Using RAID technology? Alex Young provides some help.
RAID (redundant array of independent disks) technology has enabled users to get much more out of their hard disk drives, namely data protection, fault-tolerance, increased performance and larger capacity. In a RAID array, which can be configured to different levels e.g. 0, 1 or 5, data is shared and/or replicated across multiple disks. Below are the top ten tips existing and potential users of RAID technology should consider for their implementation of the technology:
1. Compare like-for-like: when comparing RAID products don’t focus on the CPU clock speed as that doesn’t always mean faster performance; you should look at the performance figures (MB/sec or IOPS) instead. Only when two RAID products are based on identical RAID ASIC, RAID firmware, and hardware design, does comparing the CPU clock speed might make sense. Each RAID product manufacturer offers different CPU, architecture, firmware, hardware, RAID ASIC, etc, hence the CPU clock speed comparison alone is not a reliable clear performance comparison criterion.
2. Keep it flowing: if you have purchased a 24- or 16-bay RAID subsystems to meet future growth and have not installed disk drives in each bay, install empty trays in the chassis as this will ensure sufficient air flow.
3. Needle in a haystack: if you, like many RAID users, have deployed dozens or even hundreds of RAID subsystems, let the software display a text string that the administrator can identify when searching for a specific unit. For example you can use the IP address, the name of the connecting host computer, or you can even give the RAID subsystem a name or number.
4. Ensure you have enough capacity: before proceeding with a RAID migration project make sure you have sufficient free capacity or unused drives in your RAID subsystem. RAID 6 arrays require at least four member drives and use additional capacity for the distribution of secondary parity. For example if you decide to migrate a three-drive RAID 5 array to RAID 6 you will need one additional disk drive or enough unused space to have a second parity.
5. Protect the cached data: the life span of a battery varies according to the number of recharging / discharging cycles hence you should replace the Battery Backup Module (BBU) after 12 months of operation in order to safeguard the cached data should a mains power failure occur.
6. Check your writes: the Media Scan should be performed regularly. Unless you enabled the Write-Verify function for the normal writes, the disk drives usually do not verify the data when writing. Performing Media Scan can decrease the risk of having multiple data blocks missing, and lowers the risk of data loss. You can have the RAID subsystem perform the Media Scan monthly by using the automatic scheduler function.
7. Reduce latency: when the cache Write-Back is disabled (Write-Through Mode) the entire host IOs are passed directly to the disk drives after RAID operations. All the disk drives will be accessing the data blocks in an order related to the host, and most of the time will be moving the read/write arm and waiting for the data blocks (the so called Latency Time). When the cache Write-Back is enabled, the Write data from the hosts are collected in the cache memory, optimised with the cache algorithms and then flushed to the disks by the RAID controller. The Write-Back cache mode does save a big percentage of disk drive latency time and provides a much better Write performance in most situations, compared to the Write-Through mode (Write-Back disabled).
8. Beware of slow PCI slots: you might be struggling to reproduce the highest performance figure of the RAID subsystem. This might be due to the host computer where the SCSI or Fibre Channel HBAs are usually installed. Often there will be only one or two PCI slots in a computer and while other PCI slots might look the same they might be running at lower speeds. Depending on the computer’s internal design, in many situations multiple PCI devices will have to share the PCI bandwidth and these will all limit the maximum performance that can be performed by the SCSI / Fibre Channel HBA, and affect the performance test results.
9. Plan for growth: when creating RAID Logical Drives, plan to accommodate any future drive capacity variation and be aware that drives from different manufacturers which are supposed to offer the same capacity will actually vary in size. The capacity of the disk drive is measured by the numbers of available data blocks, often labelled on the drive as ‘LBA’ (Logical Block Address, each block is 512 bytes). In a RAID Logical Drive all member drives will be used up to the maximum common denominator so if for example three disk drives in a RAID 5 Logical Drive, have 100, 99 and 101 blocks, only 99 blocks will be used on each disk drive when creating the RAID 5 onto these three disks. This discrepancy means that when a drive has failed and the replacement drive is slightly smaller in capacity the ‘rebuild’ won't start. The solution? When you create the RAID Logical Drive use 1 percent less capacity than the ‘official’ one.
10. Hard copies not just hardware: keep a print copy of the RAID array configurations and connection schemes as in some situations if the complete system is being replaced and the replacement is not the same product (e.g. from different vendors), the original configuration file might not be work. Keeping a hard copy of these details can ensure you quickly have the new unit up and running.
Alex Young is director of technical and marketing EMEA, Infortrend www.infortrend-europe.com
•Date: 8th Dec 2006• Region: World •Type: Article •Topic: IT continuity
Rate this article or make a comment - click here