dslreports logo
site
 
    All Forums Hot Topics Gallery
spc

spacer




how-to block ads


Search Topic:
uniqs
2217
share rss forum feed


rocca
Start.ca
Premium
join:2008-11-16
London, ON
kudos:22

TorIX Upgrade

For those interested, we completed our TorIX upgrade to 10Gbps tonight.


merc782

join:2012-11-19

Nice. any plans to go into Nyix or chix, which might reduce your transit cost .

also please update your network records in peeringdb.com, it will help in future



PajaPatak

@start.ca
reply to rocca

What did you upgrade it from?

Also, r u going to add any other uplink?



rocca
Start.ca
Premium
join:2008-11-16
London, ON
kudos:22

It went from 1Gbps to 10Gbps.

We are connecting nLayer next week, and another one shortly.



JCohen
Premium
join:2010-10-19
Nepean, ON
kudos:6
Reviews:
·Start Communicat..
·TekSavvy Cable
·Rogers Hi-Speed

said by rocca:

We are connecting nLayer next week, and another one shortly.

I'm looking forward to the nLayer peering, gonna be nice to have lower latency to my hosted servers.

hmph

join:2012-10-23
reply to rocca

peering or transit?



rocca
Start.ca
Premium
join:2008-11-16
London, ON
kudos:22

For nLayer, transit.

For TorIX, peering.


hmph

join:2012-10-23

Damn that alone has me considering switching from Teksavvy.
How big of a link do you have to nLayer?
Any chance the other provider you are going to be adding shortly will be Zayo, Level3, or Inteliquent?



rocca
Start.ca
Premium
join:2008-11-16
London, ON
kudos:22

said by hmph:

How big of a link do you have to nLayer?

Starting with a gig.

said by hmph:

Any chance the other provider you are going to be adding shortly will be Zayo, Level3, or Inteliquent?

Sorry, can't say who until the contracts are done.

hmph

join:2012-10-23
reply to rocca

Any reason you are not peering with OVH Canada via TorIX?
Traffic seems to be going all over the place via bell.


dan mackay

join:2008-10-11
Barrie, ON
reply to rocca

I want HE to not route me to my nyc server by going to chicago first, but I doubt that will ever happen :P


hmph

join:2012-10-23

Toronto connects to NYC by either: Toronto-> montreal -> NYC or Toronto->Chicago->NYC
Chicago has a lot more capacity then montreal, so it makes sense for HE to route it that way.



rocca
Start.ca
Premium
join:2008-11-16
London, ON
kudos:22
reply to hmph

We do peer at TorIX, perhaps they aren't advertising all their routes there?


hmph

join:2012-10-23
reply to dan mackay

After taking a second look it does seem like there is a connection from Toronto directly to NYC, I guess HE is just poorly routing traffic to the network you are on.


Navier

join:2012-10-29
reply to rocca

I googled what Torix is. Toronto internet exchange. then we haveNY and Chicago. But what exactly does this do and why such a select few are excited about it.
Just for curiosity sake.


hmph

join:2012-10-23
reply to rocca

I am being routed to your TorIX IP via OVH but only to your TorIX assigned IP, not sure why regular traffic is not routing through there as well.

Goes through TorIX here:
ping 206.108.34.187
PING 206.108.34.187 (206.108.34.187) 56(84) bytes of data.
64 bytes from 206.108.34.187: icmp_req=1 ttl=251 time=8.60 ms
64 bytes from 206.108.34.187: icmp_req=2 ttl=251 time=8.64 ms
64 bytes from 206.108.34.187: icmp_req=3 ttl=251 time=8.71 ms
64 bytes from 206.108.34.187: icmp_req=4 ttl=251 time=8.93 ms

Traceroute is pretty much blocked for this route, so I won't bother.

