Re: (FWD) Results of stress testing the current version of the Onion Router

[Bruce Montrose <montrose@itd.nrl.navy.mil>]

At 03:19 PM 9/5/2002 -0400, you wrote:
>On Thu, Sep 05, 2002 at 02:05:16PM -0400, Bruce Montrose wrote:
> > Check out hps.pdf and fps.pdf reports at http://freehaven.net/~montrose/
> >
> > I think these custom reports are much more telling than the previous 
> canned
> > ones.
>
>Well, it seems to agree with my earlier guess that it was about 3 ..so
>that's a good sign :)
>What's min-test and max-test?
>
>What are each of these load units doing? Pulling down one file? How
>big is the file? Can you try it with small files and again with large

What is going on here is that each Load Unit (virtual client) repeats a 75k
transaction over and over until the test terminates. The test creates more
load units incrementally over time until there are a total of 256 running.

The min-test would be the minimum hits-per-second or failures-per-second
respectively over all the transactions at that particular load size.

I think I will create a test that keeps the load size constant and ramps up
the transaction size. That way we can easily determine the optimal transaction
size. I have to write a different WebLoad program to accomplish this.

Once I have this test created, I will run it with various settings for MaxConn
so we can determine the optimal setting for that parameter as well.

>files? (That will give us an idea of whether the symmetric crypto is
>playing any part.) Also, how about making MaxConn on the routers something
>like 40 -- does that help or hurt our 3 transactions/second figure?
>
>--Roger