I've decided to run some speed tests.
A while back while testing some kernel things for 'Psycho, I found a website ( Pingdom Tools ) that measures webpage load times. As far as I can tell from the site, it uses YOUR Internet connection, not theirs. The evidence from the above link to the kernel post and the following data supports that. The site supports saving tests for future reading/analyzing. The links are included in an attachment.
- Phone: Pre-
- OS: webOS 2.1
- Kernel: F105-71
- Carrier: Sprint
The attached files are:
- "Tests_And_Links.txt": The tests and links to the test I ran.
- "Speed_Test_Results_CSV.txt": A spreadsheet (in csv form), sorted by Time ascending, showing various times for various tests. Any Spreadsheet app should be able to import a csv just fine.
I tested changing only MaxConcurrent and MaxRecvSpeed. I did not play with MaxQueueLength as I don't really know what that means in this context.
I used "msnbc.com" as the test website. I figured it was a bloated enough page for "real world" testing.
- Change value
- Clear Browser Cache
- Run test
Results (Sorted by Time Ascending):
I think the data speaks for itself. The "Cap" of course is "65536" (Palm's setting), and un-capped is "-1" (Jason's setting). I also tested with "disabling" a hard-coded MaxConcurrent number, those tests are the "-1" in the "Threads" column.
I ran each MaxConcurrent change with and without a hard-coded MaxReceive number. The difference there is obvious. It seems that MaxConcurrent > 10, it starts to fall down. I didn't test higher than 15 MaxConcurrent. It's entirely possible that devices that have more RAM than my whimpy Pre- would perform better above 10 MaxConcurrent (or a Pre- running webOS 1.4.5 - there's MUCH more available RAM there).
I hope people find this interesting and maybe pass this info on to others to dissect as well.
For those on Windows, you'll have to open the "Tests_And_Links.txt" file with WordPad (or Notepad++) because I use Linux and Windows uses a different EOL than the rest of the world does.