Many disk vendors allow multiple disks to be daisy-chained to one controller. Thus, although you might have six physical disks, there might be only one or two disk controllers. That won’t buy you much. Ideally, there should be a one-to-one correspondence between a disk and a controller, unless the controller throughput is substantially higher that the throughput of single disk, in which case the controller throughput ought to match the sum of the maximum potential throughput for all disks. However, as the databases get bigger and bigger and the number of disks grows exponentially, and organization will find it difficult to invest in an equivalent number of controllers. Accordingly, you should understand how many controllers are present and how many disks are connected with each controller. It helps to be familiar with the conventions that vendors use to indicate disk and controller relationships. Usually, once the configuration has been set up, certain disks map to certain controller only. So prior to that actual setup, any disk may potentially be accessed by any controller. After setup, however, once a disk slice has been created, it may be accessed by only one controller. So if controller is busy working and other controllers are free waiting for an I/O request, they still cannot speed things up by accessing disk slice. Accordingly, tasks for controller will have to be serialized. Keep such facts in mind while distributing data across different disks. Find out what kind of channel the vendor has used between the host computer and the external disks. Also be familiar with the kinds of controllers being provided by the vendor. Are they based upon SCSI standards or something else? Know what standards are available in the market and which of these standards have the support of mainstream vendors. If you are still in the vendor selection stage, select the one who offers the fastest and the best medium, if price is restraining factors then analyze the price versus performance and availability ratio. This area is really critical; you will spend a lot of time in the future tuning I/O and related aspects. So a little bit of homework toward this end will not hurt. If your organization has already purchased the hardware, any time spent in familiarizing yourself with the components will be time well spent. This is very important thing to understand in dba services. It is very important to check your system contains how many disk controllers. How those disk controllers are connecting with disks. This information is very useful to understand logic of disk I/O and performance bottleneck of physical I/O of Oracle server for tuning prospective. So many times, disk I/O bottleneck found at Oracle database server but we never get actual root cause of performance degradation. Oracle disk I/O tuning is very critical and vital for ever Oracle server. If we configure Oracle server system accurately with keeping all requirement and future issues in mind during deployment period then it would be easy to analyze any disk I/O bottleneck well in advance. Oracle server would be running smoothly with high performance and availability. Author is providing DBA Services and using strong database administration .
Related Articles -
Oracle tuning, oracle performance tuning, oracle disk io, disk io tuning, oracle performance,
|