The 4th Annual EMC Innovation Showcase has come and gone, and an idea that my co-workers and I proposed was chosen for funding. It is quite a difficult task to survive the whittling down of 1500+ submissions, but we had a distinct advantage this year: the CSX technology.
One more thing: the mid-summer acquisition of GreenPlum didn't hurt either.
Anatomy of an Idea
Chuck Hollis mentioned yesterday that I am a big fan of collaborating with contrarians and risk-takers both inside and outside of EMC (thanks for the kind words, Chuck!). This particular idea was no exception.
I was hanging out last December with customers at ITAC, the Industry Technical Advisory Council (described here). The topic of "big data analytics" came up. This discussion was the first time that I heard about GreenPlum's approach to scalable analytics, and I returned to EMC to learn more about our efforts to launch a GreenPlum appliance.
At the same time, my group continued to experiment with the CSX technology, componentizing critical data path assets so that they can run on any platform. I wondered if GreenPlum had an easily-downloadable component of their own that I could play with. I found out that they did. I started to ponder the combination of the GreenPlum component with the catalog of toys that we have in our repository: distributed caching algorithms, RAID, file systems, object-stores, etc.
I asked my boss for a one-month intern during the winter break (I didn't tell him why). I asked the intern to build a Virtual Machine with GreenPlum inside, and combine it with some of our other technologies. I hoped that he could finish in one month.
He finished in less than a week and then asked me for more work (we hired him full time. But that's a different story!).
Scale-Out
I now had a prototype VM that could accept database I/O (along with some other nifty things supported by different internal EMC assets). I began an internal search for skunkwork-worthy co-conspirators. I decided not to use people in my own business unit because we were already collaborating on so many backroom skunkworks that their managers would probably sniff this one out. So I decided to try and find some help outside of my group. I looked around EMC for some people that had a very different mindset (and skillset) from my own: the massive scaleout of virtual machines. I patched together volunteers from both the West Coast and East Coast of the U.S., and also pulled in some expertise from Russia as well.
And off we went.
Within a few weeks (and actually a few phone calls to the pre-acquisition GreenPlum team!) we had scaled out our one-week prototype to be a distributed GreenPlum database running alongside CSX-ported EMC assets.
In June we submitted what we had done into the annual innovation contest.
And Then EMC Bought GreenPlum
To put it mildly, that really helped the cause of the idea. Actually, I had never been more confident that an idea would be chosen for funding.
In the weeks that come we will articulate what resources we need to take the idea to the next level (I plan on asking for a fully-loaded vBlock, can I get an AMEN from the vSpecialists out there???).
The summary here is that the component assembly approach leads to an enormous amount of permutations when it comes to combining software assets. In years past, whenever EMC acquired a company, it was a difficult proposition to integrate that asset into our storage systems. For example, we'd have to port GreenPlum to Windows (for CLARiiON), or into DART (Celerra), or into Symmetrix, Centera, etc.
But now we have options. We can "CSX-ize GreenPlum", which would allow it to be inserted as a re-usable component in multiple product lines (this is what we do with RecoverPoint assets, for example).
Or, we can take the approach above and integrate GreenPlum as-is into a newly-assembled and purpose-built product.
I'm glad the proposal was chosen for funding.
It makes it a lot easier to tell my boss what I've been up to.
Steve
Twitter: @SteveTodd
Comments