10 Informix Dynamic Server 11: Advanced Functionality for Modern Business
specific columnar values in addition to the partitioning scheme applied to the rest
of the table. Simple LOBs can and need to be fragmented into simple
blobspaces. However, because they are black-box objects as far as the instance
is concerned, no further fragmentation options are possible.
Indexes can also be partitioned using an
expression-based partitioning scheme.
A table’s index partitioning scheme does not need to be the same as that used
for the associated table. Partitioned indexes can be placed on a different physical
disk than the data, resulting in optimum parallel processing performance.
Partitioning tables and indexes improve the performance of data-loading and
With expression-based partitioning, the IDS cost-based SQL optimizer can
create more efficient and quicker plans using partition elimination to only access
those table and index partitions where the data is known to reside or must be
placed. The benefit is that multiple operations can execute simultaneously on the
same table, each in its unique partition, resulting in greater system performance
than typical data servers provide.
Depending on the operating system used, IDS can use
raw disks when creating
dbspaces to store table or index data. When raw disk space is used, IDS uses its
own data storage system to allocate contiguous disk pages. Contiguous disk
pages reduce latency from spindle arm movement to find the next data element.
They also allow IDS to use direct memory access when writing data. On
Windows-based platforms, this access is not possible so standard file system
files are used.
With IDS 11, if the Linux or UNIX operating system permits, the data server
bypasses the system’s I/O buffers and writes directly to
cooked or operating
system flat files that support chunks and dbspaces. With this feature enabled,
you can approach the same I/O performance levels as with raw space. Given
that all operating systems (OSs) fragment files on disk based on block size, the
data pages within a chunk are not contiguous as with raw space. For many
clients, that is an acceptable trade for the ability to manage IDS storage space
with normal OS disk tools.
1.1.4 Leveraging the strengths of DSA
With an architecture as robust and efficient as IDS, the engine provides a
number of performance features that other engines cannot match.
The High Performance Loader
The High Performance Loader (HPL) utility can load data very quickly, because it
can read from multiple data sources (such as tapes, disk files, pipes, or other