10 comments

[ 5.9 ms ] story [ 40.4 ms ] thread
huge geek karma, but i suspect bound to fail

this is basically a consulting gig...i doubt they intend to try to compete head-head with the major web cos with their giant colo farms. and who would fund a new massive server farm after famous flame-outs like cuil?

what is the market? those companies technical enough to know hadoop exists, know they need it, yet incapable of implementing it while being cash-rich enough to pay someone else to do so. are there even five potential clients?

i love hadoop, its awesome...but my guess is that like mysql, apache etc, people will just look to redhat etc to do the integration on private networks, or just use it as a service on something like amazon's platform

I do not quite understand why people build hadoop and lucene with java. Is a python & c++ implementation of mapreduce a better choice?
you can write hadoop client tasks in any language you like, afaik
Because Java is an incredibly powerful tool that works across a large selection of platforms with minimal effort.

And, Hadoop is only the infrastructure, they have APIs for C, C++ and shell for the applications. http://wiki.apache.org/hadoop/FAQ

Is this the result of that cyprus trip?
Really cool. Hadoop is pretty mature and Yahoo's shown it works well on large clusters, but it can be a bear to install, maintain, and learn. I'm sure many companies wouldn't hesitate to pay a few bucks to get another company to help them get up and running faster.
after doing consulting i can tell you that you need a very big client list to make a go of it full time. you need companies like telcos, insurance companies, food industry...not just core tech companies...that won't last long. i just don't see hadoop being a pressing need in the wider consulting space

then again, maybe they just want to get their names out there...as long as they keep overhead low, might be worth it

(comment deleted)
"Their welfare depends on the usability and reliability of the product being mediocre or worse. "

Some shops have an attitude that is the opposite of "NIH" and "DIY", preferring to source existing experts instead of building in-house talent. There are few things that are easy in the edge cases, and fewer that are easy and worthwhile. With Hadoop, however, there is a lot of money to be made supporting the edge cases.

EDIT: I replied to a comment that has since been deleted? Neato!