For database environment problems arise in the use of the database, it may be caused by any of the components of which a table space, file system, data files, processes, etc.,
There may even be due to a particular SQL performance statements cause poor system performance issues. Therefore, when the root causes of database problems, a thorough investigation of the problem to be repeated, complicated labor,
But before troubleshooting, performance testing as a technician should first understand what tools, which monitor database performance indicators clear, targeted, no gain, as oracle monitor performance indicators and the use of monitoring and analysis methods:
Index Name |
Problem Description and Diagnostic Analysis |
|
Oracle Data Access Data way |
||
Full table scan / sec (Full Scans/sec) |
Indicators Description |
It refers to the number of full table scans per second. Full table scan can be substantially full table scan or index scan. Because full table scan takes a lot of time, so a full table scan frequency too high CPU utilization is too high, it will affect performance. |
Diagnosing |
1, obtained by using Spotlight conduct SQL problem. 2, analysis get the object in question through SQL database logs |
|
indicator anaysis |
1 , by using Spotlight or Profiler analyzes to see whether you need a full table scan by the way, to optimize the SQL query performance. 2 , there are a large number of cases for process historical data look-up table data, you can improve query performance by analyzing whether the partition table. |
|
other |
Indicators Description |
SQL syntax wording problems, leading to excessive consumption of resources CPU / memory |
Diagnosing |
1, obtained by using Spotlight conduct SQL problem. 2, analysis get the object in question through SQL database logs |
|
indicator anaysis |
Too many nested sub-query or sort grammar and so on, by modifying the SQL implementation |
|
ORACLE database instance parameters Analysis |
||
Buffer Waits Buffer wait |
Indicators Description |
It refers to the probability of obtaining waiting in the buffer Buffer |
Diagnosing |
Get metrics corresponding values or by using a snapshot Spotlight |
|
indicator anaysis |
Spotlight by using the performance index to be close to 100%; |
|
Buffer hit ratio Buffer Hit % |
Indicators Description |
It refers to the data block in the data buffer hit ratio of |
Diagnosing |
Get metrics corresponding values or by using a snapshot Spotlight |
|
indicator anaysis |
This value should generally be 90% or more; |
|
Shared area hit rate SharedPool % |
Indicators Description |
Shared use of pool |
Diagnosing |
Get metrics corresponding values or by using a snapshot Spotlight |
|
indicator anaysis |
By using Spotlight analysis, this value should be generally 90% or more; |
|
Sort of memory usage |
Indicators Description |
It refers to the ratio of the sort operation in memory. When a query requires a sort, and now sort of memory, lack of memory will use the temporary table space disk sorting, causing IO problem, affecting efficiency. |
Diagnosing |
Get metrics corresponding values or by using a snapshot Spotlight |
|
indicator anaysis |
This value is typically close to 100% better performance; |
|
Log File Synchronization Log file sync |
Indicators Description |
This wait event means that when a session is complete a thing, LGWR process must wait until the session's redo information is written to the log file from the log buffer to continue execution. |
Diagnosing |
Get metrics corresponding values or by using a snapshot Spotlight |
|
indicator anaysis |
This time the event is too long to wait, probably due to submit or update data once too often or too long write time logs; |
|
Redo Buffer Waits rate Redo wait |
Indicators Description |
Waiting for the chance to acquire an in Redo buffer Buffer |
Diagnosing |
Get metrics corresponding values or by using a snapshot Spotlight |
|
indicator anaysis |
The performance indicators should be close to 100%; |
|
oracle locks (Locks) performance counter |
||
Queue waiting (enqueue(cs)) |
Indicators Description |
The locking mechanisms to protect shared resources, such as data records in order to avoid two personal update the same data at the same time. enqueue comprising a queuing mechanism, i.e., FIFO ( first in first out ) queuing. |
Diagnosing |
By using Spotlight, snapshots, LR monitor to obtain the corresponding index value |
|
indicator anaysis |
如果enqueue等待事件比较显著,则需要根据enqueue等待类型,采取相应的优化方法。 防止多个事务并发使用资源,尽可能少使用锁来提高性能开销。 |