When response times degrade for hosted applications, or Internet access seems to slow to a crawl, network managers have had to endure the finger pointing that ensues with service providers. But new technology from a small startup could give them the proof they need to take to the service provider to get the problems resolved.
Network engineers at the not-for-profit Providence Health System found with the unique appareNet network intelligence system the source of the performance problems they were having with a hosted human resources application. The tool, released for general availability late last month by Jaalam Technologies, gave them the documentation they needed to prove the fault was with the service provider, according to John Proffitt, lead network engineer at Providence Health System Alaska in Anchorage.
“It was very slow. It could take minutes to go from screen to screen, and sometimes it just crashed. We went to them, and they claimed it was our Internet connection. They said it was because we were in Alaska and they were in the lower 48. They said you need to get a better ISP,” recalled Proffitt.
Before bringing in appareNet, Providence Health System Alaska used commonly available tools to measure page loading times, but they were not able to pinpoint the source of the slow response time or determine the cause of it. But network engineers at the health care provider knew intuitively that the problem was with the ASP.
The appareNet performance measurement and analysis tool can determine whether performance problems are with the application or with the network by identifying the unique signatures of certain problems that occur in IP-based networks. Jaalam technologists created algorithms that can identify those signatures based on test packets that are sent out from a server and make a round trip from the specified end point. As the test packets traverse all of the links from one hop to the next, they are deformed by events occurring on those links that are associated with problems.
The software, which runs on an application server, includes an analysis engine and database of problem signatures that determines the cause and source of any problems it identifies along the way.
When Providence network engineers ran the apparaNet test packet sequencer from the server at the ASP, it was finally able to determine that the performance problems were both in the service providers network and on its servers.
“We found a half/full-duplex conflict. They had a misconfiguration on their network, and we were also able to determine that their code wasnt up to it,” said Proffitt. “It was the first time we were able to say, Heres why we know its not the network.”
The appareNet network intelligence system also changed the way Providence approaches new service providers. “Anytime we look to get a service over the Internet, it is a prerequisite to use apparaNet to snoop around and see if their Internet connection is OK. We now in a sense pre-certify ASP vendors before we go ahead with them,” said Proffitt.
Providence also discovered with the tool that not all Ethernet NICs are created the same. “Even a $50 NIC can make a big difference. Its amazing what you can discover with this little tool. You begin to make more informed choices in terms of what youre going to purchase,” he said.
Although Providence hasnt determined what kind of return it has gotten from the software, which starts at $30,000 for a single server or $100,000 for an unlimited enterprise license, Proffitt is convinced that it quickly paid for itself.
Another early user, DHL Systems Inc. in just one instance discovered right off the bat that a new $30,000-per-month asynchronous transfer mode Permanent Virtual Circuit (PVC) was only performing at half the contracted variable bit rate. The problem was due to a misconfiguration in the ATM carriers switch, according to Richard Muller, senior manager in Global Network Engineering Group for the systems unit of DHL Worldwide Express in San Francisco.
DHL also uses the tool to establish a baseline benchmarking the performance of service providers before it contracts for their services—including enterprise global carriers, prospective ISPs or ASPs and VPN services. “This is especially valuable in a B2B environment where multiple providers are typically used as the transport for the same information across multiple enterprises,” he said.
The age old finger pointing between the network operations staff and the applications development groups can also be put to rest using the tool, DHL found.
“It also includes the ability for us to prove or disprove to other IT departments–i.e., application developers–when performance issues are network-related and when they clearly are not, using the empirical data gathered from appareNet testing,” said Muller.