Databases Reference
In-Depth Information
Table 14-2.
Common Exadata Statistics
Statistic
Meaning
cell physical I/O interconnect bytes
Number of bytes transmitted over the InfiniBand interconnect
cell physical I/O bytes eligible for predicate
offload
Number of bytes eligible for predicate offload, an indication of
smart scan
cell physical I/O interconnect bytes returned
by smart scan
Numbers of bytes transmitted over the storage interconnect as a
result of smart scan
cell physical I/O bytes saved by storage index
Number of bytes saved by a storage index; this is a reflection of how
many physical disk I/O bytes (and by proxy, requests) were saved
due to a storage index eliminating candidate blocks
cell flash cache read hits
Number of read requests satisfied in Smart Flash Cache
■
this recipe is noticeably void of scripts and examples because in Chapters 15 through 19, we will
demonstrate using exadata statistics and counters to measure specific characteristics of exadata performance in
Chapters 15 through 19.
Note
How It Works
Monitoring with statistics and counters is a performance methodology with a long history. Combined with the
database time and wait event performance methodologies, the counter-based performance monitoring strategy
rounds out the performance engineer's monitoring tool set.
When painted with a broad stroke, monitoring with statistics and counters typically has limited benefits for the
performance engineer because statistics alone often have no bearing on what an application or user experiences.
Users are concerned with how long it takes to complete an operation or run a program and when performance is
unacceptable, the questions that need to be answered are the following:
•
Where is the program or session spending its time?
•
What resources is the program or session waiting on, if any, and can these bottlenecks be
addressed?
•
Is the program or session issuing SQL statements to the database efficiently?
As is evident for most experienced Oracle DBAs and performance engineers, raw, counter-based statistics do not
answer these questions beyond a hypothetical or correlative level.
Exadata statistics and counters, however, do provide some value to the Exadata DMA. With Exadata, you
presumably wish to unleash the power of the software that runs on the storage cells. The cell offload statistics can
present a nice means to measure the effectiveness of the storage server software and help determine whether your
workload, programs, sessions, or system is benefiting from Exadata's features.
In Chapters 15 through 19, we will demonstrate statistics-based techniques to measure the effectiveness and use
of these Exadata software features.
Search WWH ::
Custom Search