“Latency kills revenues.”
If you had the pleasure of attending VMworld in San Francisco this week, you probably heard this in the keynote presentation. And it’s absolutely true. Let me explain...
The unprecedented growth of data is causing application, server, and desktop sprawl. Both server and desktop virtualization technologies have been deployed to relieve the strain that this growth creates and its effects on the compute and storage infrastructure. These virtualization schemes, however, are negatively impacting network throughput and user productivity.
Users attempting to access their applications suffer excessive wait times caused by boot storms, decreased performance, and increased latency. And it gets worse as virtualization is more broadly adopted. As the number of platforms an organization manages grows, virtualization density (the percentage of systems that are virtualized) also grows, putting more strain on the underlying infrastructure and dramatically impacting the latency of each application, each virtual desktop, and each user.
Traditionally, magnetic hard drives (HDDs) have been deployed in virtualized environments to accommodate this kind of growth and have served as the shared storage that virtual servers access. Because of the inherent design of (even the fastest) HDDs, latency issues and their resulting effects are compounded. The I/O blender and HDDs’ inability to handle the load causes boot storms, which results in an idle, less productive staff. They can no longer be relied on to provide a productive and efficient virtualization environment. With the advent and increasing maturity of solid state drives (SSDs) and flash memory technologies (which have given us near price parity with magnetic HDDs), those negative effects no longer need to be a factor.
With virtualization becoming commonplace in the data center, latency and performance issues have a significant effect on virtual system performance and their users. Deploying Micron SSDs in platforms hosting either desktops or applications running in virtual machines can result in a more responsive application; better throughput; lower OpEX (floor space, rack space, power, and cooling); lower CapEX (because fewer hosts are required); and, most importantly, increased productivity and, ultimately, revenues.