EMC World Day 1: VMAX Update

VMAX 3 is now hitting its stride.   In Q4 2014, it got the full reach of it’s core data services with SnapVX and  SRDF.  RDF has always been the hallmark of the platform and the defining element of “Tier 1 storage” – and remains that way today.   In Q1, it showed in the VMAX results (ultimately – all public companies cannot hide – the facts are the facts).

Here at EMC World we’re announcing the next wave of enterprise data services on VMAX3.  

It’s worth reiterating something I’ve been on the warpath internally and externally:

• The era of “consolidate all your workloads in one place” has been over for a LONG TIME.  Anyone suggesting “consolidate all your workloads in one data plane/array” is off their rocker.   It’s as silly as saying “build one giant RDBMS as the “single record of truth for the enterprise” – we live in an era of polyglot persistence :-)
• The era where the “Tier 1 Enterprise arrays” where the place you went for maximum performance has been over for a LONG TIME (though they remain an excellent performer for many mixed workloads – just not necessarily the lowest cost way to do it if that’s your only driver).
• That all said - the era of the “Tier 1 Enterprise arrays” as the place you go for the ultimate in availability services for workloads that depend on infrastructure resilience has NO END IN SIGHT.

I want to be explicit about what I mean about that 3rd point.

• Its not about being “reliable” (time between failures).  AFAs are “reliable”.  It’s also not about availability in a purely statistical sense.  AFAs and SDS models are both “reliable” and “available”.   Specifically, the XtremIO install base data is showing 99.9995% availablity.  It has the highest customer sat scores we have.   VNX is highly available.   ScaleIO is highly available.   Isilon is highly available.
• It IS about availability services like:
  • 3 site replication.0 RPO or near 0 async replication – but at huge scale (in terms of thousands and tens of thousands of devices, changes, bandwidth).
  • consistency groups at large scale (thousands or tens of thousands of devices).
  • T10DIF for when you have hundreds of terabytes or petabytes of Oracle databases – and cosmic ray silent data corruption is  real concern (which is why it’s a feature that shows up on hybrid arrays born with classic enterprise workloads as the target)
  • Remotely active replicas – and now, even active/active remote replicas.
  • And, of course, rich Mainframe and iSeries platform support.

There is no AFA that can do those things.   There are of COURSE hybrids that have existed for 10+ years (some even pre-dating the invention of NAND) that get “dressed up” as an AFA.   Will true AFAs (and IMO, this is a storage stack whose whole IO path presumes only ever persisting on NAND – I wouldn’t include VNX, VMAX, VSAN, ScaleIO in that category) every support those use cases?  Maybe.  I doubt it.  That set of workloads is critical, but not growing at a huge rate, so not a priority in the AFA roadmaps.

Now, that list of Enterprise availability services may seem “foreign” to some.  To some, it may seem outright “alien”.   To others, those requirements represent a critical part (rarely the only part) of their persistence needs.

I want to be clear.   Just like the market that is the “Tier 2 storage market” – this segment is in the process of being disrupted.  The list I noted above of “Enterprise availability data services” – well, it’s a subset of the workloads that the “Enterprise Tier 1 storage” (not just VMAX) market serves today – but what an important set of workloads.  Society runs on these – and that’s not an exaggeration.

We’re not overly concerned about VMAX becoming the “Enterprise Data Service Platform” vs. the “monster workload consolidator”.  We’re not concerned that workloads that DON’T have those requirements are being cannibalized by things like XtremIO and ScaleIO and the like – in fact, we encourage that.

That all said – there are very important workloads that DEMAND those sorts of data services – and you would be amazed at what they underpin, including cloud SaaS apps that we use every day – not just banking/financial/healthcare apps that people naturally associate with “Tier 1 mission critical”.

The name of the game is to embrace disruption, and expand the capabilities of the whole portfolio, each in their domains!   In that vein of continuing to expand the capabilities of the Enterprise Data Services platform that is VMAX what’s new with VMAX3?   Answer: A huge software update:

• FAST.X – extending the data services (RDF, Host IO QoS, ProtectPoint, SLO based management) to things like XtremIO and public cloud object storage via CloudArray.
• SRDF/Metro – bringing active/active storage models natively for customers with VMAX<->VMAX scenarioes.
• Extending the HyperMax model to incorporate more containerized functions – namely the ViPR Controller.   We have been delivering the containerized embedded NAS (eNAS) function through the tail end of Q4 and now Q1 2015.   It’s an illustration of how we’re innovating in the core code stack – enabling adding embedded functions without adding appliances.  The VMAX always had a rich SLO-based provisiong model (check it out at the booth to see how much has changed)
• Expanding ProtectPoint support to broader workloads (originally was Oracle via RMAN, now SAP via BR*Tools, DB2 via ACS, and also general purpose Windows filesystems, and Exchange, SQL Server and more).

VMAX is holding off the zombie apocalypse – that is the domain of the workloads that demand RDF, demand mainframe support, T10DIF and more.   It isn’t the answer for the entire world of persistence (nothing is – unless you work for a one-product company).   But – it IS the answer for some workloads that are very important to all of us!

Comments

[John Ford]

Can't wait till someone expands on srdf metro. What does my expanded vmware cluster look like with it? Can they be separate FC islands?? Do I need to extend my fc, etc.