’S ARCHITECTED TO STRETCH TO ASYNC USE CASES (the idea of distributed cache coherency).
• IT’S ARCHITECTED TO GO TO MULTIPLE SITES.

VPLEX changs the game as it stands now with the capabilities shipping since April, but the stuff to come, well – the whole landscape changes.   You could have a “global datastore”, backed with all sorts of different storage.

So – in the EMC supersession – we showed VPLEX vTeleporting a VM from Boston to Buffalo (14ms ~ 730km), then from Boston to Chicago (26ms ~1600km), then from Boston to Omaha (44ms ~ 2300km).    Those distances are a bit conservative (we looked it up based on AT&T latencies).   If you have dark fiber, 44ms @ the speed of light is actually 13,200km – that’s a third of the way around the globe.

Here’s what it looks like…

You can download this demo in high resolution WMV, MOV formats.

The fact that this even worked at this early stage is awesomesauce.   VPLEX team, you rock.  Praphul, you rock.  Stephen, Matt and the rest of the vSpecialist who worked on this one, you rock.

Comments

[Ivanfetch]

Hi Chad - we spoke at VMworld about Vplex and VMware HA. You mentioned a command which would allow us to mark paths dead (vs. resume IO), but I am not seeing this command in the KB. Is the use of this command available today?