The latest buzzword in virtualization, yet for me the technology it describes is old hat (in the I.T. world old hat isn’t all that long ago). Let me explain…traditionally a ‘converged’ system is simply a combination of 2 (or more) great bits of technology with very different roles combined into one. An example of a converged system is VCE, where I still think of it as the ‘V’Mware, ‘C’isco and ‘E’MC alliance:
- VMware – provides the virtualization function
- Cisco – provides the network and server layer (with a little help from Intel!)
- EMC – experts in storage, so you can guess what they provide!
[With Intels contribution it should really be called ‘ViCE’ 😉 ]
Together that means a joined up system, a VBlock, that you simply deploy then use as a converged compute system. Want more performance? Then add more CPU or RAM or Storage…
…and that is where Hypercovergence differs. Instead of isolated blocks of converged compute you have ‘blocks’ that can work together and scale out, want more performance? Add another block to an existing one via a network cable and BOOM! You have more power. Add 10 blocks. Or 50!
Why did I say it was ‘old hat’ I hear you ask? Well, that’s exactly the way MongoDB works, it scales out in pretty much the same way. When your databases reach a certain size and you need more oooomph, traditionally you would need to migrate the workload to a beefier machine. What if there was a better way, one perhaps that could make use of some of the spare CPU cycles available in an existing machine or one that allowed a redundant piece of kit become useful again? I’ll explain with pictures:
Poor chap, a lowly P75 system crunching away at that data. Need to urgently number crunch the number of stars in the universe and the probability % of habitable planets? Well you need more ooooomph, so scale out like thus which MongoDB has been doing for years (since 2007 while VMwares bitter rival Nutanix first released their Virtual Compute Platform in Q4 2011) :
OH look at that, my Xeon buddies have joined in the game. Now with all that Quad core Hyperthreading with a bit of clever sharding on the MongoDB config you’ll be finished calculating in no time.
So that’s what Hyperconvergence is pretty much. The ability to add more by simply using Ethernet. No need for messy transitions or complicated integration paths and reams of consulting days. Buy it, plug it in, switch it on, use it.
Of course Hyperconvergence is a little more than my simplistic analogy, it’s changing the landscape for virtualization and storage. Previously you would need to integrate 4 or 5 vendor offerings to get your virtual compute platform running. Now you don’t have to. Buy just one (very expensive) hyperconverged box and spin up 100’s of workload VMs to do your grunt work. Potentially you can reduce significantly the number of racks of servers you have, and power/storage costs anywhere between 20 to 80%. Impressive stuff
The following are ones to watch:
Nutanix – possibly more famous for rowing with VMware
SimpliVity – simple isn’t it! Get a free ‘For Dummies’ book here
PernixData – just like The Flash, these guys are fast
I wonder what NetApp are thinking right now…?
Probably enjoying the ever growing spat between VMware and Nutanix, my buddy Chuck started it all with this > 10 reasons why vmware is leading hyperconvergence