I/O performance solutions

Virtualization technology fixes a lot of problems with server utilization and desktop management. Unfortunately, it ruins storage utilization. Welcome to the “I/O blender”.

How, exactly, is virtualization breaking storage? One physical server can easily support upwards of 8-12 virtual servers or 50-70 virtual desktops, each of which generates its own I/O stream.

As a result, the I/O patterns that virtual environments generate are far more random and write-intensive than in a physical server environment – which translates into massive degradation of storage performance.

The “I/O blender”

Virsto calls this phenomenon the VM “I/O blender.” It’s bad in virtual server environments, and incredibly bad in virtual desktop environments. The common response to this problem is to over-provision storage for each virtual machine, which gives some relief. But multiplied across typical enterprise environments, this brute-force tactic requires the purchase of enormous storage resources that are barely utilized – 20-30%, enterprise-wide, is not uncommon.

Why storage can’t keep up

Legacy storage architectures, spinning disk in particular, perform at their worst with very random, very write-intensive I/O patterns. The rotational latencies and seek times of spinning disk start to dominate data transfer times, and input/output per second (IOPS) per spindle drops like a stone.

Solid-state drive (SSD) storage doesn’t necessarily help much. While it is very good at handling reads and sequential writes, SSD is only slightly faster than high-performance spinning disks at handling random writes – unless it is front-ended by additional technologies designed to minimize the randomness of the workload. SSD is still priced at a premium, and the black boxes and appliances required to improve its performance add further cost and management complexity.

What (not) to do?

The three fixes to the I/O blender that most IT organizations consider are:

• Adding more disk spindles
• Upgrading to a higher-end, more expensive class of disk array
• Buying SSD

...all of which drive storage costs through the roof.

The I/O blender has implications for snapshot/clone technology as well. With conventional snapshot technologies, the more snapshots and/or clones you create, the slower they all run. Since you’re already chasing performance, you might look at very high priced, enterprise-class, array-based snapshot technology. No surprise here – storage costs go up.

Chasing performance? You’ll never catch it

Finally, native thin-provisioning virtual disk options offered by various hypervisor platforms might be attractive if storage consumption is pushing up costs. But these options perform much more slowly than fully allocated virtual disks until they reach a “steady state.”

Again, since you’re already chasing performance, and you’re adding or removing VMs all the time, you’ll probably select the fully allocated storage option if you can’t afford the enterprise-class array. This means you are forced to trade poor storage capacity utilization in order to get good performance.

The bottom line: the VM I/O blender drives storage performance down and storage costs up.

Virsto has the answer

But all is not gloom and doom. Virsto has the solution to the I/O blender problem. At its core, our Virsto storage hypervisor has a unique I/O system designed specifically for virtualized environments. It speeds installed legacy storage and accelerates SSD – and allows you to use a lot less of it to get the same results.

With Virsto, you can achieve up to 10x better performance and up to 90% better utilization of the storage you already have. Here’s how we do it .