There’s been a recent flurry of activity from search vendors (and those larger companies that have been buying them) around the theme of Big Data, which has become the fashionable marketing term for a sheaf of technologies including search, machine learning, Map Reduce and for scalability in general. If anyone impertinently asks why company X bought company Y the answer seems to be ‘because they have capability in Big Data and our customers will need this’.
Search companies like ours have been working with large datasets since the beginning – back in 1999/2000 the founders of Flax led a team to build a half-billion-page Web search engine, which as I recall ran on a cluster of 30 or so servers. Since then we’ve worked with other collections of tens or hundreds of millions of items. Even a relatively small company can have a few million files on their intranet, if you count all those emails, customer records and Powerpoint presentations. So yes, you could say we can do Big Data – we certainly know how to design and build systems that scale.
However it makes me nervous when a set of technologies that could (in theory) be used together are simply lumped together for marketing purposes as the Next Big Thing. The devil is as always in the detail (and the integration) and it’s important to remember that just because you can fit all your data into a system doesn’t mean that system will help you make any sense of it. A recent term for unstructured data (which of course us search developers have been working with for decades) is Dark Data, which implies that it is mysterious and hidden – but that doesn’t mean it has any actual value. Those considering a Big Data project should be aware that in any computer system GIGO is still an issue.
