beefing up proxy server performance...

[Steve Holdoway]

I've got a problem with slow downloading files... they're coming from
Amazon, and if I test against my Canadian server, I can get speeds of
c.1.8MB/s. Here on voda it's down to c.250KB/sec, and if I use SNAP it's
about 100KB/sec.

However, if I parallelise - for example using aria2c - then i can
increase the download speed... at least doubling it.

Does anyone know of a way of implementing this across the board?
Integrate the concept with a proxy server maybe? Can't seem to find
anything on google that suggests it's possible...

Cheers,

Steve

[Hadley Rich]

Post by Steve Holdoway
and if I test against my Canadian server, I can get speeds of
c.1.8MB/s. Here on voda it's down to c.250KB/sec, and if I use SNAP
it's about 100KB/sec.
However, if I parallelise - for example using aria2c - then i can
increase the download speed... at least doubling it.

I had this issue with Snap's International transit 18 months ago. Very
poor performance but if I opened multiple TCP connections I could get
nearly 1MB/s. Their helpdesk at the time said that they knew about it
but it wasn't being treated as an issue. I solved the problem by moving
providers.

hads

[Steve Holdoway]

Post by Hadley Rich

Post by Steve Holdoway
and if I test against my Canadian server, I can get speeds of
c.1.8MB/s. Here on voda it's down to c.250KB/sec, and if I use SNAP
it's about 100KB/sec.
However, if I parallelise - for example using aria2c - then i can
increase the download speed... at least doubling it.

I had this issue with Snap's International transit 18 months ago. Very
poor performance but if I opened multiple TCP connections I could get
nearly 1MB/s. Their helpdesk at the time said that they knew about it
but it wasn't being treated as an issue. I solved the problem by moving
providers.
hads

Looks like that was a wise choice then... ):

Steve

[Mark Foster]

Max speed on a TCP Session basis is largely determined by Latency.

https://secure.wikimedia.org/wikipedia/en/wiki/Network_performance#Relationship_between_latency_and_throughput

https://secure.wikimedia.org/wikipedia/en/wiki/Bandwidth-delay_product

http://www.speedguide.net/faq_in_q.php?qid=93

If the bottleneck you're seeing is associated with the aggregate amount of available bandwidth to the destination that the ISP has access to, then you might see a performance shift if you move ISPs.
However if your baseline latency end to end doesn't change (or change much) after the shift, then Latency is probably involved.
Of course this is also mangled by the actions some ISPs take with their traffic (optimization above layer 3).

Mark.


-----Original Message-----
From: nzlug-***@linux.net.nz [mailto:nzlug-***@linux.net.nz] On Behalf Of Steve Holdoway
Sent: Friday, 29 April 2011 1:36 p.m.
To: NZLUG Mailing List
Subject: Re: [nzlug] beefing up proxy server performance...

Post by Hadley Rich

Post by Steve Holdoway
and if I test against my Canadian server, I can get speeds of
c.1.8MB/s. Here on voda it's down to c.250KB/sec, and if I use SNAP
it's about 100KB/sec.
However, if I parallelise - for example using aria2c - then i can
increase the download speed... at least doubling it.

I had this issue with Snap's International transit 18 months ago. Very
poor performance but if I opened multiple TCP connections I could get
nearly 1MB/s. Their helpdesk at the time said that they knew about it
but it wasn't being treated as an issue. I solved the problem by
moving providers.
hads

Looks like that was a wise choice then... ):

Steve

--
Steve Holdoway BSc(Hons) MNZCS <***@greengecko.co.nz> http://www.greengecko.co.nz
MSN: ***@greengecko.co.nz
Skype: sholdowa


_______________________________________________
NZLUG mailing list ***@linux.net.nz
http://www.linux.net.nz/cgi-bin/mailman/listinfo/nzlug

[Steve Holdoway]

fyi latency via voda = c. 275 ms, snap = c. 220 ms ( and my bionic
Canadian server is 23ms ). So it's more than just that as voda shows a
lower latency than snap.

Are you suggesting I use some kind of netcat-based solution? I've tried
lowering 'latency' by tunnelling to buffers on a *BSD server in the USA
over openvpn on tcp/80 before, but it's a load of work just to game a
system that really shouldn't need gaming at all!


Steve

Post by Mark Foster
Max speed on a TCP Session basis is largely determined by Latency.
https://secure.wikimedia.org/wikipedia/en/wiki/Network_performance#Relationship_between_latency_and_throughput
https://secure.wikimedia.org/wikipedia/en/wiki/Bandwidth-delay_product
http://www.speedguide.net/faq_in_q.php?qid=93
If the bottleneck you're seeing is associated with the aggregate amount of available bandwidth to the destination that the ISP has access to, then you might see a performance shift if you move ISPs.
However if your baseline latency end to end doesn't change (or change much) after the shift, then Latency is probably involved.
Of course this is also mangled by the actions some ISPs take with their traffic (optimization above layer 3).
Mark.
-----Original Message-----
Sent: Friday, 29 April 2011 1:36 p.m.
To: NZLUG Mailing List
Subject: Re: [nzlug] beefing up proxy server performance...

Post by Hadley Rich

Post by Steve Holdoway
and if I test against my Canadian server, I can get speeds of
c.1.8MB/s. Here on voda it's down to c.250KB/sec, and if I use SNAP
it's about 100KB/sec.
However, if I parallelise - for example using aria2c - then i can
increase the download speed... at least doubling it.

I had this issue with Snap's International transit 18 months ago. Very
poor performance but if I opened multiple TCP connections I could get
nearly 1MB/s. Their helpdesk at the time said that they knew about it
but it wasn't being treated as an issue. I solved the problem by
moving providers.
hads

Steve
--
Skype: sholdowa
_______________________________________________
http://www.linux.net.nz/cgi-bin/mailman/listinfo/nzlug