Product specifications

• client proxies in a 2-tier deployment, on the client tier
• server bridges in a web tier deployment, on the web tier
Note:
Connection Server also provides an application-level lookup mechanism, which helps it to choose the
right server instance for a job. Connection Server binds the client to the appropriate server, depending
on the type of data source requested by the client and on what network layer and database pair the
server supports.
Remember:
Load balancing is performed after the lookup. It is available on all platforms.
4.5.1 Finding the best server
Connection Server scalability issues may come from the following:
• The number of opened database connections
Issues dealing with the number of database connections are already handled by the connection pool
mechanism.
• The CPU used by Connection Server
Connection Server may use a noticeable amount of CPU when converting data or when writing and
sending back big response buffers to the client. However, Connection Server is not an intensive
CPU consumer.
• The number of jobs running
The number of jobs that run on a server increases its workload.
• The amount of memory allocated by either Connection Server or the middleware it embeds
The amount of memory used depends on the SQL queries executed. For example, a single SQL
statement against a large table may have a bigger impact on scalability than a series of small SQL
queries.
Connection Server implements load balancing with the help of the most relevant indicator of server
workload, which is in this case the amount of allocated memory. The less memory a Connection Server
server uses, the healthier it is.
Note:
Connection Server may also use the number of jobs as health indicator in future releases.
The following formula calculates the HEALTH of the server:
HEALTH = (available memory / max memory) * constant
where:
• max memory is the maximum amount of allocatable memory, for example 2GB on a 32-bit MS
Windows platform
2012-03-1533
Learning data access specifics