Sep 042013


I recently attended EMC Forum, where some fairly impressive figures were rolled out relating to what EMC has spent on R&D and acquisitions over the last 10 years*. Backed up by their corporate profile, those figures are:

Our differentiated value stems from our sustained and substantial investment in research and development, a cumulative investment of $16.5 billion since 2003. To strengthen our core business and extend our market to new areas, EMC has invested $17 billion in acquisitions over the same period and has integrated more than 70 technology companies.

So over ten years, that’s almost 35 billion dollars invested by EMC into products, technology, skills and innovation. No matter how you slice that, it’s an impressive commitment towards capabilities growth and leadership in data.

EMC demonstrates that commitment yet again with the Speed2Lead launch in Milan.

I’ll be in Milan for the launch, a guest of EMC**, and I’m looking forward to getting some additional details of the new VNX range in particular – but the information EMC has whet my appetite with is pretty impressive so far.

XtremSW Cache 2.0

v2 of XtremSW Cache sees it really take off via:

  1. Integration with EMC Arrays;
  2. Working with:
    • IBM AIX;
    • Server flash;
    • Oracle RAC (coming in October);
    • VMware vCenter Integration;
  3. XtremeSW Management Center will provide strong control and efficiency options – a single point of management when deploying multiple cache instances.

The EMC array integration I think is going to prove to be particularly popular:

  • VMAX:
    • Strong support out of the box for VMAX integration – XtremSW Cache will be manageable from within Unisphere so that administrators can choose which LUNs should be cached based on trending analysis;
    • Prefetching entire tracks – so if you’ve got a read-intensive application leveraging VMAX storage you’ll get a considerable performance boost – IO rates can increase by 25%;
    • Cache coordination (optimised read miss) moves the read cache tier to the host. That allows the array of course to use resources elsewhere, which can lead to increasing IOPS by as much as 2.5 times.
  • VNX:
    • For the VNX series, too, XtremSW Cache 2.0 starts with Unisphere Remote, which will recommend LUNs to cache based on trending information. (EMC say over time the integration between VNX and XtremSW Cache will be on par with the VMAX support);
    • Additionally, you’ll be able to monitor performance and health, as well as the configuration for XtremeSW Cache all from one location.


VNX has jumped to being highly multi-core aware in its processing and capabilities. While VNX systems have used multi-core CPUs for a while, the architecture wasn’t taking full advantage of those powerful cores. In particular, depending on the workload involved you might see one or two cores particularly heavily utilised, and others mostly idle. This new design, MCx, sees EMC VNX arrays fully utilising up to 32 cores, with significant boosts to performance: full symmetric multi-processing. Some of the performance improvements being cited are going to be very popular:

  • 6.3x faster for host-side IOPs;
  • 4.9x better on file serving IOPs;
  • Capability of delivering 1 million IOPs deliverable at sub 1-ms latency.

No, there’s no typos on that last point.

Capacity, Performance or Both?

  • With support for up to 1,500 storage devices (SSD, HDD or any mix you want), VNX is offering serious capacity and performance capabilities. Depending on your layout of HDDs vs SSDs your performance and capacity will scale considerably in one of several directions – the more hard drives, the closer you’re going to be to hitting maximum capacity; the more SSDs you have and the closer you’re going to be to hitting maximum IOPS;
  • Of course, if you want to get maximum bang for back with Flash (and Fast), and your data profile supports it, deduplication may very well be a good thing to turn on;
  • VNX is moving to active-active storage-processor capabilities. That’s starting with traditional, thickly provisioned LUNs, but will over time move to encapsulate the rest of the VNX functionality. This is a big change – and a big win for customers who want higher performance but for pricing considerations need to stay within the VNX range;
  • The new VNX 8000 being released in Milan is an absolute beast, with scaleability of to 6PB and support for running a workload of 8000 virtual machines, it’s going to be a major boost to datacentres and cloud environments.

There’s more, of course – much more. I’ll be particularly looking forward to some discussions with EMC folk regarding the performance increases we’re likely to see out of the new VNX MCx architecture when it comes to NDMP.

EMC AppSync

Demonstrating its continuing focus on protection and recovery, EMC’s AppSync system offers a new storage focus to protecting critical applications. With a configuration system based around SLAs, you can define AppSync protection strategies based on Gold (Concurrent remote and location protection), Silver (Remote protection) and Bronze (local protection). Of course, you can change those names to suit your environment, but the Gold/Silver/Bronze often works quite well to define protection levels.

The advantage of that of course is that once you’ve got those SLA policies defined, deciding the protection strategy for an application comes down to picking whether you want Gold, Silver or Bronze…

The applications covered by AppSync are an important collection:

  • SQL Server 2008;
  • VMware NFS file replication ;
  • Exchange 2010 and 2013 (EMC cite being able to protect a 22TB Exchange database in under 5 minutes);
  • VMware Datastore Protection.

[not closing] Thoughts

A few years ago during a blogger forum organised by SNIA-AU, I was part of a group of people who visited EMC, IBM, HDS and NetApp. One of the most telling things said during the day was from Clive Gold at EMC Sydney. It was such a simple statement that I’m trusting myself to quote him verbatim after all this time:

People buy storage for capacity, but they upgrade storage for performance.

Developing big storage is almost a no-brainer. Think about it: your enterprise could attach 12TB of DAS USB-3 storage to most servers for less than $1000 per server. Totally crazy stuff, of course – the management overhead alone would be a nightmare. But that’s the thing: if you’re not worried about performance, storage is easy.

Speed2Lead shows me that Clive Gold wasn’t just speaking from a marketing statement – it’s something EMC fervently believes in: capacity is one thing, but delivering on performance is more important, because it’s performance that customers notice. As a storage admin, you’re not going to get any pats on the back that you’ve got 1PB of storage free and unallocated if the 500TB you do have allocated can’t service the IOPS requirements of the business.

EMC are calling this Speed2Lead … and they’re certainly speaking the truth.

Stay tuned for more. Following the event when I’ve had more time to digest the information further, discuss it with colleagues and customers, I’ll be posting some additional details.

For now, check out EMC’s launch page for Speed2Lead.

* Some companies tend to be dismissive of money spent on acquisitions, but I have a different perspective – companies that aren’t willing to acquire are steadfastly following the “not invented here” approach. In order to survive and grow, a company has to be willing to both invest in research/development and other companies that provide synergistic product sets or skills.

** Disclaimer: EMC flew me from Melbourne to Milan via Qantas and Emirates economy class, and put me up in the Westin Hotel. There were some transfers and a couple of meals included.

Sorry, the comment form is closed at this time.

%d bloggers like this: