The outage was caused by unexpected difficulties in a planned migration of Alabanza customers acquired by NaviSite to new systems. A planned 12 hours of downtime that began Saturday, November 3 stretched into several days, then almost a week. By Thursday afternoon, NaviSite issued a statement saying it had 90 percent of the equipment back online and expected everything to be up on Friday.

On Friday afternoon, the company posted a notice "from the desk of CEO Arthur Becker" on its Website, thanking customers for their forbearance, patience and understanding.

"I understand that this has affected your business and your customers," says Becker, "and I apologize for the inconvenience and disruption. This is not the standard for NaviSite. While we had planned this migration for months, we did not anticipate a number of scenarios that became major issues during the execution of the migration. Despite the fact that we have acquired a number of companies during the past few years and both consolidated data centers and migrated data, we had never encountered the series of problems that we saw earlier this week."

As of Monday, a few complaints were still trickling in from customers reporting that their servers were offline.

According to Rathin Sinha, senior vice president and chief marketing officer at NaviSite, the migration was officially completed early Friday evening, and any downtime a customer may have encountered since then would instead be considered a matter of normal maintenance and dealt with in the normal course of support.

"There are some customers whose service might have gone down or up," says Sinha, "but those are normal production or maintenance related issues, like rebooting the servers. So what we have done at the conclusion of the migration is we are treating these issues as our normal support issues."

Although NaviSite has undertaken other such integration and migration activities in the past, says Sinha, three separate issues arose during the Alabanza migration that combined to slow the process of bringing customers back online so significantly.

"There were really three unanticipated issues" he says. "One was - as we planned to do it by bringing the data over a dedicated line - we had data transfer issues. It was much slower than anticipated. Second was a major issue around the nameserver, where we could bring up the IP address, but the URL was not having a one-to-one relationship with the IP address. And the third problem was the ARP request, which came when many IP addresses were back live. A lot of people got into the network and there was slowness."

Troubling for NaviSite is the fact that the downtime that came with the migration has overshadowed the fact that the company set out to move the Alabanza customers on to a better platform.

"We made investments on creating a better infrastructure," says Sinha. "The goal was to move all of the Alabanza customers to a state of the art infrastructure with Sun machines and virtualization from VMware. This was really a migration for something better. Obviously during the process, the outage became much longer. And obviously many customers were distressed. It was a major disruption for them and it created a major inconvenience of them. And we got many, many calls."

According to Sinha, NaviSite feels it did the best it could with a bad situation, with all the company's technical resources working round the clock to get equipment online and creating extra support lines to handle the greatly increased volume of customer calls. He says the company issued regular updates via its Web site, though the messages have since been replaced by the CEO's, and NaviSite says it doesn't have a file of the messages it can provide at this time.

There are certainly some customers who feel the company didn't handle the situation well. Message board, blog and comment posts from as late as Monday, November 12 suggest that at least some NaviSite customers were still waiting for their sites to come back, and for an explanation from the company.

Sinha, however, paints a rosier picture of customers response to the outage.

"In fact," he says, "a lot of customers - and I have quite a few examples here - when their services came back, they thanked us for working around the clock and thanked the support personnel and the people who took their phone calls. They complimented them in every possible way."

NaviSite seems content to consider the online outcry the work of a vocal minority of customers, though the sheer volume of activity on message boards (185 posts on a Webhostingtalk thread at last count) would, at least by that measure, make the incident one of the more talked-about outages in recent Web hosting history.

However it chooses to characterize the criticisms, NaviSite's next step is the same. Sinha says the company is looking at its terms for issuing amends to customers whose guaranteed services levels weren't met. He says the mode of calculating compensation is in almost every case a condition of the service contract.

"Now we're definitely looking at what are the appropriate things to do," he says, "in terms of measuring the inconvenience and the disruption and quantifying that and seeing what customers really need. We definitely will do the appropriate things to serve these customers."

It may prove difficult at the moment to measure the precise impact of the outage on NaviSite's business, but there will most certainly be a monetary cost in the form of compensation and, it is more than likely, a loss of customers.