Forum OpenACS Q&A: Stress-testing OpenACS sometime?

Collapse
1: Stress-testing OpenACS sometime?
Posted by Gregory McMullan on 05/11/00 12:00 AM

In responding to a question about "Why should I consider Postgres instead of Oracle for my ACS system?" (see http://www.photo.net/bboard/q-and-a-fetch-msg.tcl?msg_id=000saW&topic_id=21&topic=web%2fdb for more), I had a possibly-crazy thought. Do we want to consider setting up a box with a decent OpenACS setup somewhere and post to say photo.net and slashdot asking people to actively use it for a while, so that we can see what kind of load it can actually take? Of course, there are considerations like bandwidth and such, and who pays if hosting limits are overrun. This seems like something that would be handy in tooting our horns later - "We survived slashdot on a mere PII-300 (or whatever)!"

No, I don't have time to put more into this at the moment - maybe in a few weeks I will, though. Just some food for thought...

Collapse
2: Response to Stress-testing OpenACS sometime? (response to 1)
Posted by Ben Adida on 05/11/00 12:16 AM
a good idea. We just got our free server from ArsDigita. Now that
PG 7.0 is out, I'm going to set it up. We can stress-test on that
box, no problem. It should be interesting...

Note that the best way to Slashdot our system is to say bad
things about MySQL... and since my rebuttal is coming up, it
might make a good bait to make this happen, again...

Collapse
3: Response to Stress-testing OpenACS sometime? (response to 1)
Posted by Gregory McMullan on 05/11/00 02:58 AM
Yes, baiting slashdotters is an easy way to get lots of traffic. Enough so that I thought about suggesting that we could use the tactic in my initial message.  I decided against it because I am very tired of the cycle of taunting that baiting engenders.  I'd much rather that we consciously chose to take the high road, and refrained from badmouthing anyone as much as possible.

This isn't to say that we can't post something to slashdot that says "We've revised our explanation about why OpenACS doesn't use MySQL here.  We're testing a new server, too, so please feel free to come in, look around, and use the features.  If we can't provide good response time under a good slashdot load, maybe we'll re-evaluate our use of Postgres".

Collapse
4: Response to Stress-testing OpenACS sometime? (response to 1)
Posted by Don Baccus on 05/11/00 08:29 PM
I couldn't help but notice that Slashdot itself was quite slow earlier today, despite running MySQL on a quad P500 box with .5 or 1 GB RAM, fronted by two dedicated boxes to serve static pages.  Brand new hardware all around.  Oh well, at least you could get in, that's an improvement.

If we really want to test load, we could always post "Why the AOLserver team chose Tcl rather than Perl" and "Why we don't use Java to generate content" along with Ben's response to the MySQL fanatics.  Post 'em all at once and the box will take a few hits...

When I tested my P200 classic with AOLserver and PG a year or so ago, I was able to serve two insert pages and ten select-only pages a second, total of 200 selects and 40 inserts a second (two long transactions in each insert page).  I just tell people that and point out my new dual P450 will serve a heck of a lot more, and don't worry about it...

Collapse
5: Response to Stress-testing OpenACS sometime? (response to 1)
Posted by John-Mason Shackelford on 05/20/00 07:47 AM
It seems that a good shell script could do the job. If a few people were running multiple instances of a shell script that would make  rapid-fire requests, you'd know what you wanted to know without stirring up trouble :)