SQL/MX 2.x Database and Application Migration Guide (G06.23+, H06.04+, J06.03+)

Converting SQL/MP Applications to SQL/MX
Applications
HP NonStop SQL/MX Database and Application Migration Guide—540435-005
10-5
Parallel Operations for Scalable High Performance
Parallel Operations for Scalable High Performance
NonStop SQL/MX can perform some queries in parallel across multiple disk volumes
without the use of executor server processes (ESPs). NonStop SQL/MP always
requires ESPs for parallel queries.
NonStop SQL/MX can execute large queries that scan large portions of the database in
parallel. NonStop SQL/MX supports three types of parallelism: pipelined parallelism,
independent parallelism, and partitioned parallelism. Pipelined and independent
parallelism are inherent features of NonStop SQL/MX, and you cannot force plans to
use them.
Partitioned parallelism is the ability to divide the data to be processed into partitions
and work on each partition in parallel. In a partitioned parallel plan, multiple operators
all work on the same plans, and results are merged by using multiple pipelines.
Although NonStop SQL/MP uses some partitioned parallelism to process in parallel,
NonStop SQL/MX takes advantage of both ESP parallelism and Data Access Manager
parallelism.
For more information about parallelism, see the
SQL/MX Query Guide.
Reasons for Keeping SQL/MP Applications
In some cases, it might be preferable to keep SQL/MP applications instead of
converting them to SQL/MX applications.
Large Number of Indexes
If you use a large number of indexes in your SQL/MP database, reconsider converting
your SQL/MP applications to NonStop SQL/MX. A large number of indexes in your
database can result in longer SQL compile times for SQL/MX applications. In NonStop
SQL/MX, an excessive number of indexes increases the search space for the optimal
plan.
Replicated Indexes on Multiple Nodes
Your SQL/MP database might use replicated indexes on multiple nodes so that a query
on a node can use the local index for access instead of a remote index. Be aware that
NonStop SQL/MX currently does not recognize this arrangement and might use a
remote index instead of a local index.
Single-Row Queries
Currently, single-row access has a smaller path length in NonStop SQL/MP than in
NonStop SQL/MX. If your SQL/MP application performs many single-row queries, it
might be preferable to defer converting your application to NonStop SQL/MX. As
NonStop SQL/MX matures as a product, the set of queries that perform better in
NonStop SQL/MX are expected to grow.