While data center managers are seemingly enamored with the cost benefits and adaptive advantages of virtualization, not everyone is so comfortable with the technology. Take customers of managed services provider the Revere Group, for example.
Building the case for virtualization
Like every managed services provider, the Revere Group formally lays out its deliverables with customers via service-level agreements (SLAs) that outline uptime, CPU utilization rates, memory and the like. With an environment that includes "tons of VMware," in Margolis' estimation, the role of SLAs for the company takes on an added dimension; SLAs are used in the traditional way to verify pre-established service levels. But the company also uses SLAs to lend creditability to virtualization. "We use our SLAs to prove the case [for virtualization] to some of our customers who are skeptical," Margolis said.
For the Revere Group, virtualization is a necessity; the company could not cost-effectively provide hosting services to every customer asking for a dedicated physical server. Currently the company has two data centers in San Jose and Chicago, with a third being built in Sterling, Va. Both data centers run VMware ESX Server on Hewlett-Packard Co. ProLiant DL380s.Since August 2006, Margolis has used Nimbus, an application performance platform from Nimsoft Inc. With NimBUS, Margolis is able to augment the performance metrics provided by VMware's VirtualCenter infrastructure management tool with other metrics to monitor and report on the objectives in SLAs.
VirtualCenter, for example, monitors CPU utilization; if a customer is promised 1 GHz of CPU every month, Margolis can take the utilization information from VirtualCenter, feed it into SLA Manager and create reports showing how the Revere Group is delivering on that promise. "We can show customers that we're hitting our metrics 99.9% of the time every month," Margolis said.The Revere Group uses Nimbus throughout its infrastructure, including routers, switches and firewalls. "For the price and functionality, we couldn't do the level of monitoring without Nimbus, Margolis said. To set up a server with Nimbus, Margolis installed what Nimsoft calls "probes" -- that is, agents on servers that monitor typical characteristics, including CPU utilization, disk space, memory. The data is collected and sent into a SQL Server database. The SLA Manager application within Nimbus takes the server performance data and uses it to calculate and measure over time the various service-level objectives stipulated in SLAs.
Adjusting service levels on the fly
The reporting is one of the most useful aspects of Nimbus, Margolis said. He can create reports for customers based on data within certain parameters; for example, during maintenance windows, data can be excluded from an SLA report so as not to skew the metrics.
Let us know what you think about the story; email Megan Santosus, Features Writer.