Goes over bell here:

ping start.ca
PING start.ca (204.101.248.50) 56(84) bytes of data.
64 bytes from ws01-010.start.ca (204.101.248.50): icmp_req=1 ttl=58 time=24.1 ms
64 bytes from ws01-010.start.ca (204.101.248.50): icmp_req=2 ttl=58 time=23.9 ms
64 bytes from ws01-010.start.ca (204.101.248.50): icmp_req=3 ttl=58 time=21.9 ms
64 bytes from ws01-010.start.ca (204.101.248.50): icmp_req=4 ttl=58 time=22.0 ms

start.ca
traceroute to start.ca (204.101.248.50), 30 hops max, 60 byte packets
1 > (>) 0.726 ms * *
2 bhs-g1-6k.qc.ca (198.27.73.17) 0.673 ms * *
3 mtl-2-6k.qc.ca (198.27.73.4) 2.206 ms * *
4 198.27.73.178 (198.27.73.178) 146.046 ms * *
5 bx5-chicagodt.bell.ca (206.223.119.66) 17.936 ms 17.947 ms 18.217 ms
6 newcore1-chicagocp_xe-4-0-0_0.net.bell.ca (64.230.186.105) 51.864 ms 33.518 ms newcore1-chicagocp_xe3-1-1.net.bell.ca (64.230.186.81) 31.535 ms
7 tcore4-toronto12_POS0-6-0-0.net.bell.ca (64.230.147.173) 39.216 ms tcore3-toronto12_POS0-6-0-0.net.bell.ca (64.230.147.165) 35.608 ms tcore3-toronto47_POS0-2-0-0.net.bell.ca (64.230.186.125) 37.862 ms
8 CORE1-LONDON14_POS4-0-0.net.bell.ca (64.230.149.225) 35.577 ms CORE2-LONDON14_POS12-0-0.net.bell.ca (64.230.149.221) 34.011 ms CORE1-LONDON14_POS4-0-0.net.bell.ca (64.230.149.225) 35.839 ms
9 dis20-london14_7-0-0.net.bell.ca (64.230.112.6) 69.630 ms 69.593 ms 69.612 ms
10 ws01-010.start.ca (204.101.248.50) 23.782 ms 23.765 ms 23.987 ms



rocca
Start.ca
Premium
join:2008-11-16
London, ON
kudos:22
reply to Navier

It's a peering exchange where networks can see each other directly without having to go out to the 'general internet' to pass traffic. Generally this results in shorter paths, lower latency, etc for both networks involved in the transmission.



rocca
Start.ca
Premium
join:2008-11-16
London, ON
kudos:22
reply to hmph

Hard to say, we advertise our space to them via the TorIX fabric, they must have some reason they are selecting our Bell route instead. If you want to PM me your colo box IP I can see what the reverse looks like.


merc782

join:2012-11-19
reply to rocca

your IP range is advertised by both your AS and Bell. i think thats the reason for the routing issue

»bgp.he.net/ip/204.101.248.50

@OP, try traceroute to ns1.start.ca 64.140.120.21, this is annouced only by Start, not Bell


hmph

join:2012-10-23

merc, it had the same outcome. Start buys Bell transit (at least I think they do considering Bell doesn't connect to anyone unless they are paid quite a bit) so it is still in their BGP mix. It is possible that in this case it could be OVH's fault. Currently waiting for trace from Start->OVH path to determine if both the paths are identical.



rocca
Start.ca
Premium
join:2008-11-16
London, ON
kudos:22

traceroute to ## , 30 hops max, 40 byte packets
1 core1-london1-ge1-120.net.start.ca (64.140.120.1) 1.275 ms 0.268 ms 0.281 ms
2 core1-toronto1-te3-2.net.start.ca (64.140.112.86) 4.915 ms 5.778 ms 4.538 ms
3 gw-ovh.torontointernetxchange.net (206.108.34.189) 4.418 ms 7.607 ms 4.282 ms
4 mtl-2-6k.qc.ca (178.32.135.71) 22.669 ms 22.836 ms 22.479 ms
5 bhs-g2-6k.qc.ca (198.27.73.7) 23.755 ms * 23.654 ms
6 bhs-1-6k.qc.ca (198.27.73.14) 59.77 ms * 30.847 ms
7 ## 45.651 ms 23.579 ms 21.713 ms


hmph

join:2012-10-23

Alright, OVH's fault. Start is routing nicely. Time to complain to OVH...



jmck
formerly 'shaded'

join:2010-10-02
Ottawa, ON

are you actually seeing performance issues going from ovh to start? I ran some tests a few weeks ago and saw none. if ovh is having any capacity or performance issues going to start through TorIX you might not want them to statically route it.


hmph

join:2012-10-23

1 edit

OVH has 2x 10gbit connections to TorIX that generally peak at 10gbits. Start has 10gbit to TorIX with large amounts of spare capacity. I have already opened a ticket with OVH to make the change. In the end this should shave off ~10ms from the route. Why allow Bell to route the traffic all over the place when you can have a direct route between Start and OVH. It's a lot more cost effective and better for all end users performance wise including you. This is what TorIX is for after all...



jmck
formerly 'shaded'

join:2010-10-02
Ottawa, ON
Reviews:
·TekSavvy DSL
·Start Communicat..

i understand that, and usually providers will give preference to public/private peers to cut transit costs too so my concern is that they're avoiding TorIX for performance reasons.

