Andrea Danti - Fotolia
As growing volumes of big data increasingly factor into business strategies, organizations face infrastructure bottlenecks and other technical pain points that come with them -- specific business problems that few one-size-fits-all big data technologies can adequately satisfy.
That puts CIOs at a crossroads, and they must choose: Should they go ahead and invest in products that might not be best suited to their organizations' specific needs, or look at other platforms and strategies to support their analytics, such as NoSQL databases, open source and cloud? More daunting yet, should they build these capabilities in-house, use a blend of products from different vendors, or combine purchased and custom-built functionalities?
The number of big data technologies available for the picking is overwhelming, and before investing in any of them, CIOs must be clear on what big data really means, especially with regard to their business' unique requirements. In this #CIOChat recap on the topic of architecting for big data, our tweet jam experts from Centerstone Research Institute -- Tom Doub, CEO, and Wayne Easterwood, CIO -- as well as SearchCIO editors and followers, sounded off on how enterprise CIOs can sort through the building-versus-buying dilemma for big data architecture.
When should I invest in a single stack or best-of-breed for my big data architecture?
When asked under which circumstances CIOs should invest in a big data technology stack that consists of products from a single vendor versus create one that consists of solutions from various vendors, many #CIOChat participants were quick to respond with use cases for a best-of-breed architecture. Essentially, the decision comes down to the diversity of specific business requirements and the number of clients and sites, as well as other nontechnical factors:
SearchCIO executive editor Linda Tucci brought up a caveat when looking at only a single vendor, a point of view shared by Forrester analyst Brian Hopkins in our recent big data architecture coverage. "There are no nice, neat packages you can acquire," he said, if you're looking to solve today's breed of data problems:
A1. Single stack isn't there yet and the tools aren't mature enough...easier to build best of breed currently #CIOchat— Brian Katz (@bmkatz) September 24, 2014
Hopkins recently shared similar sentiments regarding single-stack vendors, particularly as they relate to megavendors such as IBM and Oracle. Despite how complete their stacks seem, "I don't think it works as nicely as they think, and neither should the buyer," he said.
When should I buy a product versus build big data functionality in-house?
Deciding between buying an entire stack of big data tools from one vendor as opposed to multiple providers is one thing, but when is a custom build the best fit? Doub pointed out that decisions around scale and security might tip the scales toward a custom build:
#CIOChat A2: Personally I'd rather buy than build when I can find a solution that meets the business need. Then Invest time in customization— Tom Doub (@tomdoub) September 24, 2014
Participant Gloria J. Miller listed examples of complex business requirements that could complicate the build vs. buy decision:
In addition to ensuring that internal resources and talent are available to build these big data functionalities in-house, it's also important to focus on business outcomes before moving forward with an architecture decision, according to Stephen Laster, chief digital officer of McGraw-Hill Education, in a recent SearchCIO piece.
Once that's been determined, "We take the implementation apart into small pieces and we go and see for each piece -- is this something that's commodity or that [has] been solved by the market? And if there are pieces that are unique or haven't been solved well before, we invest in building those," he said.
In addition to looking at particular big data tools and how suited they are to complex business needs, CIOs also must look at the vendors themselves, #CIOChat-ters said:
What factors do you consider in determining whether to build, buy or blend a big data architecture? Sound off in the comments section below.