Timesten 学习笔记（四）环境变量设置

Shared library path environment variable

On Linux systems, add install_dir /lib directory to the LD_LIBRARY_PATH environment variable.

 

Set the CLASSPATH variable

set CLASSPATH=.;install_dir
/lib/ttjdbc6.jar

 

ODBCINI SYSODBCINI environment variable

TimesTen 使用 odbcini 文件确定 数据库资源和连接属性。 如果用户自己不指定odbcini，系统自动徐兆 安装目录下的/info/ sys.odbc.ini 。

 

具体内容如下

 

[my_ttdb]
Driver=/home/tt1/TimesTen/tt1121/lib/libtten.so
DataStore=/home/tt1/TimesTen/tt1121/info/DemoDataStore/my_ttdb
PermSize=20960  /*The ttMigrate and ttDestroy utilities, and can only expend can not desize*/
TempSize=3200 /* Tempsize = 40 MB + ceiling(PermSize / 8 MB) */
PLSQL_MEMORY_SIZE=64
PLSQL=1
DatabaseCharacterSet=WE8MSWIN1252
OracleNetServiceName=LNPDUMP
LogBufMB=500   /*MB*/
LogFileSize=512
#MemoryLock=4
LogDir=/home/tt1/TimesTen/tt1121/log/my_ttdb /*place the transaction log files on a different I/O

                                                                                   path from the database checkpoint files. */
CkptFrequency=300          /*seconds*/
CkptLogVolume=0
CkptRate=10     /*MB*/
CachegridEnable=0

 

有个connection需要特别指出。系统默认是64 个， 最大2047。 下面是一段对于 concurrency的描述

 

In a shared environment, concurrent connections to the database become a key limitation: A separate Oracle process is required to handle each connection in an Oracle database, and each process may require up to 2 MB of memory. All database installations--even Dedicated installations of Oracle and others--limit the number of concurrent connections based on the upper limit of memory and CPU available.

Example: Consider a shared Oracle server with 30 customer accounts at 100 concurrent connections per account. If each customer averaged only 30% of their maximum--or 33 connections per customer--the total number of concurrent connections would average 990. At 2 MB per process, the machine would be using 1980 MB RAM. This does not include shared memory for database buffering, or any memory for core OS/DB processes. In addition, at 990 concurrent connections, a great deal of processor time is burned just handling context switches. Under these circumstances, the server would slow to a crawl very quickly.

Limiting concurrent connections enables us to dedicate more memory to each individual process to dramatically increase the performance of your queries. Because of these precautions, complex transactions run much faster than with other databases. More dedicated memory is available to each process, and a high amount of system memory is dedicated to buffering.