I've talked to OVH network group a bit the last few months and managed to get TSI and OVH to setup a private peer at TorIX to get around bad performance through the public peer. They would likely be open setting up a private peer too with Start. If rocca's interested, he can reach their network group by contacting support@ovh.ca.


ultramancool

join:2004-12-22
Schenectady, NY

Uhh, trace from teksavvy to OVH, it goes through torix over public peering.

Tracing ...........*T
TTL LFT trace to XXXXXXX.cable.teksavvy.com (XX.XX.XX.XX):80/tcp
1 192.95.27.253 1.4ms
2 bhs-g1-6k.qc.ca (198.27.73.13) 0.5ms
3 mtl-2-6k.qc.ca (198.27.73.4) 2.1ms
4 tor-1-6k.on.ca (178.32.135.70) 9.8ms
5 gw-teksavvy.torontointernetxchange.net (206.108.34.67) 8.9ms
6 2120.ae0.bdr02.tor.packetflow.ca (69.196.136.65) 65.3ms
7 richmond3.cable.teksavvy.com (24.52.255.74) 15.0ms



rocca
Start.ca
Premium
join:2008-11-16
London, ON
kudos:22
reply to hmph

BGP makes path selection by prefix size, and then AS path length (not number of hops) and it appears they are directly connected to Bell in their Montreal DC, so their router looks at these options:

577 (Bell) -> 40788 (Start)
11670 (TorIX) -> 40788 (Start)

...so you have equal length paths, one you have to use your backhaul to Toronto for and one you get to dump off in Montreal. Before the holidays we sent them a direct peer request via the TorIX portal, didn't get a response - I'll followup sometime next week. If we did directly peer then the TorIX ASN becomes removed from the equation and it'd just look like:

577 40788
40788

...in which case the later would be selected (assuming they don't pad out their connection to Toronto).


k3d3

join:2013-01-18
reply to jmck

shaded, if start has a fairly little used 10gbit torix link and OVH has a half-used 20gbit (total) link, then you will pretty much be guaranteed a good link. In this case, performance is not the reason Bell is preferred. Also, this is on OVH's end, though you've also mentioned this.

rocca, I am currently with TekSavvy. If I switch to Start, do I incur the setup fee?

Thanks



rocca
Start.ca
Premium
join:2008-11-16
London, ON
kudos:22

said by k3d3:

If I switch to Start, do I incur the setup fee?

No if you have an active cable internet connection at the same address and just transferring it to us then there is no installation fee applicable.

Johny

join:2012-12-06
Reviews:
·Start Communicat..
reply to rocca

Looks like you guys got nlayer hooked up, and with that I've noticed a 70ms increase to a certain ip range that I use alot. What's up with that?

Tracing route to unknown.iad.scnet.net [75.102.43.162]
over a maximum of 30 hops:

1 16 ms 9 ms 10 ms 10.106.19.129
2 12 ms 11 ms 11 ms 69.63.242.29
3 38 ms 11 ms 23 ms 69.63.249.173
4 17 ms 19 ms 23 ms v514.core1.tor1.he.net [216.66.0.53]
5 13 ms 12 ms 25 ms torix.tge2-3.ar1.tor1.ca.nlayer.net [206.108.34.
53]
6 75 ms 26 ms 25 ms xe-1-0-1.cr1.nyc3.us.nlayer.net [69.22.142.61]
7 32 ms 29 ms 38 ms xe-0-2-1.cr1.iad1.us.nlayer.net [69.22.142.92]
8 134 ms 148 ms 137 ms as23352.xe-1-3-0-101.cr1.iad1.us.nlayer.net [69.
31.30.21]
9 44 ms 60 ms 46 ms 50.ae0.cr2.iad1.us.scnet.net [64.202.119.69]
10 38 ms 33 ms 42 ms as36352.ge-1-1-5.cr2.iad1.us.scnet.net [75.102.4
3.122]
11 119 ms 116 ms 122 ms unknown.iad.scnet.net [75.102.43.162]

Trace complete.