Current time: 08-17-2017, 11:36 PM Hello There, Guest! (LoginRegister)

Post Reply 
 
Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
Best EC2 istance to approximate average user PC?
06-13-2017, 09:03 PM (This post was last modified: 06-13-2017 09:27 PM by datadiggers.)
Post: #1
Best EC2 istance to approximate average user PC?
I just realized this post fits better in Private Instances subforum, can it be moved there? Thanks.

Is anyone aware of recent-ish study on performance of typical user PC? How many cores it has? I googled for it but haven't found much.

I'm trying to find the right EC2 instance for test clients. Currently I'm using m3.medium and I feel the results are not representative of average user experience. It seems to perform certain tasks really slow - for example the same page (inlined CSS, CSS purged of unused rules, no images, no js, no complex layout, no layout trashing, no CSS animations) served through HTTPS vs HTTPS + HTTP/2 Push (with ~ 30kb of pushed content) has difference in Speed Index of 200 ( 500-700 vs 800-900) - verified with ~ 500 test runs per each scenario. I thought it might be due additional work CPU has to do to unzip SVGs I was pushing so I checked with PNGs with the same results. Note that those images are only pushed through pipe, they are not used on page (so no decoding should be involved). Also m3.medium seems to take a lot of time to do initial Layout (Chrome timeline) - 100ms+ (on my X1100T Amd it takes several ms).

What's more I find it almost impossible to go below SpeedIndex of 500 on m3.medium (granted, 200-300 of SI is due to HTTPS connection, but it also might be due to poor CPU performance as on my PC the whole connection takes ~150ms)

I have my guesses on why m3.medium performs this way, however please note that the question is about "best" EC2 instance to approximate average user PC, not "why my SpeedIndex is slow".
Find all posts by this user
Quote this message in a reply
06-14-2017, 12:16 PM
Post: #2
RE: Best EC2 istance to approximate average user PC?
(06-13-2017 09:03 PM)datadiggers Wrote:  I just realized this post fits better in Private Instances subforum, can it be moved there? Thanks.

Is anyone aware of recent-ish study on performance of typical user PC? How many cores it has? I googled for it but haven't found much.

I'm trying to find the right EC2 instance for test clients. Currently I'm using m3.medium and I feel the results are not representative of average user experience. It seems to perform certain tasks really slow - for example the same page (inlined CSS, CSS purged of unused rules, no images, no js, no complex layout, no layout trashing, no CSS animations) served through HTTPS vs HTTPS + HTTP/2 Push (with ~ 30kb of pushed content) has difference in Speed Index of 200 ( 500-700 vs 800-900) - verified with ~ 500 test runs per each scenario. I thought it might be due additional work CPU has to do to unzip SVGs I was pushing so I checked with PNGs with the same results. Note that those images are only pushed through pipe, they are not used on page (so no decoding should be involved). Also m3.medium seems to take a lot of time to do initial Layout (Chrome timeline) - 100ms+ (on my X1100T Amd it takes several ms).

What's more I find it almost impossible to go below SpeedIndex of 500 on m3.medium (granted, 200-300 of SI is due to HTTPS connection, but it also might be due to poor CPU performance as on my PC the whole connection takes ~150ms)

I have my guesses on why m3.medium performs this way, however please note that the question is about "best" EC2 instance to approximate average user PC, not "why my SpeedIndex is slow".

By default I believe WPT creates m1.smalls in ec2, I've always updated this setting to use m3.larges because they were the smallest to offer 2 vCPU cores. I believe 2 vCPUs actually 1 single core. They have a bit too much RAM but that's a different story...


vCPUs on AWS are a bit opaque but there has been some experimentation:
https://www.pythian.com/blog/virtual-cpu...-services/
Find all posts by this user
Quote this message in a reply
Post Reply 


Forum Jump:


User(s) browsing this thread: 1 Guest(s)