End to end network reliability is limited by a number of factors, including physical redundancy, router reliability, DoS attacks, etc. A recent study you might want to take a look at is End-to-End WAN Service Availability, by Chandra et al, which tried to quantify the impact of routing failures, and how much CDNs help at hiding failures (despite the hype, answer is: not much).
DoS attacks and success floods are also an increasingly big factor limiting end to end reliability. Anyone with a grudge can download a tool to take out virtually any Internet connection (we have even seen attacks aimed at everyone from major web sites to individual home users!), and the tools are becoming even more sophisticated and devastating. Re-routing along a redundant link doesn't help here (except in a few special cases) -- the DoS attack traffic would re-route along with the good traffic. Fortunately, there are products available (from us and others) to solve the DoS problem, and solutions to other Internet reliability problems are being developed.
Bottom line: demand will drive Internet reliability to get better; it is not a technological problem.
Slashdot Mirror
End to end network reliability is limited by a number of factors, including physical redundancy, router reliability, DoS attacks, etc. A recent study you might want to take a look at is End-to-End WAN Service Availability, by Chandra et al, which tried to quantify the impact of routing failures, and how much CDNs help at hiding failures (despite the hype, answer is: not much).
DoS attacks and success floods are also an increasingly big factor limiting end to end reliability. Anyone with a grudge can download a tool to take out virtually any Internet connection (we have even seen attacks aimed at everyone from major web sites to individual home users!), and the tools are becoming even more sophisticated and devastating. Re-routing along a redundant link doesn't help here (except in a few special cases) -- the DoS attack traffic would re-route along with the good traffic. Fortunately, there are products available (from us and others) to solve the DoS problem, and solutions to other Internet reliability problems are being developed.
Bottom line: demand will drive Internet reliability to get better; it is not a technological problem.