I got an email from an angry Adobe Connect Professional customer asking if I knew anything about the "Adobe Connect crash" on Thursday. I hadn't heard about it, so I contacted Adobe and talked to David Slater in product marketing and Vincent Toesca in product management for Connect Professional.
They confirmed that there was a drop in the hosted service yesterday some time around 9 or 9:30am Pacific time. They say it lasted for roughly 30 minutes or so. The customer I spoke to estimated 45 minutes of service interruption. Unfortunately that came right while they were trying to host a webcast with hundreds of attendees. Ouch.
David and Vincent were unable to tell me the cause of the outage. They said that their IT team was still looking at logs and diagnosing the situation. I asked whether Adobe uses outsourced data centers and they told me that they hosted and controlled all servers themselves for the Connect hosted service, so it's not a problem with an external data center going offline.
Although Adobe has failover switching to backup servers, the large instantaneous load caused some of those to fill up and require a secondary switchover, which would have extended the down time for certain customers. From what David and Vincent tell me, the severity of the interruption was different for different customers. Some had a momentary loss of connection and then came right back. Others suffered some performance degradation, for instance losing streaming audio over VoiP, but kept their meetings going. And some lost the entire connection and use of the service.
I asked about the frequency of failures on the Connect servers. The men said that the only other outage this year was at the end of June.
I'm not on a witch hunt here. Large online services fail. They all do. I lost my cable service for half a day last month. I've lost connection to WebEx twice this year. AOL, Amazon, and Netflix all have had publicized service drops. It's a hassle and infuriating. The big question is whether the IT department can learn from the incident and build another layer of protection and better response. In this case, it sounds like failover load balancing might be improved upon. But we won't know the details for a while. I asked Adobe to follow up with me when they get more information about the cause and how they plan to respond.
Chalk up another debate point in the battle between installed software and Software as a Service!