Pat Gelsinger, EMC's COO for Information Infrastructure Products, recently imparted a new vision for the future of IT to a group of analysts gathered in Hopkinton, Mass.
Gelsinger, who now manages some of the company's crown jewels like the storage products division and is EMC's executive sponsor for VMware, said EMC is out to change the structure, technology, and possibly the behavior of the IT community.
That's a tall order to fill even for someone as obviously energetic and experienced as Gelsinger, and so if you react to that statement with a measure of skepticism, you're not alone.
However, it is well worth looking closely at what's at the core of this vision. What he and EMC really want to do is to challenge the notion that applications and application data are essentially stationary. Critical business application software and the data sets that support critical business applications are difficult to move around quickly and particularly so over great distance. As fast as light speed is, communications networks as we know them are still a bottleneck. Latency is still an issue. Applications stay put. On the other hand, imagine an IT world where an application could wake up in North America, then move to Asia, then on to Europe, continually circling the globe and following the sun.
What EMC plans to offer by way of a product that turns vision into reality is a distributed data caching appliance (V1 of the vision). Perhaps the concept is better described as a network of geographically distributed data caching appliances that maintain coherency and data consistency with respect to one another. This network is capable of spanning great geographic distances between network nodes. The term Gelsinger uses to denote how individual caching engines relate to one another and to an application environment supported by VMware is "federation." The result is that EMC has, in his words, solved the problem of cache coherency over distance. That, combined with VMware's ability to move applications, yields application fluidity. It also, according to Gelsinger, creates the illusion of from an application perspective of having "petabytes" locally when they could actually be physically living in storage devices thousands of miles away. One could think of this as adding data abstraction on a potentially global scale to VMware's ability to do server abstraction.
I hear some of the storage industry cognoscente out there saying that they've seen distributed data caching before. Yes, Gelsinger did allow that the underlying technology is based on an EMC investment in, and the subsequent acquisition of, a somewhat obscure Canadian big-box storage array company named Yotta Yotta (as in yotta-byte scale storage). Nonetheless, YY was years ago and couldn't get market traction, even when it came to shows with the cute fuzzy dice that chirped "Yotta Yotta." EMC today can apply it's highly respected marketing savvy to getting distributed data caching into mainstream IT architectures. Version 1 is in beta test status now.
Gelsinger sees distributed cache coherence as fundamental to the private cloud in that it supports around-the-clock data availability, the ability to aggregate and pool infrastructure and application resources on a grand scale, and the ability to geographically move critical applications at will.
Could this be game-changing, disruptive, and (insert you favorite superlative here)? I think that for sure, users of this technology combined with VMware can now challenge application vendors to come up with new licensing models--as if VMware wasn't challenging enough for the likes of Microsoft and Oracle. Remove application and data dependency on physical location and, in at least some situations like the "follow the sun" example above, you remove the need for multiple application licenses for multiple locations. I also believe that the federated view of distributed IT resources will indeed impact they way enterprise IT architects design systems. But as I mentioned, I also think that coherent, distributed storage caching will challenge some long-time EMC partners in the both the applications and networking spaces.
But that's just for starters. As an application administrator or application architect, I can now think in terms of having a data abstraction layer underneath a virtualized OS layer. I now no longer have to worry about where the data is physically. I could have multiple, globally distributed instances of an application using the same data repository no matter where the application instance lives. If you want to call that a private cloud, I won't argue the point. What I think the distributed cache network does do is create a dedicated storage cloud that can be given the security and manageability attributes common to storage arrays and supports the creation of highly mobile applications running on both stationary and mobile platforms.
Bottom line: EMC has to demonstrate that distributed cache coherency really works and delivers the presumed benefits. If it does, the superlatives are earned and those who struggled at Yotta Yotta to realize their original vision will not have struggled in vain.