Moving enterprise applications to the cloud requires close monitoring to ensure performance doesn't suffer. Several vendors have developed cloud monitoring tools that set benchmarks and pinpoint potential bottlenecks. Whether these tools give enterprises a full picture of what's going on in the cloud is still up for debate.
"Cloud application monitoring tools can be helpful, but they often are not as full-featured as longer established performance monitoring solutions," said Lori MacVittie, principal analyst at Rishidot Research LLC.
Still, performance monitoring tools have evolved. Traditionally, they were designed to examine various autonomous components, such as servers, network connections and storage systems. The rise of virtualization has blurred the dividing lines among these devices.
- BMC's End User Experience Management tool tracks user activity at the individual or group level and distinguishes between broad and targeted slowdowns.
- Cloudkick Inc., which was acquired by Rackspace Hosting in 2010, gives customers elasticity; as new servers come online, new monitoring functions become available. Cloud admins can customize alerts and send them via email or SMS.
- LoadStorm has a real-time cloud monitoring tool that provides graphs with performance metrics updated in one-minute intervals. Among the data highlighted are average response time, error rate, requests per second, throughput in KB/second, peak response time and number of concurrent users.
- SevOne's cloud performance monitoring tool identifies changes in performance behavior through automated base-lining across various data center elements. The company's traffic analysis functions also support differentiated service models.
- SOASTA's CloudTest software creates complex tests without the need for programming. Tests can be paused or stopped when an issue is found and restarted once it has been resolved.
But each of these cloud tools comes with a few caveats. One challenge is presenting enterprises with a unified view of their network, server and application performance across physical and virtual infrastructures. Performance data has to be funneled to a central location while still allowing users to drill down to troubleshoot individual elements.
Timeliness is another issue and performance problems don't happen in regular intervals.
"Performance problems can arise in an instant," MacVittie noted. Most of the tools poll locations are various intervals, so a bottleneck may occur a moment after a poll, but remain undetected for 15 minutes.
Additionally, product pricing can be difficult to decipher. Vendors develop various pricing models based on items like the number of servers monitored, the number of users or the volume of data. Consequently, making apples-to-apples comparisons of cloud performance monitoring tools can be difficult.
Paul Korzeniowski is a freelance writer who specializes in cloud computing issues. He is based in Sudbury, Mass., and can be reached at firstname.lastname@example.org.