This opportunity to examine the practical, day-to-day end user computing experience from the outside in, rather than from the data center system outward, means problems that end users have come to tolerate, because they can't do anything about them, may be possible to identify and diminish or banish altogether.
There are several end user-oriented performance monitors in play, such as HP's Mercury Interactive products, Compuware, and Empirix. One cited in the 2010 Gartner report, "APM Innovators: Emerging Vendors Drive Market Evolution," was Aternity, which released version 5.0 of its Frontline Performance Intelligence Platform on Tuesday.
FPIP is a performance analytics engine, accumulating a data warehouse full of performance statistics for each end user, Aternity President Trevor Matz said in an interview. The end user may be on a hardware device or in a virtual machine; it makes no difference to the platform.
Aternity actually OEM's a version of Oracle with its product to accumulate, index, and partition all the stats. "It scales to terabytes of data," said Matz, which is a good thing, since FPIP is constantly collecting data on each end user's activity. It knows, for example, how long a transaction typically takes for each end user and can detect when a 1.5 second transaction has suddenly started to take two or three seconds. It recognizes individual CPU and memory usage and application response times.
The 5.0 version also has ramped up its ability to collect these stats off a host server, such as a Citrix XenApp Server. In the previous version, Aternity could monitor 70 to 100 users per XenApp Server; in 5.0 it can monitor up to 250. It also monitors VMware View hosts or any Windows host handling multiple end users, such as one running Terminal Services. It can track the activity of 30,000 to 40,000 users in implementations currently in place or about to be completed, he said.
FPIP does so, explained Matz, by placing a small monitoring agent on each client, which supplies feedback on application performance, user activity, and device performance for each end user. The information goes to a central server running FPIP, which is conducting real-time analysis as needed on the data streaming into the underlying data warehouse. A model is constructed of averages for each end user, and the FPIP analytics engine can detect when day-to-day variations start to occur in those norms.
If, for example, Microsoft Outlook is responding slowly to a particular end user, "very often the problem is not Outlook. It's something else going on the user's computer," such as a security scan or activity by plug-ins particular to that user. FPIP can correlate the effect of such activity on how well Outlook will perform under the contention and indicate to an IT manager where the problem lies.
Line of business managers as well as IT managers can consult an end user key performance indicator dashboard, which has set timeframes for user experiences for each individual in the organization and shows whether those expected timeframes are being met.
Managers may also use line of business definitions to set different metrics for different groups within the enterprise. Managers may need to know that the expected timeframes for a transaction in the organization's banking center can take longer than a similar transaction on the trading floor, where latencies in transactions are a more sensitive issue. FPIP can report the difference without setting off alerts. In short, Matz claimed, the end user monitoring can look at training groups, where user activities may be hesitant and tentative with one application, differently from call center groups, where user activity with the same application is expected to be crisp and responsive to customers.
Aternity's platform is more than just passive monitoring, however. Its analytics engine is able to drill down on a problem and show a flowchart of user activity, with the IT manager able to move a vertical bar over any portion of the flow to get more detail. "That bar is a time machine," said Matz. It can summon the user data in the database to illustrate what happened to response times as a problem occurred.
It also can drill down into something like the omnipresent Outlook server's performance to see what's going on beyond the times it takes to open a message or send a message. It can correlate Outlook's responses to its connection to a backend Microsoft Exchange Server to measure impact on performance.
In the new, all-digital InformationWeek Best Of Interop supplement: See why VMware's vCenter Operations and eight other products stood out at Interop 2011. Also in our supplement: Dell, HP, IBM, and 13 other vendors team up to showcase gear based on OpenFlow, the network virtualization standard. Download it now. (Free registration required.)