Compaq Backup and Recovery for Microsoft SQL Server 6.x
initialized and never used pages only, and yields very high compression ratios and very high backup
20
throughput.
The "Repeated Patterns" column represents a data set with a high percentage of repeated
data and unused space, thus relatively high compression ratios and high performance. The "Random
Patterns" column represents a data set with a very low to zero percentage of repeated data and unused
space, thus low or no data compression. The chart should give you an idea how the compression ratios
and backup performance vary, depending on the nature of data.
From our results we can conclude the following:
1.
Hardware data compression is typically more effective than software data compression, when both
are operating on the same data set. This is an advantage for sites that want to conserve backup
storage, and are not pressed to perform backups within the smallest time window possible.
2.
Software data compression can sometimes be faster than hardware data compression, since it can
utilize a more powerful and faster system processor (depends on the processor type, of course), and
because it actually sends less data to the tape drive controller. Performance of software data
compression is an advantage for sites that have a need to back up as fast as possible, with less
concern over the backup storage consumed. Software compression by the application is only
applicable to off-line backups, since it is not provided by SQL Server itself.
3.
It is very important, with both hardware and software data compression, to keep the tape drive
21
streaming
streaming, you may be able to achieve better performance without using any data compression.
Here is why. Under no data compression, data is written to the tape "as is." Since no compression
occurs, enough data is delivered to the tape drive to keep it continuously busy. With data
compression, data is analyzed, compressed and then only a fraction of the original amount of data
is actually being written. If the compressed data stream stops due to the host not keeping up, the
20
The case of 'unallocated space' showing increased compression does not apply to a SQL Server based (online)
backups.
21
Streaming means that the system can and does deliver enough information to the tape drive to keep it
constantly writing, as opposed to the tape drive having to stop and wait for new information.
Chart 1 - Data Compression Ratios and Performance
8QDOORFDWHG
8QDOORFDWHG
5HSHDWHG
6SDFH
6SDFH
3 DWWHUQV
6:'&
+:'&
6:'&
at all times to achieve optimal performance. If the system cannot keep the tape drive
5HSHDWHG
5DQGRP
5DQGRP
3 DWWHUQV
3 DWWHUQV
3 DWWHUQV
+:'&
6:'&
+:'&
Page 18
%DFNXS7KURXJKSXW
&RPSUHVVLRQ
5DWLR