Need Help For some critical issues of MySQL server

[SANAUL HAQUE]

Hi,
I have been trapped in a technical difficulty with MySQL server 4.1.11 Beta since last one week. The difficulty is that "I am trying to use MySQL database server on linux OS {Kernel- 2.6.9-1.667} with the assumption of heavy load (Nearly 300-450 users will hit MySQL Server simultaneously)" I have tested it with my benchmark but there is little difficulty in it. I am creating 50 test databases and 2000 tables in each database. The tables consists of 30 records of 12 fields, In runtime I am also updating records in all created tables. After running my test program I am viewing the memory and CPU usage thru TOP command of linux. At the end of the program I saw that around 700MB RAM of total 2GB Memory has been in use. Then I closes all opened connections.
And finally closed my program also. Then I saw memory and CPU usage again 2 or 3 times, strangely It was showing the same memory usage 700 MB (after closing all connections and program). Then I again did one test "After 4 hours I droped all databases which was created during testing" Then I saw that the memory usage come to the normal state, That means 700 MB memory is than added to the free memory.

My Analysis:
Here my analysis is that after closing the connection the connection object is released and it is not appearing in the NETSTATISTICS or in the SHOW PROCESSLIST but the memory and buffers used by the connection during its transaction period is yet in use, as it is showing in the result of TOP command of linux, that is looking strange to me. Do you have any suggesstion in this regard. I will appreciate this help very much.

Configurations of the Test server---->
HardWare Configuration:
CPU : 3.2 GHZ
Board : Intel 865 Desktop Board
RAM : 2 GB
HardDisc : 70 GB SCSI (7200 RPM) Disc
NetWork Speed: 100 MBPS LAN

MySQL Server Configuration--------->
We are using 90% innodb tables and 10% MyISAM tables. The current configuration of the server is as follows:

[mysqld]
port = 3306
socket = /tmp/mysql.sock
skip-name-resolve
skip-locking
skip-bdb
log_slow_queries = slow_queries.log
long_query_time = 3
key_buffer = 16M
max_allowed_packet = 8M
table_cache = 128
sort_buffer_size = 2M
read_buffer_size = 1M
read_rnd_buffer_size = 2M
myisam_sort_buffer_size = 2M
thread_cache = 100
query_cache_size = 8M
thread_concurrency = 8
max_connections = 500
datadir = /var/data

#innodb configuration, right now all default settings is being in use no tuning has #been done here. Below is not taken from my.cnf, it is the query result of the query "show variables like '%inno%'"

innodb_additional_mem_pool_size 1048576
innodb_autoextend_increment 8
innodb_buffer_pool_awe_mem_mb 0
innodb_buffer_pool_size 8388608
innodb_data_file_path ibdata1:10M:autoextend
innodb_data_home_dir
innodb_fast_shutdown ON
innodb_file_io_threads 4
innodb_file_per_table OFF
innodb_flush_log_at_trx_commit 1
innodb_flush_method
innodb_force_recovery 0
innodb_lock_wait_timeout 50
innodb_locks_unsafe_for_binlog OFF
innodb_log_arch_dir
innodb_log_archive OFF
innodb_log_buffer_size 1048576
innodb_log_file_size 5242880
innodb_log_files_in_group 2
innodb_log_group_home_dir ./
innodb_max_dirty_pages_pct 90
innodb_max_purge_lag 0
innodb_mirrored_log_groups 1
innodb_open_files 300
innodb_table_locks ON
innodb_thread_concurrency 8

Plz also suggest if there is any problem in above Hardware or Server configuration setting for the following usage scenario:
Number of concurrent users: 300-500
Engine type for tables : InnoDB
Number of Fields used in tables : 150
Transaction : Insert is very high, Then comes select <= Insert and finally updates in relatively Low.

Thanks in advance, lookingforward to hear from your side.

Regards,
SANAUL HAQUE,
COMPETENT SOFTWARE PRIVATE LIMITED,
NEW DELHI, INDIA.