There is a growing conflict surrounding virtualization that has nothing to do with its relationship with cloud computing. While IT strives for higher virtual machine density to improve TCO and ROI that increase can cause depletion of compute resources at a much faster rate. This effectively decreases efficiency of the data center and tanks the TCO and ROI that made virtualization attractive in the first place.
Perhaps worse for users and those responsible for resolving application performance issues is that the depletion of resources has a direct impact on application performance and, ultimately, availability. I/O intensive operations end up bogged down in the virtualization translation layer while saturation of the network card and software switches turn the network into a bottleneck, reducing available bandwidth, increasing latency, and degrading application performance.
IT staff end up caught between budgetary demands and unhappy users, with virtually nowhere to turn. You’ve been here, I’m sure. It’s like you’re in the Twilight Zone, transported back to the situation you were in before you started your virtualization venture. You’re out of resources.
There is a solution: Offload the most intense processing to external solutions. Doing so consumes fewer resources on the servers which makes the virtual machines more efficient. The result is sustainable availability and performance levels.
Offloading the most compute intensive processes to a purpose-built application delivery controller (ADC) is a virtually non-disruptive process requiring no modification to applications or virtual machines. The ADC virtualizes the applications and presents itself to the client as a broker; mediating between client and server. As requests and responses are received the ADC performs the most intense processing– TCP termination, SSL processing, compression, caching – leaving VM resources focused on the tasks it performs best. In addition, since the ADC is managing all of the application traffic virtualized application can be spread across multiple physical servers providing additional benefits in scalability, performance and availability.
Because the processing performed by the ADC is hardware accelerated and the ADC can determine what content will benefit -- and what will not -- from such assistance, application performance is not only restored but generally improved.
Offloading intense processes also removes the network as a bottleneck by decreasing the amount of data transferred. The encryption associated with SSL necessarily increases the size of data. Moving this processing off the VM reduces the amount of data needing to be handled which improves network performance.
So before you throw your hands up in frustration because the consolidation you hoped would come from virtualization didn’t, think about using an ADC to recover resources.
Posted by Karl Triebes on 05/26/2010 at 12:47 PM2 comments