It wasn’t meant to be this way. Platform health was supposed to transfer directly to user experience. But the prevalence of online applications, and the use of simple office applications in WAN environments, has required a move beyond a components-based approach.
User performance management isn’t a new concept, but it can still be hard for some administrators to be told that a green light/red light, fast/slow approach to networks, CPUs, packet loss, and disk space, may be unrelated to the end user’s experience.
“There is a huge rift between platform health and user experience,” says Alistair Croll, senior vice-president marketing for Coradiant software, a user performance management company with R&D headquartered in Montreal.
“When looking at complex, distributed applications it makes more sense to start with a user experience,” Croll says. “If they report a problem, you can assess what they had in common.”
Brad Wilson, chief executive officer and president of QoS Networking, a company out of Albany, N.Y., that provides, among other things, application performance tuning via Compuware’s Vantage product, agrees. “The place to look is the end-user,” he says. “Just looking at the circuits doesn’t tell me what the user experience is.”
Products like Vantage are designed for companies requiring a process that aligns to the service delivery process components of ITIL. The problem is that complexity has led to specialization, and most infrastructure tools aren’t picking up on the problems that end-users report. What’s needed is an approach that gets everyone on the same page.
Brian Handrigan, director of sales for Telnet Networks Inc. in Elizabethtown, Ont., says that newer technologies such as VoIP only reinforce the need for a three-pillar approach to the entire network. “From a hardware perspective we need device management, but also end-to-end performance, and then, more specifically, application performance management,” he says. That third component requires not only managed response times for each application on the network, but also an ability to assess the effects of new applications.
“We need to know the end-user response time for every application,” he says, “whether it is Web-based, custom, or an ERP module.”
One of the problems, says Croll, is that only a small percentage of end-users will report an issue, yet it is from end-users that IT managers find most problems, not from traditional monitoring tools.
For Web-based applications, Coradiant’s customers watch end-users via its TrueSight appliance. This can be deployed as a single box in an enterprise’s data centre, tracking behaviour from an array of applications including PeopleSoft, Siebel and SharePoint.
“We’re able to understand the entire HTTP conversation, thousands of pages a session, in real time,” says Croll. “We then know how much delay is due to network, how much to server.”
Martin Morel, VP production for Taleo, an on-demand talent management company and Coradiant customer, reinforces the need for a 360-degree view.
“We have total visibility from within our network infrastructure all the way over to the end user,” Morel says. For a company like Taleo, and other popular on-demand ISVs such as Salesforce.com, it is crucial that there be visibility beyond the perimeter. “The device sits on the outside edge of our network and can measure network latency, the time waiting on different servers, all the way the edge of the customers network,” Morel says.
This can lead to some powerful results, particularly in crucial areas such as mean time to recovery for online applications. For many enterprises, however, such front-end heavy, Web-native business and technology models are still down the road, and the real struggle is opening a Word document in a WAN environment.
“MS Office, e-mail, and other applications designed for a LAN environment, when put on WANs, can be slow,” says Jayanth Angl, research analyst with Info-Tech Research Group. “People can run into a lot of problems in terms of performance.”
“You need monitoring at the application layer,” he says, pointing out, as does Handrigan, that VoIP can certainly complicate matters. “When you have different real-time applications like voice and video on the network, it is not just a matter of bandwidth, but also latency, jitter, and packet loss.”
Handrigan points out that determining a slow response time is useless without the how and why. “We need to know the applications running at that time, see if the routers and switchers were at 100 per cent CPU utilization, then check for the time and any anomalies,” he says. “You don’t want to change the whole network for something that will only happen once.” There are appliances that address application performance issues over WANs, and they’ve been shown to offer considerable improvements.
“This can include optimization of the actual TCP protocol,” says Angl, “as well as certain compression features. Most enterprises want to keep their spreadsheets in-house, but certainly on-demand offerings can address this.” J.R. Simplot, a $3-billion a year private agribusiness company based in Boise, Idaho, with 230 Canadian employees in Portage la Prairie, Man., has announced an ambitious $3.45-million plan to roll out AT&T’s voice, IP-enabled frame relay and asynchronous transfer mode network services. However, at present VoIP is only used internally, and application availability is a real concern.
“We are doing VoIP across frame relay today, and are moving to an MPLS network. Data and application availability are concerns. This will be phased in over five years. We have a different group looking into Blackberry infrastructure,” says Simplot’s chief information office Roger Parks. Given the distributed nature of the enterprise, securing availability by taking niche applications off the network altogether and going for an on-demand solution makes sense, and this is what Simplot has done.
“We have an ASP for rail car management that integrates into JD Edwards at the back end, and that works fine. We have several hundred rail cars throughout the US and Canada.” The food services group also outsources for contract management. Anything that reduces loads helps, especially when your own people put on a low-scale DOS attack.
“We got hit hard by the NCAA tournament,” says J.R. Simplot’s network technology manager Mark Sachs. “Fantasy football, too. We need guidance from upper management for this; we don’t want to be the bad cop. But our proxies keep logs and we could see what happened.” In the end, application availability is also an HR and policy issue. “We have already adapted new policies for MP3 downloads,” says Sachs.