UPDATE Almost a week after many of its users lost service, Microsoft's MSN Messenger continued to be plagued by problems on Monday.
Microsoft said that 10 percent of its MSN Messenger users remained without service Monday. The problems prevent users from accessing their buddy lists - online address books that enable users to open message boxes with other users.
"We're still working on restoring the service so everyone worldwide can log in," a spokesperson said on Monday. "We're working around the clock to restore service."
The break in service is a huge embarrassment for Microsoft, displaying a vulnerability the Redmond, Wash.-based software giant can ill afford if it is to go forward with its ambitious plans for Internet services. Those plans involve the building of multiple data centers to host a range of Web services, dubbed HailStorm, for millions of users. Microsoft envisions that the single gateway, the Passport identity-authentication service, for all of these Web services, including MSN Messenger, will run on servers that securely and reliably host the data of hundreds of millions of users.
The MSN service interruption began late on July 3. Microsoft blames a hardware failure for the problem. "An extremely rare set of circumstances occurred when one of our database servers had a disk controller fail," the software giant explained in a statement. "The backup for this controller also had an error occur."
Microsoft claims some 32 million users of the MSN Messenger service. According to a May survey by Gartner, 36 percent of consumers and 40 percent of business users use MSN Messenger, compared with 52 percent of consumers and 51 percent of businesses that use AOL's rival service. The new Windows XP operating system due in October will bundle the instant-messaging service into the OS, a move that threatens AOL's lead. (In part to protect its market lead, AOL has ensured that the two services remain incompatible.)
When HailStorm was launched back in March, group VP Bob Muglia, emphasized the need for Microsoft to demonstrate that it could deliver "operational excellence" in large-scale systems with high levels of availability. "This is not something Microsoft has always had in the core of our genes," Muglia conceded that day, though he added that Microsoft was gaining experience in running big data centers. "Being honest, some of that experience is not so good." In January, as a result of problems with servers hosting domain names, many Microsoft sites had been intermittently unavailable for several days.
For business-critical applications within enterprises, Windows has historically been no match for Unix systems. But Microsoft has striven mightily to change its crash-prone reputation in the server market. Ever since the release of Windows 2000, it has touted massively increased reliability. One of its current major ad campaigns has a theme of "Five Nines" reliability, or 99.999 percent uptime. "That translates into just over five minutes of server downtime per year," reads one ad, which states that Microsoft has set up such highly reliable systems for customers such as Starbucks and GMAC Commercial Mortgage.
But the Messenger outage is evidence that Microsoft cannot yet guarantee the level of reliability needed to run vital "always on" Web services of the kind that it plans with HailStorm and Passport. Critics, who see Passport as a potential central control point on the Internet, were quick to knock Microsoft for its current Messenger problems.
"Can you imagine when they absorb the Internet?" said Andre Durand, chief strategy officer with Jabber.com, a startup that offers a rival instant-messaging service based on open-source code. "Passport goes down. The world is not available."
Durand admitted that Microsoft has achieved scalability beyond anything his startup has achieved and is likely to achieve within the next year. Jabber just Monday announced its first enterprise customer: Disney is deploying the instant-messaging service along with its Go.mail Web e-mail service.
However, he said Jabber is aiming to provide enterprises with their own branded solutions, each running in distinct data centers, so
