K.I.S.S.’ing in the Data Center

Jonathan Kowall

by Jonathan Kowall

Hey Kowall - just like your last blog… um what?

Yeah I know, my titles are not the best.  I’m not sure about the origins of this idea but I know the Navy uses it and they’re pretty amazing at what they do.  But the idea I’m trying to convey here is that we think the simpler things are the more successful you can be and we want to bring this principle further into the data center.

Let’s level set, the data center is a combination of many complex systems working together to deliver some value.  Any opportunity to simplify one of those systems offers opportunity to increase the overall value.  One example is increasing the efficiency of the temperature control systems.  The better the room can be cooled the more servers the room can handle, increasing processing capabilities, which can deliver more VMs, which runs more applications, serving more customers in parallel or gets end of month billing done quicker so more invoices can be sent out today instead of tomorrow… and so on.

And since I’m the infrastructure guy, where can we make things simpler in the compute and storage space?  The easy answer is you get rid of stuff in the stack that you don’t need anymore.  For a long time the only way to deliver consistent performance between the compute and storage was a dedicated specialized network most often built on Fibre Channel switches.  Many people felt and still feel that IP could not handle the requirements of storage.  I think this now firmly falls in that 80/20 rule.  80% of the workloads run just fine on IP backed storage networks.  Those workloads have already been virtualized so we know they can collapse the compute layer and share server resources, let’s collapse the networking and storage layers and make virtualized compute environment even more efficient.

Make this (3 tier architecture):

K.I.S.S.’ing in the Data Center

Look like this:

K.I.S.S.’ing in the Data Center

This is what the industry is calling hyper converged infrastructure. IDC has a definition for it, Gartner still lumps HCI into the CI category.  CI or converged infrastructure is the ability to consume a three tier architecture as a single product.  CI is what VCE started with vBlocks.  If you have a hypervisor running on a server with external storage, you have CI.  HCI pulls that storage back into the server and aggregates the storage for shared use via software.  There are many discussions to be had around HCI but the one thing that is important here is how it simplifies the stack.  Keeping it simple here removes the extra management, cabling, power consumption, maintenance streams, training, run books, licenses, rack space, troubleshooting, support calls, etc.. that are needed in a 3 tier architecture. If you can remove all those things you get the efficiency to do more with less.  The requirement to do more with less is the first thing we hear from every customer.  And if you put less things in the stack, you can do more.