 |
|
Oracle Tips by Burleson |
Oracle
RAC Cache Fusion and I/O Bandwidth
Oracle RAC Cache Fusion uses a high-speed IPC
interconnect to provide cache-to-cache trasfers of data blocks between
instances in a cluster. This is called data block shipping. This
eliminates the disk I/O and optimizes read/write concurrency. Block
reads take advantage of the speed of IPC and an interconnecting
network.
The cache-to-cache data transfer is performed
through the high speed IPC interconnect
. The Oracle Global Cache Service(GCS) tracks blocks
that were shipped to other instances by retaining block copies in
memory. Each such copy is called a past image (PI). The GCS, via the
LMSx background process, tracks one
or more past image versions (PI) for a block in addition to the
traditional GCS resource roles and modes. In the event of a node
failure, Oracle can reconstruct the current version of a block by
using a saved PI.
Simultaneous Reads on
Different RAC Nodes
Simultaneous reads of the same data block on
multiple nodes cause bandwidth bottlenecks, but they can also cause
cache conflicts within RAC. Real Application Clustersresolves this situation because multiple instances share
the same blocks for read access without cache conflicts. Conflicts can
only occur for read/write and write/write situations.
Simultaneous Reads and
Writes on Different RAC Nodes
Simultaneous reads and writes on different nodes
are the dominant form of concurrency problems in Online Transaction
Processing (OLTP), hybrid OLTP, and
data warehouse applications. A read of a data block that has been
modified recently can be for either the current version of the block
or for a read-consistent previous version. In both cases, the block
will be transferred from one cache to the other via the high speed
interconnect.
Simultaneous Writes on
Different RAC Nodes
When one or more users need to modify the same
block, this can lead to simultaneous writes being triggered on
multiple nodes. These simultaneous writes lead to contention and
require lock messaging and conversions. By using proper blocks sizes
(large blocks aggravate the problem, small ones reduce it) this issue
can be mitigated. When using SSD with high bandwidth the duration of
transactions, and thus the incidence of simultaneous writes are
reduced.
Now that information has been presented on how SSD
removes the bandwidth bottleneck for high access files, attention can
now be turned to how to locate high concurrent access data files in an
Oracle RAC database.
The above book excerpt is from:
Oracle RAC & Tuning with
Solid State Disk
Expert Secrets for High
Performance Clustered Grid Computing
ISBN
0-9761573-5-7
Donald K. Burleson & Mike Ault
http://www.rampant-books.com/book_2005_2_rac_ssd_tuning.htm
|
