Content area
Full Text
This article details a performance study of Wide Ultra (WU) SCSI and SSA (Serial Storage Architecture) architectures. The main objective of this article is to address the architectural performance of WU SCSI and SSA interfaces. A secondary objective is to identify numerous industry applications and how these applications store and retrieve information using these interfaces. Results and conclusions are drawn from real lab experiments and they are intended to inform and educate the reader on the architectural performance of both interfaces. (Architectural Performance: the ability to analyze the performance of IO (Input/Output) interface architectures by eliminating differences in as many system components as possible, e.g. the host, host adapter, storage devices, etc.)
This article contains one method of benchmarking performance. The performance benchmark tool used in this study was developed by me and Solution Technology's engineering staff. These benchmarks have taken into consideration industry experience gained by working with performance engineers from major computer and peripheral manufacturers that are involved in such undertakings.
Actual Performance Versus Modeling
Although many papers have been written based on modeling techniques, this article contains actual measured performance using real hardware. In order to obtain actual architectural performance data, the architectures must be measured as identically as possible by using the same host computer and operating system, disk drives mechanics (same HDA, hard disk assembly), measurement tools and host adapters.
Typically, the objectives above are impossible to achieve unless one vendor makes all of the components. However you can come close by using the same setup and components with major similarities. For instance, in this analysis the following were used:
200Mhz Pentium based File Server with Windows NT 4.0 Operating System; IBM 9.1GB UltraStar WU SCSI and IBM 9.1GB UltraStar SSA disk drives;
Solution Technology's Extreme SCSI Performance benchmarking tool was used;
Host adapters from Adaptec and IBM. The Adaptec 3940 Dual Channel WU SCSI adapter was chosen because its architecture closely resembles the IBM SSA 6209 Dual Channel host adapter.
Keep in mind that this article's objective is to determine the performance of the different architectures and not the performance of the adapters and disk drives. Therefore, IBM disk drives were chosen because IBM is the only manufacturer that makes both SSA and WU SCSI disk drives. The disk...