The new article on web hosts state that a CPU usage of 1.00 is equal to 100%. Mine often goes above 4.00. Does this mean i have a slow server?

I pinged the server in MsDos and found that mine is faster than sitepoint or even yahoo.

Also, i highly dought that my server has more than one processor as they are running using AMD CPUs.

Ping really has nothing to do with the load the server is under at any given time. Ping can give you a basic estimate of how fast it takes to send 32 bytes of data from your computer to the host machine and back. Ideally if this is under 200 ms then you should be fine.

However 32 bytes of data cannot test true performance of a server under stress.

CPU usage on the other hand tells you how much work the processor is doing.

With a TOP of 1.00 the processor is working at 100%. A modern processor and operating system can go over this amount because of page swapping (Physical RAM) and virtual memory (hard drive). However both of these technics will slow performance of your machine because they are slower than the processor.

Running TOP now on the SitePoint server I get:
12:09pm up 144 days, 7:09, 0 users, load average: 0.84, 0.58, 0.59

Which is decent for a one processor machine and the load we have.

Going to 4.00 doesn't mean you have a slow server, just an over worked one.

Okay thanx, the load on my server at the moment is:

3.09 3.87 4.09

I have seen it go past 6.00. Might be because the MySQL databses are on the same server as the accounts and thus there is allot of processing there

Exactly. Your server could have four CPUs, so then 4.00 would be about 100%.

When I was on Communitech.net the servers were constantly at 6.00 (on dual processor machines.) It was slower at running scripts but my pages still loaded acceptably. Of course, the lower the better.

Owen

He said that his server was using AMD processors. This strongly limits the ability of having multiple processors unless he meant they had the newer SMP capable Thunderbirds. Somehow I doubt this is the case. Even then AMD is only going to produce chipsets that support dual processors at first so again four processors seems highly unlikely.

On a four processor RS-6000 F80 running 4 333 Mhz PowerPC 604e processors and 1 gigabyte of RAM with 206 users (right now) and Sybase I get a load average of:
0.08 0.05 0.05

<Edited by W. Luke on 01-12-2001 at 02:59 PM>

Oh. you're right. And dang thats a good load average (or conversely, you're seriously wasting that machine's power).

I also like the uptime on the sitepoint machine. Pretty impressive to keep a production machine up for so long.

Owen

These are 800Mhz servers, i highly dought they are SMP capable and they only have 512MB of ram.

Still they are fast and have a good service.

We're wasting that machine's power. It is in the process of being phased out for a new system. An IBM RS-6000 S80. I had thought it was a 16 processor model but it only has 12 450 Mhz RS64III PowerPC processors and 64 Gigabytes of RAM. I can't telnet into that machine to see what its uptime is (stingy admins ) but it is running Oracle 8i with 16 databases and serves a maximum load of 1,000 concurrent users.

You can se the specs at: http://www.rs6000.ibm.com/hardware/enterprise/s80.html

It is a shame really... we are only using 20 PCI slots on the machine.

I tried talking them into letting me have the F50 for my home personal use when they shut it down but that meeting didn't go over so well. Still gonna try for another RS-6000 F50 but it only has 1 processor and 256 megs of RAM.
<Edited by W. Luke on 01-12-2001 at 05:27 PM>

How i would wish to have one of those....

*drool running down face*

Pete

1.00 listed in top or uptime is not 100% of proc utilization. I've been up to 26 and I have one processor. It is sort of a scale based on how many processors you have, etc. So if you had a 33mhz 486 versus a 64proc HPUX machine, both could be at 0.01 but the HPUX would be doing much more work. It's inversely proportional to the bogomips (a general proc+ram speed rating).

Of course, bogomips are 'bogo' or 'bogus' because they are generally just for fun and often not representative of a computer's capability. So take the load ratings with a grain of salt if you're using them for anything scientific.

P.S. Since we're bragging about uptimes, my 486 has been up for 56 days (since I bought it) and the load avgs are 0.01, 0.02, and 0.01. It's a 66MHz 486 with 32 megs of RAM.

Right. If a Gigahertz PC is running at 1% then it's doing more work than a 486 66 MHz running at 1%. If your computer is running at server load 26, then your computer is running very inefficiently. Your computer is at .01 because 3/4 of the time it is doing nothing - if your computer is idle it should be at or near 0.

Here we go:

So if you have two processors then the computer is able to run two processes simultaniously making 2.00 equivelent to 100%. In general (until it reaches or goes above 100%) it closely mirrors the CPU usage.

Owen
<Edited by Owen on 01-12-2001 at 06:42 PM>

I asked my host about this, and they said that 1.00 isn't 100% load. Even though they have 800 - 900Mhz AMD CPUS wioth 512MB ram.

Does the load mean that the CPU is full but with all the new technologies, (Virtual Disk etc) the server hasn't gone totally full capacity yet.

If this is right then i get it now.....

1.00 Doesn't mean 100% load on the server, what it means at the CPU is 100% loaded (With 1 CPU of course), i.e. It has no free clock cycles left, so anything else that wants to run will have to join the back of the queue.

OK,

thanx