APC RAID Subsystem SCSI-SATA II Technical Information Page 167

  • Download
  • Add to my manuals
  • Print
  • Page
    / 259
  • Table of contents
  • TROUBLESHOOTING
  • BOOKMARKS
  • Rated. / 5. Based on customer reviews
Page view 166
Troubleshooting
6-3
9. Add memory at host computer or RAID controller
Adding more memory to your motherboard or RAID controller helps to cache more data at memory to so as to reduce the
number I/O access to hard disks, especially helpful for data being accessed frequently. Bigger memory also helps to
avoid the performance glitch because more data can be buffered for write commands or pre-fetched for read commands,
especially helpful for multiple video streams.
10. Check data layout on hard disks
A hard disk can deliver its best performance when servicing I/O access on the inner tracks of disks, which provides data
of lower block address (LBA). As a result, to deliver high-throughput performance, place the data in the beginning area of
disk groups. To retain the performance of data at the second half area of disk groups, you may use more hard disks with
striping.
11. Make the I/O workload distribute evenly
Check below to ensure the I/O workload are distributed evenly
• Data is transferred evenly through multiple host-interface connections
• MPIO is enabled and dynamic load balancing is turned on
• I/O are processed evenly by the two controllers
• I/O are distributed evenly to multiple disk groups
12. Close the web GUI during I/O access
Sometimes, the web GUI or RAID monitoring software could be an influential factor to the performance unstableness,
because it needs to retrieve RAID system status periodically, and consumes CPU cycles of the storage processor. Close
the web GUI when you run time-sensitive applications.
13. Reduce the impact of background task
The background I/O tasks, like RAID initialization or rebuilding, have impact to the performance of your applications,
because they need to access hard disks. Even SMART monitoring could cause disturbance. You may set the priority of
background tasks to low, or schedule these tasks to run at non-business hours to avoid the impact.
14. Constantly monitor the I/O statistics
To know details of the performance of your RAID system, you may check the performance management web pages, by
which you may identify slow hard disks, slow host interface links, or unbalanced workload distribution (see 2.11
Performance Management on page 2-54). Some operating systems offer similar utilities. For example, Microsoft
Windows Performance Monitor not only displays many useful statistics but also can be configured to send out alerts
according to the threshold you set. The statistics can also be saved as a log file. You may find similar utilities from your
HBA and switch vendors.
15. Know more about performance benchmarking
And finally, you have to understand that the test result of performance benchmark tool is not always related to your real-
world performance. You have to be careful with choose right tools and right testing workload profiles to mimic your
application behaviors.
6.4 Hard Disks
Hard disks are the most important components in a RAID system because they are where the data resides. Please
contact your RAID system supplier to get the list of qualified hard disk models when you’re choosing hard disks.
1. Hard disks cannot be recognized by the RAID controller
The hard disks are initialized by the RAID controller when the controller boots up or when the hard disks are plugged into
the RAID system. If a hard disk cannot be ready within a specific period of time during the initialization, the RAID
controller will force the hard disks enter faulty state and you cannot see any model information of hard disks. To ensure
hard disks have enough time for power-on, you may extend the delay time when boot-up of the RAID controller (see 2.8.1
Hard disks on page 2-38).
2. Hard disks are offline unexpected
The RAID controller takes a hard disk offline when the hard disk cannot respond to the RAID controller after the full-cycle
error recovery procedure has been done. This could happen when the hard disk is permanently dead because of its
internal component failure, and you lose all your data on the hard disk. Another reason a hard disk is taken offline is that
its reserved space for meta-data (RAID configurations) cannot be written, which means the reserved space for bad block
reallocation in the hard disk has been full. This is unlikely to happen because two copies of meta-data are reserved, and
a hard disk is offline only when both areas cannot be accessed.
An offline hard disk might also be transiently down because of its disk controller firmware lockdown or mechanical
unstableness. In this case, the hard disk is still accessible and you may reuse it, but the hard disk might fail again.
Page view 166
1 2 ... 162 163 164 165 166 167 168 169 170 171 172 ... 258 259

Comments to this Manuals

No comments