Ranking by Fewest Failed Requests,
September 1st - September 30th
After August, which contained the Blaster Worm, the SoBig Virus, and the North East US/Canada power outage, September was a much more benign month for the internet, with the top two sites showing remarkable reliability.
Just one of the 2976 requests made from our five performance measurement points to the British Telecom site failed, and only two out of 2976 requests failed to the Dell Host site.
These are the most reliable statistics that have been posted to date on a calendar month basis, and also the first time that a site running Windows has been placed top.
Many events that could adversely affect performance and reliability hit the hosting industry during August including Distributed Denial of Service Attacks, the Power Outage in the North East US and Canada, a much shorter outage in London, the Blaster Worm, the SoBig Virus, and a deluge of related mail from poorly written anti-virus programs.
Ranking by Failed Requests and Connection time,
August 1st - August 31st
Remarkably the numbers of failed requests to hosting company sites did not rise significantly during the month; a testament to the durability of the Internet. However this hid some big variations:
Rackspace's site was down for around two and a half hours
early this morning [BST], in what may likely have been a repeat of Tuesday's distributed denial of service
The response times to our own performance collector on Rackspace's network indicate that the attack did not adversely affect response times to other machines at Rackspace.
The recent spate of distributed denial of service attacks has diversified, with some attackers apparently now targetting hosting companies.
On Tuesday rackspace.com was attacked just one day after issuing a press release
launching a service to mitigate the effects of DDoS attacks, while early this morning[BST] Rackshack appeared to suffer a similar attack
Nick Marsh points out that the www.sco.com
site has been down for a long time - now over two days so far according to our performance measurements
At the moment, it is not known whether the SCO site has been successfully attacked, intentionally taken down, has lost connectivity or has simply broken.
On Friday Microsoft changed its DNS so that requests for www.microsoft.com no longer resolve to machines on Microsoft’s own network, but instead are handled by
the Akamai caching system, which runs Linux.