Taking charge of online application performance

by Martin LaMonica

April 23, 2001 —

The proliferation of service providers offering a wide array of hosted services has once again brought SLAs (service-level agreements) to the fore. The industry has long tried to measure performance of online applications in business terms, but the technology has generally lagged, particularly on the Internet. InfoWorld Executive Editor Martin LaMonica spoke to Trent Hein, the CTO of Xor, to discuss how his company is trying to make the Internet a more reliable platform for deploying business applications.

InfoWorld: How do you define the work that you do at Xor?

Hein: Our core business is around applications and systems management. As I'm sure you're aware, that part of the industry is really interesting right now, because the traditional client/server model, where the industry developed application management pretty well, is [already] understood. Things like service levels are commonplace. Some aspects [of Internet applications] behave like client/server applications, but in this distributed world other aspects of them are completely different.

InfoWorld: What has changed since the client/server days?

Hein: We no longer have the problems that we had, say, even five years ago where you were worried about hardware redundancy. We have RAID arrays. We have [similar backups] with routers and switches. We have all the operating system and infrastructure layers that allow an individual site to provide a level of redundancy where you can guarantee service levels. Today, we have a new problem: It's so distributed, you've got maybe six, seven, eight network providers on a common path between the central host of an application and the end-user. But you need to manage services levels across that [path]. Managing applications of performance level across that [path] is a whole different ball game.

Trent Hein -- Xor

Age: 32

Title: CTO

Biggest successes: Pivotal role in launching Xor's service with instrumentation technology for outsourced services

Key challenges: Demonstrating and evangelizing that the Internet brings revolutionary technology to the traditional business world by sharpening competition

Personal note: Enjoys hiking, skiing, and camping in the Colorado Rockies with his wife and energetic husky/Lab-mix puppy

InfoWorld: What is one of the big sticking points with SLAs via the Internet? What can service providers do to address that?

Hein: Today it's kind of a black box. Most of the companies you go to, they'll say, 'Well, we'll be happy to provide you a service, but we're going to monitor your box from this box that sits across the room and we'll make sure that the [connection] works well all the time.' Unfortunately, that doesn't give you very good insight into a lot of things. ... Where we specialize is in instrumenting applications, instrumenting networks, so that we can see all the different layers and then have ways to monitor and manage that to meet the user's needs.

InfoWorld: What's the end goal of adding this instrumentation into applications?

Hein: At the end of the day, what we would like to be able to do with the user end is [find out] what the business metric is that they're [trying to manage to]. They want to do so many trades of a commodity. They want to sell a certain number of widgets. They want to service a certain number of employees on their online HR system. What's the metric they want to manage to and the associated user response to that? So, what's the user experience? How happy are they with that experience? And then [we can manage] everything behind that. We can manage the application in the system and the network infrastructure to make that happen.

InfoWorld: What's the advantage of going to a hosting company for this function?

Hein: Monitoring it isn't enough. You have to react when there is a problem. Just putting it in a reporter, a blinking red light somewhere, that doesn't make the user happy. So, we have teams of people that are there to do applications maintenance. If there's something wrong ... a table overflowed or something's changed on the network or in the application or in the data-set so that we need to go in and do some coding, we [are available to] do that 24 hours a day.

InfoWorld: The Internet has been around for several years. Why should people consider dropping their ATM or frame-relay dedicated networks?

Hein: I think a lot of people are looking to improve their service levels and save cost. Two reasons that folks have avoided [using the Internet over dedicated networks] is that, in the '90s, [the Internet] wasn't so super reliable, and it definitely wasn't sufficiently secure. Now we have commercial-grade VPN technology. We have commercialgrade geographic distribution technologies so that even on a bad day we can get business transactions across the Internet.

InfoWorld: What do you think needs to be developed?

Hein: One thing that the industry really needs -- there are little sparks of this in corners but it just isn't widespread -- is integrated network management at the backbone level. When we talk about the Internet as a communications media, you have all these islands that manage their portion of the Internet. And sometimes the islands communicate, but they sort of communicate like [how] you would think of medieval times or something, where they're sending up smoke signals or they're waving flags to communicate the status within their network.

Integrated network management, where they provide real-time information from backbone to backbone and to customers and to other service providers like ourselves, and [providers] do things like transfer tickets electronically between all these entities, I think that's where the industry is going in the next three to four years.

And that's going to be huge. It will be very similar to the way the public switch telephone network works today. Most of those entities transfer tickets back and forth electronically. As the customer, when you make a long-distance call on your telephone from Boston, you have no idea which providers that call crosses or if it had to route around problems or anything like that, because they're well integrated.

On the Internet today, that's not the case. Oftentimes the user will actually know [that] the reason he can't run this application today is because there's a problem in Sprint's backbone or UUNET's backbone. And that's just not acceptable long-term. We have to get to the point where that's completely transparent to the user.

InfoWorld