toddf: (short answer on the MTU stuff is that it also needs to be applied to kvm's tap bridge... At least, that's what was needed when setting up the second NIC on my vps to interface with the backup server)
up_the_irons: my speed is faster on nlayer than ntt
err wasf aster
oh my normal ip going through ntt too now
yeah i announced that /20 while i'm still here
up_the_irons: steadfast have 10 gigabit to ntt i'm pretty sure
so it's just ntt transitting back to los angeles, so no reason it should be slower ?
maybe i'll try udp test
yeah some loss receiving no loss sending
but way lower jitter receiving
is it actually gigabit to the nodes from the new router now?
jbergstroem: the IO scheduler is mostly concerned about when to perform each disk IO operation (the IO scheduler is only in the disk/block device code). The noop scheduler just dumps to disk immediately. cfq=completely fair queuing tries to group operations that would be near each other on the HDD (optimizing for disk-latency), and it also keeps disk IO spread evenly between users. (There's also deadline, which I'm told is good for ...
... databases because it puts reads above writes, and anticipatory... which I don't know much about)
brycec: i've found deadlien to work well generally
I use noop... but that's because my system disks are SSDs :D (otherwise I just leave it defaulted to cfq)
i've never found any scheduler that makes linux magically work well on sata hard-disks with mixed read/write sequential loads though
mercutio: no gigabit to the vm hosts yet
up_the_irons: ahh ok
jitter is pretty good then really
i wonder if the tinet packet loss is still there'
mercutio: what host are you on?  maybe i'll run a cable to it tomorrow (will be at data center) and we can experiment.  if i bridge the new cable to existing host, should be pretty straightforward
kvr15 i think
i'll check panel
yeah 15
hmm tinet isn't as lossy as before
but it's still routing via san jose
and it's still lossy
mercutio: brycec : ok, i'll run a cable for kvr07 and kvr15
i wish telstraglobal would just peer freely on any2ix
yep
i wonder if they would
up_the_irons: oh cool!
brycec: not promising i'll have time to make the whole setup actually work, but i'll try :)
up_the_irons: could i get bgp session to the second router too?
maybe on private ips for link address
i think we could work something out.  i may price a second session at $5
well everything is going to move over to the second one later anyway isn't it?
actually is bridging going to work to get gigabit to the host?
won't it have to go through the gateway unless the other router is on the same /30 ?
ie "hiding on one of the extra ip's"
i suppose it could flow from old router to new router via gigabit (or more?) then from new router over gigabit to the kvm host
but ntt traffic would still have to come in new router, then route through the old router
mercutio: your vlan would have to terminate on the new router, yes
mercutio: at some point, i will migrate all vlans to the new router, then replace the 10/100 line card of the old router with 10/100/1000 card, then migrate back.
why not keep vlans on the new router, you ask?  well, with over 4 years of excellent service with the old router, i'm gonna say "if it ain't broke, don't fix it"
at least, that is my thinking today...
My $.02, up_the_irons - Move to the new router, and leave the old one running. Should the new router explode, then you can easily switch back.
brycec: roger
apple fritter iced latte hello friday
amirite
phlux: lol
we now have 206.125.168.0/21 announced to NTT
:o
yes, phlux, you are right