eG Monitoring
 

Measures reported by SybDeviceIOTest

This test monitors the I/O activity on each device on the Sybase server, and proactively alerts administrators to contention for device I/O semaphores. To ensure that this test runs smoothly, enable the enable monitoring configuration parameter on the Sybase server.

 The measures made by this test are as follows:

Measurement Description Measurement Unit Interpretation
Reads Indicates the number of reads that occurred on the device during the last measurement period. number  
APF_reads Indicates the number of APF reads that occurred on this device during the last measurement period. number Asynchronous prefetch (APF) issues I/O requests for pages before the query needs them so that most pages are in cache by the time query processing needs to access the page. High APF activity can hence improve performance for:
  • Sequential scans, such as table scans, clustered index scans, and covered nonclustered index scans
  • Access via nonclustered indexes
  • Some dbcc checks and update statistics
  • Recovery
Writes Indicates the number of writes that occurred on this device during the last measurement period. number  
Request Indicates the number of requests from this device during the last measurement period. number  
IO_waits Indicates the number of I/O requests to this device that were waiting during the last measurement period. number  
Tot_io_time Indicates the time taken to read from or write to this device during the last measurement period. msecs A high value of this measure indicates that the device is taking too much time to process I/O requests. This could be owing to any of the following reasons:
  • An I/O overload on the device;
  • Disk fragmentation on the device;
  • Table fragmentation
Pct_io_granted Indicates the percentage of IO granted. Percent Ideally, this value should be high. When Adaptive Server needs to perform a disk I/O, it gives the task to the semaphore for that device in order to acquire a block I/O structure. On SMP systems, multiple engines can try to post I/Os to the same device simultaneously. This creates contention for that semaphore, especially if there are hot devices or if the data is not well distributed across devices.

A low value for this measure indicates that many I/O requests were waiting for the semaphore - in other words, it indicates a contention for the semaphore. One solution might be to redistribute the data on the physical devices