25594 - socket transport degredation

Status: VERIFIED FIXED

(Core :: Networking, defect, P3)

Description

[Judson Valeski]

A few months ago Rick noted that socket transport performace took a nose-dive. 
If I recall correctly his tests showed an 8-10x slowdown. I'm filing this bug to 
track this occurrance.

Comment 1

[Chris Waterson]

bug grafitti

Comment 2

[Chris Waterson]

how are you guys measuring this?

Comment 3

[Judson Valeski]

I don't recall the details. I believe rick noticed this running one of the 
socket test harnesses.

Comment 4

[leger]

tever, find out from engineers best way to measure this and let us know hos much 
slower than 4.x

Comment 5

[Phil Peterson]

PDT needs data to make a call on beta1. 

Comment 6

[rpotts (gone)]

I was using the testprotocols.exe test program.  This program acts as a 
standalone data sink for necko that simply throws away the data as it arrives 
from the network.

The benchmark I was using was to download a 10mb file via http.  I was using 
the following command: testprotocols http://warp/u/rpotts/big.html

The last time I looked, it was taking several minutes for testprotocols to 
download this file...  However, using today's tip, I got a much more reasonable 
times of < 30 seconds.

Here is some data from my laptop (using a 10 MBit ethernet card)
  Throughput: 3564202 bps  (Read time: 24.405 seconds)
  Throughput: 7449204 bps  (Read time: 11.667 seconds)
  Throughput: 7367185 bps  (Read time: 11.807 seconds)

If I remember correctly, the old netlib could process this file in about 30 
seconds (via a similar stand-alone test app called nettest.exe)

So, it looks like socket transport problem that I was seeing has gone away :-)
 

Comment 7

[Tom Everingham]

Well, using the same program that Rick used I am getting similar
results.  3 runs 

7385952 bps  11.777 sec
7442830 bps  11.687 sec
7323765 bps  11.877 sec

Not too shabby.  Marking as fixed
 


 Look's pretty decent so I am going to close this one down.

Comment 8

[Tom Everingham]

verified

Comment 9

[leger]

Putting on PDT- radar for beta1.

Comment 10

[BMO Automation]

Attachment: [mozilla-mobile/fenix] CI for 25594 (#28719)