From: Sean Donelan [mailto:sean@donelan.com] Sent: Tuesday, April 03, 2001 5:26 AM
On Mon, 02 April 2001, Roeland Meyer wrote:
From the article: <quote> "Half of the companies that are multihomed should have gotten better service from their providers," says Patrik Faltstrom, a Cisco engineer and co-chair of the IETF's Applications Area. "ISPs haven't done a good enough job explaining to their customers that they don't need to multihome." </quote>
Why would a rational customer pay for a second connection (usually more than doubling their cost) if a single connection was satisfactory? Although providers try to vertically integrate their operations, time and time again, vertical integration tends to increase the risk for the customer.
Because, if the ISP goes out of business the lead-times for getting a new connection exceed 40-days. This was my point on invoking the DSLnetworks/Covad spectre (it's the one I know the most details of). Also reference FlashPoint, NorthPoint, et al. We are now suffering over 100,000 businesses being disconnected from the Intenet because their upstream went tits-up. So, don't tell me it doesn't happen. This is the original reason for having multi-homing with two different upstreams. In most cases, the connection itself was rock-solid up until the time the lights went out. In many cases, there was no warning. In ALL cases, the outage exceeded 4-weeks unless the site was multi-homed somehow. My latest quote for a new DSL connection is 42 days. T1s are 6-8 weeks. Bigger pipes are 2-6 months for the fastest response, many cases are longer. Back to business failures, the DSL world is looking rocky at best. Rythms isn't looking very solid and neither is Covad. Some of the NorthPoint customers are not going with either, electing to go with the local RBOC instead (despite the hassles and long lead-times). What many businesses really need is true multi-homing.
Mid-level providers serve an important function in the Internet hierarchy. Multi-homing works well with mid-level providers aggregating local routes, and managing redudancy between long-haul providers. If you don't use a mid-level provider, to achieve the same reliability you end up needing to be your own mid-level provider.
Exactly. You also save some monthly recurring cost.
Why can't a large provider operate their network as a set of mid-level networks, and connections to multiple long-haul networks. They could.
But, they dont.