i got a 404....could you please fix the link?
This url should work:
It's holding up better than I expected. It's a little slow at the moment, but hasn't fallen over. I was taken by surprise by the slashdot article, but so far it's working out well.
I'm not sure where the slowness is coming from, my bandwidth and cpu are both pretty under-utilized, then again I haven't tuned JBoss or Apache for this type of load, as I wasn't expecting any publicity at all.
I finally found and fixed the timer issue i was having, so things are running well for now. It's actually pretty impressive how well an out of the box install of jboss, and a first time seam developer's first app, can handle a slashdotting, following a tech.yahoo.com-dotting, following a digg.com-dotting.
My hat is off to Seam and JBoss.
Thanks for the links! I had a look at the site and am really impressed by the performance! Since there is neither too much CPU load nor "out of memory" problem under such stress, your application must handle objects really well. Congratulations!
Thank you very much! I'm very happy with the performance of the app, given the circumstances. I'll definitely keep working with Seam and learning as I go.
IS NICE! :)
Devon, could you share with us your server setup and the average CPU / memory load during this digg / slashdot period?
I think part of the reason Seam holds up so well is that it destroys objects immediately at the end of each conversation. If the app had to maintain a long life HTTP session for each user, it would have a massive memory leak in a slashdot "attack". But the tradeoff is increased CPU use sometimes. So, I'd very much appreciate your insights on the CPU load in your peak period ... Thanks in advance!
sorry for the long delay, I've been very busy at the day job lately.
The server is a 2.8 GHz Intel single proc box, with only 512 MB of Ram (if I start doing larger Seam apps, I think upgrading the ram will be a big win). It runs about 30 domains of varying traffic, and handles a few thousand e-mail a day (before 10minutemail.com anyhow). It's running debian stable, using Java 1.5.0_05, postfix, IMAPS and apache 2. I have jboss running behind apache (to avoid using an off-port and to avoid using an extra IP.
I have done zero performance tuning of apache, jboss, etc...
During the peak load (about 40,000 unique visitors, and 400,000 pages views in one day) the site was slow but responding. I'm not 100% certain why it was slow. The load average ranged from 0.5 to 2, although was mostly in the 1.0-1.2 range. There was more CPU idle than I expected. The network wasn't even close to maxxed out. I think I pushed about 1 Mbit/sec peak, and I've pushed out a full 10 Mbit from that server before on another website.
It may simply have been the high number of connections waiting to be handled that made the site seem slow, as I didn't seem to be killing the CPU.
Also keep in mind that during the same period tens of thousands of incoming e-mail were pouring in, being handled, and then served out to my application over IMAPS.
I'd like to see what I can do with a tuned set of apps, and greater understanding of what was slowing me down.