SQL/MP Installation and Management Guide

Managing a Distributed Database

HP NonStop SQL/MP Installation and Management Guide—523353-004

12-11

Using Remote Servers

When you use a network-distributed database, you can often control whether remote

data is updated directly by a local server or indirectly by a remote server. Any local

program can update or retrieve data directly by using the remote I/O capabilities of the

Guardian file system and disk process.

Alternatively, when you need to update data stored at a remote node, you can send a

message containing an update request to a server at that remote node. Ultimately, this

issue might be one of performance and processing power distribution across nodes.

One of the main advantages of using a remote server for distributed processing is to

reduce the amount of data sent across communication lines. One message makes the

request of the server at the remote node. Then, that server manages all access to, and

updating of, the remote data. This approach reduces message traffic on the slower

communication lines and increases performance.

Managing Processor Usage in a Distributed

Environment

For a query that executes in parallel in a distributed system or network, you can

choose a specified set of processors in which the query will run. The remaining

processors are free for other tasks—for example, executing a different type of query for

another application.

You control processor usage by using the _SQL_CMP_CPUS DEFINE. Before

compiling a query, you add this DEFINE to select a set of “usable” processors. The

optimizer chooses an access plan that uses only the allowable processors for executor

server processes (ESPs) and for temporary files chosen for repartitioning.

The DEFINE influences all parallel execution plans compiled while the DEFINE is in

effect. You can reset the DEFINE to change the usable processors before compiling

other queries, so different queries can have different “usable” processors.

By determining that certain sets of queries (or applications) run on certain processors,

you can improve the performance and manageability of a distributed system or

network—especially one used for multiple purposes. You can select usable processors

in a single node or across multiple nodes.