Okay i remember several years ago the term bandwidth transfers was really confusing to me, but now it's getting firm grasp on what server hits is and how it fits into the capacity of a dedicated server.

1 page has 3 images would equal = 4 hits if loaded once right ? so if a site had 12 hits per page and 300,000 pageviews/month = 3.6 million server hits/month

would stats packages such as analog and weblazier include cgi hits with server hits ?

what i am trying to understand is how much a dedicated server can handle in terms of hits. i.e.

Pentium 3 700mhz, 512mb server can handle ??? hits per month
Pentium 3 750, 512mb can handle ? hits per month
P3 800 mhz, 512mb can handle ?? hits per month
P3 866mhz 512mb can handle ?
Dual P3 800, 512mb ?
Dual P3 866, 512mb ?
Dual P3 933, 512mb ?
Dual P3 1000, 512mb ?

Currently my sites generate 550 hits per minute on my Dual P3 866, 768mb server and it's very very lightly loaded.

Anyone care to share their own experiences or theoretical opinions on how many hits the above servers can handle per minute and/or month ? :D

Im not the best on web page hits but i would personally say, it depends on what services are running on the box and which of these services the page is utilising

CGI, PHP, SQL, Etc...

if it is simply HTTP/HTML then i can imagine the server could support MILLIONS of hits per day.

Anybody else know otherwise?

I use analog and they consider .cgi requests to be hits. It calls hits requests. I imagine just about every line in the access log counts as 1 hit. " wc -l accessfilename " and an analysis of the file should come out with about the same line/hit count.

I think it would be better to count a hit as a single unique IP address, although, i can see the dificulty here with proxies, firewalls, and dhcp

eva2000.com -- It is my experience, especially with Microsoft products... that "hits" aren't that useful to track... at least for purposes of server capacity.

Static items like graphics and .html pages can be served until the 100Mbps network adapter is crammed full :) Split the network from the machine in your study.

Lots of tiny packets can congest Ethernet pretty quickly, and HTTP GETs are a big factor. So you may not have that much bandwidth in use... but your adapter can be at full capacity. FULL DUPLEX SWITCHED networking helps a lot!

I've seen machines with 1GB adapters at 5% CPU doing thousands of "hits" per minute. But again, static content like a 3K HTM page.

==========

Dynamic content, include files, any type of server-logic, database, server side variables ("sessions") -- well, that changes everything!

Microsoft servers allow you to caputure a couple of key things in the activity logs -- the data bytes sent, data bytes received, and time-taken. Combine those bits with your "hits" and you have more useful information.

Other (non Microsoft) servers can obviously do this too.... I'm just trying to give a simple answer and sticking with an example of one platform.

"hits" (in general) don't matter on server capacity -- what you are HITTING is what matters!

There, the $0.05 answer.