 |
|
Oracle Tips by Burleson |
A Review
of Existing SSD Research Findings
Different researchers are coming to different
conclusions about the applicability of SSD to Oracle systems. There
are three research papers on SSD and each arrive at similar
conclusions about the use of SSD with Oracle. Complete references are
included in the References section at the end of this section. Here
is a summary of the findings from each study.
James Morle
According to (Morle, 2002), SSD is great for
Oracle redo logs, undo tablespace (rollback segment tablespace in
Oracle8i), and the TEMP tablespace. He notes that for rollback
segments, SSD is a great help:
“This is where SSD
can help out. By deploying a single SSD, all redo logs can be
located away from the RAID 1+0 array, whilst providing low latency
writes and high bandwidth reads (for archiving).”
Morle also asserts that full-caching of a database
on SSD may not improve performance:
“If the whole
database were running from SSD, there would be enormous pieces of
unnecessary work going on, such as:
Management of the
buffer cache
Context switches
into kernel mode to perform I/O
Conversion of the
request into SCSI/Fibre Channel
Transmission across
the SAN
And all the way
back again
In comparison to
disk I/O, this whole process is stunningly fast. In comparison to
just reading the data straight from user space memory, however, it
is incredibly slow!“
Morle notes that a typical OLTP system maintains a
working set of frequently referenced data blocks, and those might be
good candidates for SSD. For DSS and Data Warehouse systems, Morle
advocates moving the current table partitions onto SSD devices,
leaving the others on traditional disk.
Dr. Paul Dorsey
In another landmark SSD study in 2004, Dr, Paul
Dorsey showed that the SSD data transfer rates are always better than
traditional disk:
|
Device |
Test#1:
Buffered Read |
Test #2:
Sequential Read |
Test #3:
Random Read |
Test #4:
Buffered Write |
Test #5:
Sequential Write |
Test #6:
Random Write |
|
RamSan |
95 |
98 |
98 |
86 |
84 |
82 |
|
IDE |
85 |
40 |
6 |
65 |
38 |
11 |
|
SCSI |
65 |
33 |
9 |
49 |
33 |
11 |
Dr. Dorsey concludes:
“Technologically,
SSD is one of the best sources of performance improvement for an
Oracle database if you have a typical OLTP system including many
transactions which access different small amounts of random data and
lots of users.
SSDs may also
improve data warehouse applications because of the improved query
performance. There is no generic answer for all questions, but solid
state disks represent another way of thinking about managing
enterprise-wide databases.“
Woody Hutsell
In his Texas Memory Systems whitepaper titled
Faster Oracle Database Access with
the RAMSAN-210 (2001), Hutsell concludes that certain types of
Oracle databases will always benefit from SSD:
There are some
databases that should have all of their files moved to solid state
disk. These databases tend to have at least one of the following
characteristics:
High concurrent
access. Databases that are being hit by a large number of
concurrent users should consider storing all of their data on solid
state disk. This will make sure that storage is not a bottleneck for
the application and maximize the utilization of servers and
networks. I/O wait time will be minimized and servers and bandwidth
will be fully utilized.
Frequent random
accesses to all tables. For some databases, it is impossible to
identify a subset of files that are frequently accessed. Many times
these databases are effectively large indices.
Small to medium
size databases. Given the fixed costs associated with buying
RAID systems, it is often economical to buy a solid state disk to
store small to medium sized databases. A RamSan-210, for example,
can provide 32GB of database storage for the price of some
enterprise RAID systems.
The above book excerpt is from:
Oracle
Solid State Disk Tuning
High Performance Oracle
tuning with RAM disk
ISBN
0-9744486-5-6
Donald K. Burleson & Mike Ault
http://www.rampant-books.com/book_2005_1_ssd.htm
|
