Sysload Delivers Fine-Grain Monitoring for Virtual Servers
June 17, 2008 Alex Woodie
Organizations concerned that the availability of critical applications is suffering as the result of hitting the physical performance limits of their virtualized servers may want to check out the SP-Analyst offering from the French company Sysload. With the capability to monitor a wide range of servers, including i OS machines, at one-second intervals, SP-Analyst may be a good solution for spotting sudden performance spikes in virtualized infrastructures, which other server monitoring offerings with longer sampling intervals may miss.
Before virtualization become popular, it was common for industry standard servers to run at 10 or 20 percent of capacity. While this wasn’t very efficient, it provided lots of headroom for potential spikes in demand. Now that virtualization has become part of the enterprise, things have turned around, and organizations now reap 80 to 90 percent utilization rates out of their physical hardware, which is great for stretching their server investments.
“Virtualization kind of gets us back to the mainframe model in the sense that the goal is to make it as efficient as possible,” says Ted Dinkel, Sysload’s executive vice president in charge of North American sales and marketing. “Max it out, keep running right at the maximum line, with no head room.”
However, the efficiency of virtualization comes at a cost. With very little extra server capacity available, organizations may suffer from sudden spikes in demand that exceed the virtualized server’s capacity, potentially slowing down critical applications. When those applications are time sensitive, such as stock trading applications, any delay costs the organization money.
When virtualization started catching on, Sysload realized that its existing SP-Analyst product would be a good fit. With the capability to measure more than 250 different server metrics at intervals as frequent as one second, Sysload’s SP-Analyst software can help identify performance trends before they turn into problems.
SP-Analyst’s capability to monitor performance metrics such as CPU and I/O utilization every second can illuminate potential problems much sooner than competing system monitoring tools, such as IBM Tivoli and HP OpenView, which take samples every minute or more, Dinkel says.
“We start to see micro spikes and problems emerge at the two- or three-second level that you might not see for weeks or months just taking a one minute samplings,” he says. “From a business standpoint, customers are saying ‘I need to make sure this application is up 99.999 percent of the time, and to do that, I need to see just the slightest movement into some problem areas,’ and you can do that with Sysload.”
Dinkel says the agent-based architecture of SP-Analyst is the main difference from other system monitoring suites. Most other vendors have adopted an agent-less approach–where data is pulled into a central database–to minimize the impact on the servers, gear, and other devices they are monitoring. The downside of this approach is that it floods the network with information, and it costs more to run the database.
Sysload has managed to maintain the agent-based approach by developing a very lightweight agent that runs close to the kernel and minimizes the information sent to the SP-Analyst console, Dinkel says. “Our breakthrough was the agent written in binary code, written in C,” he says. “It uses no scripting, no SNMP, all the things that slow down older technology in any kind of a central database.”
SP-Analyst can be used to generate alerts if performance drops to a certain level. “You can say, ‘Send me an alert when my CPU is over 80 percent for a duration of 5 seconds and I have more than two threads in a wait state. And by the way, when sending alerts, include the actual trigger values.’ It’s a pretty powerful thing that most of the technical guys like.”
SP-Analyst runs on Windows PCs and can scale up to thousands of connected servers. The software can also be used to query the agents, where historical data is stored, to determine past performance usage and make capacity planning forecasts.
The software supports a wide range of operating systems, including i5/OS V5R3 and V5R4, Windows, Linux, AIX, HP-UX, Irix, Reliant Unix, Solaris, SCO OpenServer, Tru64, and UnixWare. A recent update brought support for VMware.
On i5/OS, the software tracks more than 200 different server metrics, including CPU utilization per job or per subsystem, memory, disk arm activity, job tracking, communication lines, applications, users, and other areas.
Sysload also sells SP-Monitor, which performs event correlation and incident management and features a Web interface, and SP-Portal, which provides a high-level view of server and application status.
Sysload recently opened an office in Boston, Massachusetts, and is ramping up sales activities in North America. The company, which was founded in 1999, is headquartered in Paris, and has about 70,000 servers under management.
Pricing for the SP-Analyst console starts at $4,000. Agents start at about $1,400 for i5/OS, which enables users to manage one instance of i5/OS running in one LPAR. For more information, visit www.sysload.com.