This post is, of course, a continuation of our top ten enterprise tech predictions for 2013. The first five predictions and the set up can be found here. Please also note that some more specific thoughts on the next year in the nascent and booming flash storage market can be found here.

Now here are our top five enterprise technology predictions for 2013:

Pure Storage
Virtualization (continued)

(#5) The server virtualization fight shifts up the platform stack – With #6, we talked about 2013 being the year VDI took off. At the other end of the spectrum, more I/O-intensive applications like databases will be virtualized in 2013. (Flash-centric storage is also key to this trend, as it provides the performance necessary for further server consolidation—Pure has had customers save enough money on reduced server software licensing fees that their storage was free.)

More interestingly, with the continued commoditization of the hypervisor (HyperV is becoming competitive while Citrix Xen and KVM are winning in the public cloud), expect the competition to shift toward software-defined automation infrastructure as OpenStack (current buzz leader), CloudStack, and Eucalyptus battle the alternatives from VMware, Microsoft, et al. (And while VMware no doubt maintains a dominant market position, those VMware renewals are going to look a bit more expensive, now that the offsetting power savings that justified initial purchases have disappeared from IT budgets.)


(#4) HTML5 wins for the enterprise’s homegrown app’s, but SaaS vendors go native – The touch screen UI strikes me as a profound innovation, rating up with the GUI and hyperlink in impact. With the proliferation of tablets and smartphones, businesses are going to port their homegrown applications to touch screen UIs (although non-core app’s should go SaaS). For most businesses, supporting multiple native platforms (Java/Android and iOS) will be too painful, and HTML5 functionality will be well good enough (especially when supplemented with code generation tools from HTML5 to native app’s such as Appcelerator). Expect SaaS application vendors, however, to follow the trail blazed by consumer tech in delivering ever richer mobile UIs via two or more native app’s. (Here’s another one I got completely wrong—to my naive thinking, HTML had more reason to win on mobile devices than it did on the PC. Millions of app’s later, we have yet more evidence you should take my prognostications with a large grain of salt.)

Everything as a Platform (EaaP?)

(#3) Software evolves into platforms – Enterprise software is ever less likely to be used in isolation. Increasingly, software must provide

  • Web APIs for integration with other SaaS and home-grown systems – By moving enterprise app’s to the cloud, the hope is to solve integration issues at scale, and actually lower the hurdles to integrating best-of-breed applications (recall the behind the firewall mess that was EAI ?);
  • APIs or REST bindings for software-defined management (i.e., DevOps everywhere);
  • ETL bindings for harvesting application data to be used for analytics, whether for structured (small) or unstructured (big) data; and
  • Security interfaces so that firms like Okta can provide authentication and authorization in the cloud.

The consumer Web has, of course, has already been delivering platforms, creating the potential for mash ups of consumer and enterprise systems. (Gracias to @quixotic for inspiring this prediction.)

Flash memory

(#2) Server flash poised for the mainstream – With 2011’s FusionIO IPO, the industry got some visibility into leading public cloud deployments (Facebook, Apple) that are using flash memory for server-local storage. Moreover, Google has been leveraging flash in the server tier for Instant Search, while Amazon is using server-based SSDs under their AWS DynamoDB (key/value store in the cloud.) With the expectation that all servers are going to ship with flash as well as DRAM on board, it is time for systems software (operating systems, virtualization platforms, databases) to start to be rewritten to take advantage of local flash cache. This will have the effect of further commoditizing flash in the server tier, and driving out proprietary caching solutions from the storage vendors. (The underlying storage ultimately cannot do nearly so optimal a job of leveraging server-local hardware than can the server software.) For most workloads, however, expect the data store of record to remain in a shared storage tier (see #1 below).

(#1) But for the data of record, shared flash storage will trump DAS – Within the public cloud, server-local (a.k.a. DAS) disk turned out to be well cheaper than deploying SAN/NAS storage. The challenge was crafting the software infrastructure necessary to ensure fault tolerance, data protection, performance, and manageability across 1000s of semi-reliable hard drives, both within and across data centers (which is after all the raison d’etre for traditional disk arrays). For many workloads, however, shared flash storage is more cost effective than the DAS alternative! The key is inline deduplication and compression. Such inline data reduction allowed Data Domain and their competitors to offer the media customers wanted (disk) at the price point of the one they wanted to replace (tape). Well, for performance storage, flash is the new disk, and deduplication and compression done right allow all-flash solutions to be price competitive with arrays of mechanical disk. In 2012 we saw proof that deduplication and compression can be done in-line without hurting performance. (In fact with writes relatively slow on flash, data reduction accelerates performance in most cases by avoiding rewriting the same data over and over.) Today, you wouldn’t buy a disk back-up solution without data reduction, and in 2013 the same will hold true for flash-centric performance storage. (One might argue for dedupe in the server tier to close this gap, but server dedupe is relatively ineffective because the data set is too small. And dedupe and compression are so algorithmically intensive that they are deserving of a dedicated processing tier, which is more easily afforded in shared storage.)

And there you have our top ten enterprise tech predictions for 2013. In fairness, some of these will ultimately take longer than 2013 to play out. And sorry no consumer tech predictions—I’m not smart enough.

Thanks for listening, and again, Happy New Year from all of us at Pure Storage.