I-managers responsible for keeping their companies web sites up and running have rewritten the rules of engagement with their outsourcers, taking the task of monitoring their sites into their own hands and forging new alliances between the makers of hardware and the providers of analytic services.
It used to be that Web hosters, just like ISPs before them, would offer uptime guarantees for Web sites. Porous at best, such contracts were supposed to protect enterprises against hardware outages similar to those suffered by eBay, when a server colocated in an Exodus Communications data center failed and the online auction house went dark. But few service-level agreements (SLAs) offered guarantees that Web sites hosted outside the hosting company remain operational on the application level.
Now, customers want advanced technical details about the interaction between the data center ecosystem and their own equipment and services. In a recent survey of its customers, traffic management equipment vendor F5 Networks found that 80 percent plan to make Secure Sockets Layer-encrypted traffic part oftheir applications. And these customers wanted to know what kind of SSL throughput they would be able to get with different elements of the data center network — such as firewalls — so that they could plan their installations accordingly and demand certain levels of performance from their data center operators.
"SLAs are becoming much more granular," says Erik Giesa, F5s director of product management. "People who are now deploying mission-critical applications over the Internet have a much stricter criteria."
Failure to negotiate strict SLAs can lead to outages, and small businesses that dont have the budget and scale to buy expensive monitoring packages for simple services are often victims. Right Brain Marketing, a small San Francisco startup, signed up with hosting and e-mail outsourcer Bigstep. While receiving domain information from Register.com, Bigstep made a mistake, Right Brain execs assume, that prevented the small business from receiving e-mail. This had been going on for several months at press time, with the customer not knowing whom to complain to at Bigstep.
"Im going to go down to their offices — I know where they are — and knock on their door and say: Hey! You dont want this," says Anita Malnig, Right Brains vice president of content development.
A Bigstep spokeswoman says that Right Brain didnt fill out all of the forms required to fire up the e-mail and first complained about the lack of service on July 8. And even if Bigstep were in the wrong, the SLA that Right Brain agreed to covers only its Web site uptime. E-mail, it turns out, is outsourced to Critical Path, and there are "glitches" in that relationship; Critical Paths SLA covers Bigstep, but not Bigsteps customers.
Ask First, Act Later
So what should I-managers do to avoid having to knock on their outsourcing providers office doors with some tough questions about outages that have already occurred? Learn about such problems before their outsourcing partner does.
"We didnt have any problems, but we just wanted to see what was our time distribution on Web site performance," says Jack Gruninger, chief technology officer of Brainbench, an administrator of technology skills tests. Brainbench is hosted by Exodus.
Gruninger says that he has no beef with Exodus, but to learn what bottlenecks could ground Brainbenchs site, he opted for a specialized software package from AperServ Technologies, a maker of software that monitors SLA performance.
Gruninger evaluated AperServ against a new hosting SLA monitoring service from Keynote Systems that was formally launched last month. Keynotes service tests the availability of hosted Web sites through its network of testing servers. AperServ polls hardware that supports the site on a more granular level inside the data center, looking at specific elements of the data center net- work. AperServ is vendor-agnostic. Gruninger picked AperServ because it allows Brainbench to control what the software monitors.
"Our technology is capable of measuring the performance of systems regardless of the load balancer [or other gear]," says Tim Keough, AperServs co-founder and CTO.
Carrot and Stick
Vendors that make monitoring software say that having a means of keeping tabs on SLA performance helps forge tighter outsourcing relationships. "Business-unit owners will use this information either as a carrot or as a stick," says Karen Styres, vice president of marketing and business development of Resonate.
Also available to customers such as Brainbench are the services of companies such as Coradiant, a managed service provider that specializes in monitoring complex Web hosting infrastructure.
Makers of data center gear are also in the monitoring game, providing information about the performance of various apps on their hardware platforms to the makers of monitoring software packages. For example, F5 recently entered a deal with AperServ competitor Mercury Interactive to share such information. So by picking the right hardware vendor, enterprise customers should get the right monitoring apps. "One of the most important things in an SLA is looking at the data center ecosystem as a whole," F5s Giesa says.