Storage Technology News:
Understanding disk storage components
By Stephen J. Bigelow, Features Writer
24 Oct 2007 | SearchStorage.com
Disk storage components
Hard drives are offering more capacity and faster interface speeds, yet are using smaller form factors to support a larger drive count in an array or server. SATA has been a driving factor in this disk evolution. For example, storage components like Seagate Technology's Barracuda 7200.11 desktop hard drives offer 500 GB, 750 GB and 1.0 TB capacities for high-volume desktop PCs and low-end nearline storage applications like archiving/content addressed storage (CAS) or disk-to-disk backups (Seagate's Barracuda family reaches 750 GB). The 7200.11 also uses a 3 Gbps serial-disk interface with sustained transfer rates of 105 MBps. Early concerns about SATA disk reliability are being addressed by disk makers -- in addition to vibration resistance, 7200.11 disks tout MTBFs up to 750,000 hours and include up to 32 MB of cache to reduce constant disk access. Still, the disk uses an average of 12 watts of power (8 watts while idle).
Serial attached SCSI (SAS) is another serial drive interface that has evolved to serve drives that offer enhanced performance between that of SATA and Fibre Channel (FC). Storage components like Seagate's Cheetah 15K.5 offer the same 3 Gbps interface speed in SAS form, but the 15,000 rpm rotational speed supports burst data transfers as high as 300 MBps. Although the Cheetah drives only offer a capacity up to 300 GB, their mean time between failure (MTBF) i rated to 1.4M hours. SAS host adapters can also support SATA drives on the same interface, so SAS and SATA drives can be grouped on the same SAS interface for storage tiering. Features like background media scanning proactively examine the disk for errors. Although SAS drives have not reached the same popularity as SATA, they have proven worthwhile in low-end enterprise storage.
FC storage components continue to set the standard for enterprise-class reliability and performance. Seagate's Cheetah 15K.6 drive family supports up to 450 GB of capacity through a 4 Gbps FC interface for integration into storage area network (SAN) environments. Long-term reliability is supported with MTBF ratings to 1.6 million hours. Sustained transfer rates up to 164 MBps benefit fast data transfers in tasks like transaction processing and Internet/e-commerce.
SATA and SAS disk users have been concerned by reliability issues. Since these storage components offer very large capacities, and rebuilds can demand many hours, there is a real statistical probability that a second disk might fail before the first failed disk is rebuilt. This would obviously result in unacceptable data loss, so RAID 6 has emerged as a major tool to guard against two simultaneous disk faults. This is accomplished through dual-parity -- calculating and writing two independent sets of parity data across the disk group. It's important to note that some vendors use the term "RAID-DP" to indicate a proprietary implementation of "dual parity."
To minimize rebuild times, today's RAID controllers utilize drive diagnostic data (a.k.a. SMART data) to identify marginal drives and implement pre-emptive rebuilds. This allows the storage system to actually rebuild a drive before it fails, placing the newly rebuilt drive into the RAID group and removing the original drive so that it can be replaced by a technician. Pre-emptive rebuilds can take more time than regular rebuilds because the RAID group must remain available to applications and users during the process -- some users may notice a reduction in storage performance during the rebuild, and this may impact the storage SLA.
While RAID 6/DP offers superior protection for SATA and SAS disks, there are two important disadvantages. First, the performance penalty needed to calculate and write two independent sets of parity information is roughly twice that of single-parity systems (e.g., RAID 5), though read performance is unaffected. Second, two additional disks are needed for each RAID 6 group, where RAID 5 groups need only one additional disk. With disk costs continuing to fall, the cost of additional storage components for RAID support is generally a small incremental cost. Not all storage systems support RAID 6 today, though the feature is increasingly popular.
Storage component enclosures
Arrays and clusters are also reflecting the changes in storage components. One important change is storage density. Not only are drive capacities growing, but the size of each hard drive is actually shrinking. A 3.5-inch form factor is commonplace today, but drives like Seagate's Savvio family use a smaller 2.5-inch form factor. Smaller products allow more storage to reside in any given chassis. Blade servers and 1U rack devices can also include 2.5-inch drives where 3.5-inch hard drives were often prohibited in the past.
Increased storage density also translates into greater power and cooling demands in all types of storage component enclosures, including RAID racks, modular arrays, monolithic arrays and storage clusters. Since each hard drive consumes power, the overall power demands from dozens (even hundreds) of disks can become substantial. Power is also liberated as heat and vibration, which can impact long-term disk reliability. Storage administrators often opt for low-power, vibration-resistant drive models intended for 24/7 enterprise operation. For example, Seagate's Savvio 10K.2 and Savvio 15K 2.5-inch hard drives all use less than 6 watts in normal operation. By comparison, full-sized 3.5-inch hard drives typically use over 11 watts.
Storage components are also experiencing a shift in features; capabilities that had once been the domain of large, monolithic storage systems are siphoning down to smaller and less expensive modular arrays and RAID racks. For example, the SATABoy is a 14-drive RAID system from Nexsan Technologies. The SATABoy offers hot-swappable drives and completely redundant components, along with features like dynamic spare pooling and dedicated spares -- once the exclusive domain of high-end storage arrays from vendors like EMC Corp. or Hitachi Data Systems Inc. (HDS). The system also provides MAID (Massive Array of Idle Disks) support that idles the drives and saves energy. Traditionally advanced software-based functions like snapshots and mirroring are now appearing with entry-level storage arrays.
Network storage components
Storage networks are benefiting from faster connectivity. For example, iSCSI networks employ readily available Gigabit Ethernet connections (GigE). Network attached storage (NAS) boxes are available with GigE ports, though some organizations opt to connect a NAS across a dedicated 10 GigE port to mimic SAN behaviors (dubbed a "poor man's SAN") allowing for high-speed data transfers along with simplified SAN management. Most fabric switches now offer GigE interswitch link (ISL) ports that allow the switches to interconnect and pass traffic between each other at higher speeds than typical 10/100 Mbps links. NAS and iSCSI storage components will invariably benefit from the migration to 10 GigE once the technology can be implemented cost-effectively over copper.
SAN switches and storage arrays now provide 4 Gbps FC ports as standard equipment -- maintaining backward compatibility with 1 Gbps and 2 Gbps data rates. Sophisticated storage components like director-class switches provide FC ports that handle data rates up to 10 Gbps though 10 Gbps FC ports are not backward compatible with slower port speeds, so 10 Gbps ports are mainly used for inter-switch links (ISLs). Expect to see 8 Gbps FC ports introduced in the near future for added speed and backward compatibility.
Another trend in network storage components is the addition of intelligent features at the fabric -- often implemented at the switch. These features include storage virtualization, data migration and replication, backup and restoration capability, better interoperability between storage components, as well as uniform storage provisioning and management.