Dear all,

I am designing a course where I teach the basics of PC hardware (architecture, CPU, memory, etc.).
In the laboratory I would like the students to measure the performance under certain circumstances.
- Linux + MySQL Server
- Linux + PostgreSQL Server
- Linux + Apache Server
- Linux + Samba Server
- etc.

Does somebody know of scripts or tools available that could help me design the course? I am thinking of using e.g. the tool ab with Apache, keep track of the CPU usage with top, etc.

Any hint or idea will be appreciated.

Thank you in advance.
Regards, Carsten

I've done a bit of this. "top" can be misleading, helpfull or counter intuiive. For example when tunning a DBMS server like Postgresql you _want_ the CPU uliazation to be as high as you can get it. Low is not good.

I tested Postgres v. Mysql and found (of course) mysql was much faster at simply querries but then we added many simultainous simulated users. When 10 users were beating up on the database MySQL handled the locking poorly letting only one user in at a time. (all users doing n-way joins and updates)

Bottom line is that your students should learn that performance testing is hard to do. and simple tests don't tell yu much.

Another example. I built application with Postgresql and used a test databse with about 25,000 rows. It was verey fast. BUT when it went into production it died, was dead slow. The reason was that my ENTIRE 25,000 row databse could cache in RAM but a multi-million row production databse would not. It made a huge difference.

So, make sure your test databset is realisticand at least 10x or 100x larger then your RAM cache. Or make two sets of test data and let your students see the difference.

I also found out it is VERY hard to stress apatche if all it is doing is serving static html pages. Even 100 users pounding keyboards could not stress a low end box

Same with samba. The 100BaseT interface is the bottleneck.

Bonnie is good at benchmarking hard drive performance.

You can use Unreal Tournament to test the speed of the computer's components like CPU, memory, video card, and hard drive. Half-Life 2 is better to test these components than Unreal Tournament as you can see at anandtech.com site.

Benchmark programs only gives you estimates that will put you in the right direction towards a goal. Benchmark programs will not show anything special of how the computer works, so I do not know why you think benchmarking a computer will show the components progress in the tests. IMHO, benchmark programs never show me how computer components work.

To really learn computer hardware is by building it. Though making 30+ students to build a computer will really be a mess.

BTW, Chris Albertson, ever heard of a spell checker.