Attempts to visit Blockbuster.com yielded either HTTP 503 (service unavailable) messages or an apology: "We're sorry, Blockbuster.com is temporarily unavailable while we make some changes to the site. We're working hard to bring you an even better Blockbuster Online experience. We apologize for any inconvenience this may cause. Check back soon!" Maybe the IT crew at Blockbuster can take comfort in knowing it wasn't alone in having a crummy day December 20. Their colleagues over at Foot Locker found themselves with a site that crashed at least three times that Saturday morning, according to Pingdom. "We got a few different error messages from them," Pingdom Analyst Peter Alguacil said. "A few HTTP error 504 (gateway timeout), a few HTTP error 503 (service unavailable). But most of the time the Web site was just extremely slow in responding." December 20 also brought a trio of brief outages at the Foot Locker site. Pingdom said Footlocker.com went down three times shortly after midnight that Saturday. "As far as I can see, the issue with Foot Locker doesn't look to be traffic related, but rather some kind of backend problem," Alguacil said. "It might well have been caused by some work being done, but it's hard to tell. They might have been doing something that affected the way their site works with Akamai, since the 503 and 504 errors we got were from Akamai servers." He acknowledged that trying to determine the causes of site problems is often "extremely difficult." Alguacil always takes into account the times when incidents take place. Weekend Work
"Considering the timeframe of the Foot Locker problem, it is quite likely that they were doing some form of work on the site and ran into trouble," he said, noting that the overnight period between Saturday and Sunday "is a pretty common choice for that, since less commerce than usual is expected to be going on, in case something goes wrong." Target.com became unavailable for about an hour on December 31, Sitemorse said. That was the same day the Lowe's site had some significant issues. Sitemorse reported Lowes.com to be unavailable at 3:08 PM New York time and said the site didn't recover until 4:36 PM. The Gap's site slipped up on the day after Christmas, Pingdom said. It remained offline (to at least some shoppers) for about a half-hour, about the same duration as an outage, spotted by Sitemorse, that hit Walmart.com on December 30. Other outages of note: CDW's site went out for about 48 minutes on December 19. Kroger's collapsed temporarily on Christmas Eve and again on January 4. Given the results of a nationwide survey from E-commerce development company Guidance, none of this will come as a surprise to many online shoppers. Guidance found that 36 percent of online shoppers "ran into roadblocks" while online during the holidays, "ranging from molasses-like Web site response to fruitless efforts to check out to outright system crashes." The good news is that about 64 percent of those surveyed said they encountered no problems doing their online holiday shopping. But Guidance found that 37 percent of the respondents said they "skipped Internet shopping altogether" for the 2008 holidays and that "a small percentage" of the people in that category said they didn't bother shopping online due to frustration from problems they encountered during prior attempts. According to Guidance, 13 percent of the people who had online shopping problems this year said they gave up because the site they were shopping was too slow. Eight percent said a site froze or crashed, 7 percent couldn't complete a purchase on their first attempt, 6 percent encountered a site that was temporarily down and 4 percent said "a purchase they thought they had completed actually didn't go through." No Excuse For Hiccups? Guidance owner and CTO John Provisor said there's really no excuse for big-time retailer Web site outages, even during peak holiday periods. "It seemed that, as a whole, the performance of many sites was poor given that retailers should have had plenty of time to estimate and prepare for the traffic," he said. "Maybe they were surprised by the volume of the traffic, but it's really a message to the retailers to do a lot better preparation. They should prepare months in advance with advanced stress testing that simulates that kind of traffic." Matt Poepsel, vice president of performance strategies at site performance company Gomez, disagreed with Provisor's unflattering assessment of site performance during the holidays. "The nature of how the retailers responded this holiday season was pretty good overall," he asserted. "Certainly, the month of December seemed to be business as usual. That hasn't always been the case." Retailers, having gone through a "gradual learning curve" over the years, are generally adept at maintaining site uptime, Poepsel said. "We've seen availability numbers hold up very well," he said. Although Poepsel acknowledged there were some high-profile site flameouts, he said retailers in general made big investments in infrastructure and that they are benefitting from more durable hardware and applications. Mid-December retail site problems can often be attributed to a "chronic buildup" of issues combined with "the stress of trying to maintain high levels of volume over a long period of time," Poepsel said. Provisor agreed, noting that retailer databases, after dealing with a blitz of orders after Thanksgiving, tend to get clogged with data, logs, backups and algorithms. "It really magnifies the demand on the infrastructure," Provisor said. "The solution is proper cleansing and pruning of your legacy and archive systems. I have seen systems that have data logs that take up all the capacity of a system. If they don't have proper pruning and purging, they're going to see a slowdown."