Consolidation was the number one reason most companies virtualized. Reducing the number of managed assets reduced costs significantly. The operational overhead of running a datacenter with 200K servers was insane.
Before virtualization, most enterprise servers ran at 5% utilization as a byproduct of traditional sizing. What do I need today, what will I need in three years when I’m due for refresh, and double all of that to avoid customer satisfaction problems. It was a recipe for datacenter sprawl, wasted floor tiles, lots of support tickets, and wasted power and cooling.
Together, we had to change our mindset about consolidation. It wasn’t about cramming more databases onto a single server. Consolidation was about breaking away from legacy relationships between workloads and physical hardware and starting to really consolidate with virtual assets. Arguments about putting “all of our eggs in one basket” became irrelevant, and building a single N+1 virtual (and consolidated) architecture was far easier than implementing 100 N+1 physical versions.
Most customers began with low risk infrastructure hosts, and soon enterprises were virtualizing the majority of workloads. But what about features? Remember all of this happened before things like VMware DRS, SRM, replication, etc. Features didn’t matter, because consolidation was the real value. Even if it only worked for 50% of my workloads because features were missing, so what? I still was able to reduce my cost in money and time significantly.
What about Software Defined Storage? I can’t tell you how many customers I’ve spoken with that buy servers with locally attached storage and then build out a SAN. Or lots of SANs. “Here’s the standard array config for my infrastructure workloads. I need 10 arrays.” It makes sense to have an enterprise class storage platform with the benefits of LUNs, replication, and single point of management. But what if you could add those benefits incrementally to the servers themselves with the local storage you already purchased and some smart software? Suddenly you can stop buying 10 arrays to deliver that value, and go from many management points to just a few. Finally – meaningful storage consolidation for the masses!
The customer I mentioned early in this series still put some workloads on bare metal, but the smart thing they did was a ‘virtualization first’ policy and addressed everything else as an exception. What if instead of consolidating 10 small arrays into 3 big arrays, you could turn them into a single physical array (or even zero physical arrays) and deliver the rest of the storage requirements with some software? You can. Check out ScaleIO. That is what it does, and what it has been doing for a few years now.
A ‘software defined storage first’ policy lowers the overall storage spend and frees time and budget to address the outliers who require hardware SAN advanced features (synchronous replication, etc) at a much lower cost.
While consolidating arrays sounds great, don’t you need lots of servers for software defined storage like ScaleIO or vSAN to work well? That’s where Pooling comes in, and is the next subject in the series.
Missed the first article? Click Here