This article can also be found in the Premium Editorial Download "Storage magazine: How to scale up with storage clusters."
Download it now to read this article plus other related content.
|SATA pros and cons|
strategies. SATA has evolved significantly from parallel ATA, but there are a number of issues that must be resolved before the drives can reliably support enterprise applications.
SATA drives deliver storage capacities equivalent to FC/SCSI drives, but for as little as a quarter of the price per gigabyte. SATA drives are also available with much larger capacities than FC or SCSI disks. The major question is whether SATA drives can satisfy enterprise-class availability, data integrity and performance requirements.
SATA evolved from the parallel ATA interface, which has been used mainly in desktop and entry-level server systems. However, the performance and reliability characteristics of ATA drives were simply not engineered for enterprise storage. The initial serial version of ATA (SATA) was developed to overcome some of the limitations of parallel ATA drives. SATA 1.0 delivered improvements in performance (1.5Gb/sec dedicated bandwidth), cabling and reliability (hot plug/ swap). SATA 1.0 drives support speeds up to 10,000 rpm and mean time between failure (MTBF) levels up to 1 million hours under an eight-hour, low-duty cycle. FC drives support up to 15,000 rpm and an MTBF of 1.4 million hours under a 24-hour duty cycle. But SATA's impressive cost advantage over FC has compelled many companies to deploy SATA in secondary and primary storage platforms, even in places where SATA may not yet be appropriate.
Four trends are driving the need for low-cost, high-capacity disk drives:
- Tiered storage
- Shift to disk-based backup
- Regulatory compliance
- Small- and medium-sized business (SMB) requirements
Enterprises are abandoning a one-size-fits-all approach to storage and moving toward tiered storage, where data is migrated to lower classes of storage as its business value depreciates. Many companies have introduced disk into their backup environments in an effort to hit tight backup windows and stringent recovery time/point objectives. Because the ratio of primary to secondary (backup) data in many environments is 1:10, the introduction of disk presents a number of economic challenges requiring a low-cost, high-capacity disk solution.
New regulatory compliance requirements mandate that businesses preserve electronic records for many years. Moreover, some of these regulations define specific recovery time objectives for data, which often means the data must be retained online. The explosive growth of online archive data sets, often into petabytes, is another driver for less-expensive disk drives.
Single point of failure
Enterprise-class data requires two independent, end-to-end data paths from the client application to the back-end disk drives. A single point of failure along the data path will compromise the entire storage architecture. While FC-based configurations ensure high availability with dual-ported drives and pairs of redundant FC loops with access to all of the drives, SATA provides only a single-ported drive.
In block-based storage, user data is stored in 512-byte blocks with disk drives typically formatted to fit one block into each drive sector. For additional data integrity, block size can be expanded to 520 bytes, with the additional 8 bytes used to protect the data in the 512-byte block. Some FC apps require drives to be formatted with a media sector size of 520 bytes. SATA drives can only be formatted with a media sector size of 512 bytes.
To overcome deficiencies in Port Bypass Circuits--rudimentary components that enable FC Arbitrated Loop (FC-AL) traffic to bypass a failed drive and continue operations--FC-AL switch devices have been enhanced with component-level diagnostics that can effectively diagnose and isolate the activity of individual FC drives. SATA-based architectures require similar or higher levels of diagnostics, particularly given their lower MTBF characteristics; SATA on its own lacks this capability.
This was first published in April 2005