Skip to content
Register

OVHcloud routing to Asia got worse, not better

Reviews by techmei 16 replies 7.2K views
techmei

Member OP

9 #1

I have two server with OVHcloud, one in London, one in Beauharnois. Both route to my office in Taipei through Cogent -> Telia -> some mystery hop -> Singapore -> Hong Kong -> Taipei. Before March 15, London used LINX direct to TWNIC peer. Now both path go through London.

This is not improvement. This is 400ms vs 180ms. I pay for premium bandwidth, not budget IP transit.

traceroute to 203.x.x.x, 30 hops max
 1  10.0.0.1  0.4ms
 2  ovhcloud.com  0.8ms
 3  cogentco.com  1.2ms
 4  cogentco.com  11.4ms
 5  telia.net  12.1ms
 6  * * *
 7  singtel.com  198.3ms
 8  hkt.com  245.7ms
 9  twnic.tw  398.2ms

I open ticket March 18. Reply: "routing optimization in progress, please wait." Today is March 24. No update. XD

Other Asia user please test and reply with traceroute. We need data, not patience. — https://bgp.tools

pieter_rtm

Member

1 #2

To be honest, I can confirm the change. As said, we shifted AMS-IX peering for cost reasons. Your traffic now exits via LINX to Telia.

  • Old: AMS-IX -> TWNIC direct, ~185ms median
  • New: LINX -> Telia -> Singtel, ~380ms median
  • Savings: €0.003/GB on transit

Dry humor: we saved enough for half a coffee per month per user.

I am pushing for partial rollback. The data helps. To be honest, I did not expect this much impact.

Containers before it was cool
KI

Member

#3
pieter_rtm said:
€0.003/GB

That's the savings? You ruined my ping for half a coffee? I run game servers! 180ms to 400ms is unplayable!

I have five server with you! All affected! This is fire!

Not optimization! This is cost cutting with a marketing hat!

Hana5

Member

GPU poor

#4

Techmei please make traceroute from Singapore too, I have server in Singapore, routing to Tokyo is OK still, ~230ms

Maybe only Amsterdam path is broken

Pieter_rtm, please give option for old path, I pay extra for premium please

8-char NTLM found in 4min 32sec
larryjeong

Member

#5

Half a coffee is insulting honestly, I'd be furious too

$3/year. 128MB RAM. Pure happiness.
wendy

Member

#6

Did you test from both London and Limburg, or just one?

Have you tried restarting it?
Nadia

Member

/dev/null enthusiast

#7

Did this ever get piped to a solution or did they just redirect the whole rollout to /dev/null

I was about to order in Amsterdam but this thread is four days old with nothing new from Pieter

"pushing" does not sound like a committed timeline to me

/dev/null: full of good ideas
LE

Member

#8

Did anyone get the partial rollback Pieter mentioned, or are we all still routing through London

I am on Leaseweb now but this thread made me check my old OVHcloud box, still 370ms to Singapore from Gravelines

Seems like the cost saving won

bellaauc

Member

3-2-1 Believer

#9

Did anyone get the partial rollback Pieter mentioned? I have a GreenCloudVPS box in Amsterdam and the routing to Singapore still looks weird from here. Not sure if related.

wendy said:
Did you test from both London and Limburg, or just one?

Also curious about this. My OVHcloud London box went from clean to messy around the same time but I never checked Frankfurt.

3-2-1 or you're already dead
craigdzir

Member

8-bit forever

#10

86 days and no rollback? Did this ever get fixed or are we all still routing the long way round

I was about to move a project to OVHcloud London but this thread killed that idea

Post a reply

You need an account to reply. Log in or register to join the conversation.

Post reply Preview Save draft