Keeping An Eye On Virtualization

Figuring out the performance of a system can be relatively straightforward. Figuring out precisely why a system performs as it does (where, to be specific, the inevitable bottlenecks lie) can be quite difficult. Figuring out why a virtualized system performs as it does can be darned near impossible. Akorri's BalancePoint 3.0, released in August, is a tool for figuring out performance in a virtualized environment so admins can plan capacity, allocate resources, or just solve problems for frustrated users.

BalancePoint 3.0 has introduced richer virtual machine and storage analysis, broader interoperability and user interface improvements to the BalancePoint product line. The product allows IT organizations to optimize virtual and physical server and storage infrastructure in order to to reduce costs and assure service levels as companies move through the stages of virtualization adoption and into private cloud computing.

Akorri ships the BalancePoint 3.0 as a cloud product -- an agent-less platform. Accordng to Rich Corley, CTO of Akorri, one of the advantages to this approach is the rapid set-up time the cloud architecture allows. He says that they can typically configure the product within an hour or two of arriving at the customer site. Because it's agentless, it makes maintenance and service much easier.

Corley says that they first get the topology of the network at the beginning of the engagement and then present a unified view of the environment to the customer for configuration discussions. Often, the view helps customers solve problems since they might never have seen a graphical representation of the network. Corley says that the picture that's presented frequently allows customers to realize where relationships exist that they didn't understand based on other representations of the environment.

Corley says that creating the dedicated data collection pieces was the hardest part of building the system. "We had to build a storage collection architecture into the product. We go to the physical server and do the same thing, which is a little easier, but still involved. Building the whole agent-less data collection system and building the database model across the various vendors was the complicated part," he says.Finally, building the analytics, as opposed to simple summary performance data, helps staff understand where things are from a management and trouble-shooting point of view. Corley says, "We take the capacity utilization information and add a performance view to it. We pull the information from across the various tools and provide a single interface to them." There are, in fact, other tools that could allow an administrator to get much of the same information given sufficient time to configure and design the system. Splunk, for example, offers a great deal of data collection and display capability. Asked about whether customers might just as easily use one of the other tools, Corley says, "To answer the question, yes you could take the 3-plus years we've taken to do something like this on Splunk, but we're the first company to take the time to make all the systems talk to one another to help people figure out where the performance problems lie."

Current economic situation is forcing many companies to adopt virtualization in a much more pervasive way. Where they originally might have wanted to get 25% of servers into virtualized setting, they now are looking at 40% or more of virtualized servers. Corley says that  Akorri hears many customers say that they're looking at the tool set, but they don't see where their problems lie. Nine times out of ten they find that it's not really in the hypervisor, but rather in the way that SAN and storage elements have been configured, but they don't see it because of limitations in the tools.

While many companies are satisfied with the benefits of virtualization, Corley says that he's seeing many customers who are using virtualization as an intermediate step toward cloud computing. Understand performance fully lets them get ready for internal (or external) cloud deployment - and the additional level of performance abstraction that comes with components in the cloud.