Big Data Hardware: 2012 Winners and 2013 Outlook

It is time to close the door on 2012 and open the door into 2013 in the realm of big data. While many such closings and openings have already been published, I decided to wait until we had fresh 2013 spending intentions data, which just came in. Subscribers will be able to access the data later this month in a report entitled 2013 IT Spending Intentions Survey. I also wanted to clear my head from the holidays in order to look backward and forward with a fresh perspective. Therefore, every day this week I will take a segment of big data and will render a final 2012 reckoning with a related 2013 prognostication in the order of: Monday – hardware, Tuesday – database, Wednesday – software solution, Thursday – cloud, and Friday an overall “ingest to insight” 2012 big data vendor of the year and macro-trends for 2013.

2012 Big Data Hardware Vendor of the Year Finalists: Cisco, EMC, Oracle

After fits and starts, Cisco broached the converged infrastructure market in 2012 with a well-thought out approach of spanning the market through a variety of targeted partnerships and packages. In the big data space, it produced several UCS offerings specifically for Cloudera, Greenplum (EMC), Microsoft, Oracle, and SAP big data solutions. Cisco also took big data steps to work with integration vendors such as Informatica and Talend, “Not Only SQL" database vendors such as MarkLogic, and stepped forward as one of the handful of certified SAP HANA resellers. Cisco put itself squarely on the big data map, perhaps to surprise of many competitors, during 2012.

It is a little difficult to find clear boundaries between EMC the storage vendor (especially Isilon in the big data context), EMS the information management company, EMC the big data platform company (Greenplum), and EMC the parent of the world's most important virtualization vendor (VMware), but suffice it to say that EMC has placed itself into play as a key strategic supplier for big data in a big way. While the natural reaction to EMC, in terms of hardware, is to focus on storage, the EMC Greenplum Data Computing Appliance (DCA) was one of the first and most well-designed dedicated big data appliances. While sales execution may not have quite measured up to EMC’s own standards, I found it telling that EMC really "gets" big data by the fact that DCA is equally as optimized for memory as for storage.

Oracle, in one of the more talked-about moves in the industry during 2012, added its Oracle Big Data Appliance to its list of engineered systems, and promises to offer its appliance "as a service" through its cloud in the near future. About a year ago, Oracle Big Data Appliance, by not just adding Cloudera software but also the recently updated Oracle NoSQL database and open source R, pioneered a new standard for how to craft a big data appliance. My only wish is that Oracle had tossed in their Hadoop connector gratis versus a paid-for add-on. In 2013, Oracle may need to respond to big data plus data warehouse in the same appliance because of the likes of Teradata, but regardless, many of those who raised eyebrows early in 2012 at the Oracle Big Data Appliance are now part of the following herd.

Winner: Cisco, for not only creating a thoughtful vision and plan for big data infrastructure, but also and more importantly executing on it in both direct and indirect channels during 2012.

Big Data Hardware 2013: Don’t DIY

One of the misleading promises associated with implementing Hadoop has been “inexpensive, commodity hardware.” At least for enterprises that take their business intelligence and analytics seriously, ESG does not recommend “do it yourself” (DIY) when it comes to spinning up and managing big data hardware/infrastructure. There are two main reasons:

  1. Few IT shops, and virtually no LOBs, possess the chops to spin up and manage their own infrastructure for a full-scale and rapidly growing production big data environment, and those that take that risk without the skills may quickly head over the edge of their own big data economic and delivery cliff. Yes, some true Web 2.0 companies possess personnel with such expertise, but brick-and-mortar heritage enterprises typically do not.
  2. Despite all the hype around an abundance of spending available to big data projects, ESG doesn’t buy it. Based on our most recent data, while BI/analytics will receive somewhat more than its fair share of IT budget, the increases are only incremental—with half of companies planning to spend about the same on BI/analytics in 2013 as they did in 2012. Those that believe CMOs and LOBs will chip in massive amounts of CAPEX for big data are also overly optimistic: There will be more investment coming from the business, but the serious projects (not the experimental ones) will still answer to the CFO, just like any other IT-oriented project, regardless of budgetary source. We believe, therefore, that serious projects should, and will if they plan to be successful, allocate a healthy, predictable, and dependable portion of budget for infrastructure—suggesting appliances that fit that criteria.

What this implies is the continuation from 2012 of the shift from DIY to appliances and cloud for big data infrastructure during 2013. It is no accident that Hadoop standard bearers such as Cloudera, Hortonworks, and MapR have all signed partnerships with appliance providers. Going into 2013 Cisco, Dell, EMC, Hitachi Data Systems (HDS), HP, IBM, NetApp, Oracle, and Teradata will be the obvious appliance players, either directly or indirectly, for enterprise big data investments, with niche players such as Supermicro taking on a unique position between the enterprise appliance and the DIY crowd. I am carefully bullish about HP making a stronger push on the hardware front of big data during 2013, and likewise believe HDS will gain share if they expand their partnerships to include more pure play big data solution providers. Dell and NetApp are also certainly not to be trifled with—both have augmented and tuned their big data commitments for 2013, and carry momentum developed during latter 2012.

The trick in all cases is to offer an appliance tuned for a particular big data software solution with applicable proof points; there has been no generalized big data infrastructure with universal appeal, and I don’t expect one to appear in 2013. From an execution perspective, what is equally as important is, when applicable, big data software solution and infrastructure providers should walk hand-in-hand into deals and bids. Enterprise big data buyers want the help and assurance from all primary vendors involved, whether hardware, software, network, or hybrid—and services.

While ESG predicts a healthy, not skyrocketing, albeit competitive year for big data hardware, looming not far in the distance is the gathering big data cloud of Amazon Web Services, threatening to steal buyer thunder from appliances. Cloud will take on an increasingly role in 2013 for big data infrastructure, just as it did during 2012. Wouldn’t it be fun if Amazon Web Services did something crazy like package up some of its infrastructure and solution partners’ software into a big data private cloud appliance with high speed data connections between public and private? No, I am sure Amazon Web Services would never consider yet another disruptive move in big data, just like Coach Belichick would never adjuist his defensive or offensive schemes during halftime of a Patriots’ game.

Topics: IT Infrastructure Data Platforms, Analytics, & AI Cloud Services & Orchestration