A Quick Look at Serial ATA (SATA) Disk Performance

From Simple Sci Wiki
Revision as of 15:33, 24 December 2023 by SatoshiNakamoto (talk | contribs) (Created page with "Title: A Quick Look at Serial ATA (SATA) Disk Performance Authors: Tom Barclay, Wyman Chong, Jim Gray Abstract: The authors investigated the use of low-cost, commodity components for multi-terabyte SQL Server databases. They referred to these servers as "storage bricks." One issue they addressed was the wiring mess, air flow problems, length restrictions, and connector failures created by seven or more parallel ATA (PATA) ribbon cables and drives in a tower or 3U rack-...")
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search

Title: A Quick Look at Serial ATA (SATA) Disk Performance

Authors: Tom Barclay, Wyman Chong, Jim Gray

Abstract: The authors investigated the use of low-cost, commodity components for multi-terabyte SQL Server databases. They referred to these servers as "storage bricks." One issue they addressed was the wiring mess, air flow problems, length restrictions, and connector failures created by seven or more parallel ATA (PATA) ribbon cables and drives in a tower or 3U rack-mount chassis. Large capacity Serial ATA (SATA) drives have recently become widely available for the PC environment at a reasonable price. The authors tested two drive brands along with two RAID controllers to evaluate SATA drive performance and reliability. This paper documents their results.

Main Research Question: Can Serial ATA (SATA) drives improve the performance and reliability of storage bricks for multi-terabyte SQL Server databases while reducing costs?

Methodology: The authors built and operated the TerraServer-USA web site, which hosts about 3.3 terabytes of user data. They used classic three-tier application architecture backed by an enterprise class Storage Area Network (SAN). The system was designed for high availability, with features like Microsoft Clustering Service (MSC) and triple-mirrored disks. However, the authors aimed to create a next-generation TerraServer-USA environment that meets or exceeds the performance and availability achieved by the SAN environment at a fraction of the cost.

They used storage bricks, which consist of seven disk 1.75 TB storage bricks with one 2.4 GHz processor, 1 GB of RAM, and 16 disks (2 TB mirrored) to support the load and tolerate disk failures. They used two hyper-threaded 2.4 GHz processors with 4 GB of RAM and 16 disks (2 TB mirrored) in their experiments.

Results: The authors found that SATA drives improved the performance and reliability of the storage bricks. They also found that the SATA connectors seemed more reliable, had a more reasonable length restriction, and allowed better airflow. The testing of two drive brands along with two RAID controllers showed that SATA drive performance and reliability were satisfactory.

Implications: The authors concluded that SATA drives can significantly improve the performance and reliability of storage bricks for multi-terabyte SQL Server databases while reducing costs. They also suggested that their design, which uses three such machines with an addition of three mirrors and a spare server, is highly fault-tolerant and has an easy recovery and management model.

In summary, the authors successfully demonstrated that SATA drives can be used to create cost-effective, high-performance, and highly reliable storage solutions for large-scale databases. Their findings have important implications for the data storage industry and can help organizations optimize their storage systems while minimizing costs.

Link to Article: https://arxiv.org/abs/0403021v1 Authors: arXiv ID: 0403021v1