Evolution of DevOps – or – Why your IT guys want to punch you in the face
The Old Way: Dev builds stuff without any operational knowledge - Throws stuff over wall to IT - Complains when it doesn’t work the way it did on their test box
Why traditional Hadoop (and indeed many analytics tools) <> DevOps – or – What is this 1993?
Clusters of monolithic, single-purpose servers are the “Insert something that used to be cool but isn’t anymore – like parachute pants or camcorders" of Operations – so very, very not happening anymore
The monolithic data-center construct died a long time ago, but new analytics apps are stuck in that past – time to move on
The middle Ground: things like VM Ware come into being and suddenly it is possible for Dev to have shiny toys too! They still don’t really understand ops, but at least their test rigs are better approximations of reality
DevOps Movement: Aligning the Dev and Ops mentality with technology to merge the ways of thinking:
Dev wants to move fast
Ops wants to be able to fix broken stuff
How did Ops get to where they are?
Shared things – storage, compute, now networking and even platforms
Virtualize ALL THE THINGS
We also recognize that big data is a team sport – we needed to support various user types including IT Operator, Data Scientist, App Developer as well as LoB user and the sponsoring executives on the business and IT
In the first instance we’ve decided to make the life of the IT Operator easy (months down to a week for standing up these environments)
Next we will focus on providing a true self service big data environment
Hadoop HAS TO BE SHARED NOTHING – why? Because it’s cheaper? LIES! It sure as hell isn’t easier for your ops guys! Why is it better?
CPU and Disk have to be together – NOPE
Performance of VM is bad – NOPE
And don’t even start on Cloud – it is just another word for Virtual Hadoop that runs NOT in your datacenter…