hm. See, I thought that provisioning/support was my weak point, not bandwidth. Bandwidth is actually an easier problem to solve, right now; I have, shall we say, massively overbought. I do need metering (right now, well, when I say I 'overbought' well, my burst capacity is two whole orders of magnitude larger than my 95th percentile. My commit is 60% of my burst capacity. I haven't had to limit a legitimate user for bandwidth usage in more than a year.) and I do need to do some general network upgrades. Right now, while I have the /external/ capacity, there are several internal bottlenecks which means that I will see some bad shit when usage hits one order of magnitude or so above what it is now. Actually, slightly less than that. I've got several points in the network where some traffic ends up doing a 'router on a stick' where the stick in question is a 1G link. Bad news, as my external capacity is 12G/sec. if Also, I don't have any metering, I mean, other than tcpdump (and perl) when shit gets bad. That's clearly a problem if I want to run at over 1/10th capacity, which I do.
I mean, on the metering side, right now, that is part of why the bandwidth quotas are so 2007; I mean, I /know/ that if I am having problems on my network due to overuse, someone is massively overrunning their bandwidth quota, so I can TC them down to something reasonable without breaking with the expectations I've set. So yeah, if you really do need serious amounts of bandwidth, (like, say, north of 200Mbps) on a VPS, I'm weak... but that's a whole hell of a lot of bandwidth.
(so yeah, I guess what I'm trying to say is that you are right if you are a serious bandwidth user, but most people aren't; and my policy of 'limit, and only if it's causing problems' I mean, rather than charging you for overages, makes bandwidth less of a big deal for most customers.)
Nonetheless, I am way too close to the 'will cause problems' bit with the bottlenecks in the internal network for me to set bandwidth expectations higher without good metering, but switches with 10Gbe uplinks are getting cheap, so it should be a fairly simple matter.
Now, provisioning? yeah, especially on the dedicated servers, I'm completely unacceptable. I've had a few customers that I strung along for weeks. (I mean, they got refunds... but that's just not acceptable.) - provisioning for VPSs is generally way better, It's rare that you wait more than a day or two, but still not anywhere near the linode/rackspace standard, and until I fix that, well, I'm dramatically less useful than they are.
I mean, on the metering side, right now, that is part of why the bandwidth quotas are so 2007; I mean, I /know/ that if I am having problems on my network due to overuse, someone is massively overrunning their bandwidth quota, so I can TC them down to something reasonable without breaking with the expectations I've set. So yeah, if you really do need serious amounts of bandwidth, (like, say, north of 200Mbps) on a VPS, I'm weak... but that's a whole hell of a lot of bandwidth.
(so yeah, I guess what I'm trying to say is that you are right if you are a serious bandwidth user, but most people aren't; and my policy of 'limit, and only if it's causing problems' I mean, rather than charging you for overages, makes bandwidth less of a big deal for most customers.)
Nonetheless, I am way too close to the 'will cause problems' bit with the bottlenecks in the internal network for me to set bandwidth expectations higher without good metering, but switches with 10Gbe uplinks are getting cheap, so it should be a fairly simple matter.
Now, provisioning? yeah, especially on the dedicated servers, I'm completely unacceptable. I've had a few customers that I strung along for weeks. (I mean, they got refunds... but that's just not acceptable.) - provisioning for VPSs is generally way better, It's rare that you wait more than a day or two, but still not anywhere near the linode/rackspace standard, and until I fix that, well, I'm dramatically less useful than they are.