Hacker News new | ask | show | jobs
by chuckgreenman 3032 days ago
I'm not sure that's something that you can take away from the article, Cogent was performing well on other ISPs, Comcast was not providing enough bandwidth for Cogent due to a disput external to Panic.

While we can't know I'd be willing to be the "unspecified" traffic re-engineering was to prioritize Cogent's Panic traffic specifically. Comcast is still the bad guy.

2 comments

Working in the field of networking I can say it's highly unlikely that Cogent is performing traffic shaping specific to Panic traffic. That's a ton of changes to a ton of hardware with a risk for screwing things up that just isn't worth it. Cogent isn't going to go to those lengths for a client that has a single connection and hosts a small / moderate site. Beyond that: the connection that Panic has to Cogent is most likely through the data center they're colocating with, not a direct connection to Panic equipment. That's not for certain, it's entirely possible that Panic paid for a drop from Cogent; however I'm guessing that's not the case.

What I find odd is that they would colocate their hardware in a DC with only a single provider. The company I work for has four transit providers so as to ensure uptime / reliability (don't get me started on how often we experience unplanned maintenances from our upstream providers). Seems like Panic may want to consider a different host for their colocation or examine (provided they're dealing with Cogent directly) a secondary ISP.

>What I find odd is that they would colocate their hardware in a DC with only a single provider.

I think it's extremely unlikely the provider only has cogent. More likely is the fact that the link between the provider and comcast preferred cogent - which panic would have no way to change.

That's not true. panic.com has an A record (and www is a CNAME to @) that points to an IP address that's not PI and advertised only as part of 38.0.0.0/8, obviously, only from Cogent.
It is. In my experience, Cogent has more peering problems than other transit providers do (e.g. L3). The submitted article is a clear example of this. I don't find these peering problems surprising, given Cogent is much cheaper vs the competition.

>prioritize Cogent's Panic traffic specifically. Comcast is still the bad guy.

No, from the article it really sounds like Comcast added more capacity to their Cogent-Comcast peering links, at their expense. Comcast was not obligated to do this for Panic in any way. How does that make Comcast 'the bad guy'? Cogent should have worked with Comcast to fix this issue in the first place. Panic is paying Cogent for transit, Cogent should be the one getting Comcast to increase their peering bandwidth. If Panic has to go around the company they're are paying to get their problem fixed, that's terrible service.