30,000 at the exact same second? Meaning..your server can handle 30,000 hits every 1 second, so in 5 seconds, 150,000 hits would happen, etc?
 
I did a simple load-test on a login process (which actually is a transaction..goes to the database, etc), and I was able to achieve 4 million hits an hour on a PIII650 with 256MB RAM and ATA33 hd. Figure a dual-cpu 1.5Ghz Intel (or 1.2Ghz Athlon), with 2GB memory, and raid 3 SCSI (I think thats the best raid config for speed..not sure though) hd's, you should be able to increase that to somewhere around 30 million hits for one server.
 
Yes, you will want to do clustering. It will allow you to easily add new servers (with identical configurations) to handle more load, but it also allows you to do Session level fail-over..so if one server went down for any reason, the other would still stay alive, making sure any current transactions don't get lost. Very good for 24/7 up-time, and the VC's like to see you spending their money on this type of config.
 
As for 30,000 hits per second, I imagine you'll need quite a farm. Are these just page hits of "static" content, or actual transactions involving databases? If they are transactions, I would definitely look into using hardware SSL devices to speed up secure transactions. There are devices that do I think on the order of 6000 SSL transactions per second, but they are about $25K or so in price (maybe less).
 
Hope that helps.
 
-----Original Message-----
From: John Flores [mailto:[EMAIL PROTECTED]]
Sent: Friday, February 02, 2001 11:15 AM
To: Orion-Interest
Subject: Scalability

Hello-
 
I am very new to J2EE enterprise server world.  I want to construct a website that will capable of serving 30,000 clients at time.  What is the maximum number of clients one Orion production site can support and what is the recommended number of Orion sites needed to support 30,000 clients?  Would this involve clustering? All help is appreciated. Thanks.
 
John Flores

Reply via email to