Volume Number: 21 (2005)
Issue Number: 9
Column Tag: Programming
by Emmanuel Stein
With all of the exisiting storage interfaces to choose from, Mac professionals are routinely faced with a bewildering array of technology and configuration options. Moreover, with Apple's entrance into the high-end enterprise server market, the identification and implementation of the array of extant interconnect technologies has become an increasingly complex task. This article represents a compass for Mac professionals to navigate the maze of available solutions. Using a modified glossary-style format, each technology is defined in detail with practical commentary on the application of the various technologies covered. Whenever appropriate, I reference how these technologies apply to the Mac and give readers a sense of what technologies are being developed and which ones will likely be supplanted in the near term.
Advanced Technology Attachment (ATA) is a parallel interface for connecting internal storage devices to personal computers and has many aliases including ATA/ATAPI, EIDE, ATA-2, Fast ATA, ATA-3, Ultra ATA, and Ultra DMA. ATA supports two devices per bus with only one drive able to access said bus at a time. ATA-7 (ATA-133), the current and final revision of the ATA specification, supports a maximum data transfer rate of 133 MBytes/sec across an 18-inch cable. Although ATA is best suited to low cost consumer storage solutions, it has been successfully adapted to the enterprise market, via clever hardware engineering on Apple's part in their Xserve RAID product, which overcomes many of the shortcomings of ATA. Every ATA drive in the Xserve RAID is connected to a dedicated bus, with two independent and non-redundant controllers that support non-native Tagged Command Queuing, an essential feature for enterprise-grade storage solutions. PATA is, nevertheless, a technology of the past and, though it is still employed in low end Macs and other devices, it is being replaced by its serial counterpart.
SATA is quickly replacing parallel ATA as the internal storage interface of choice for desktop machines and low cost server solutions. SATA made its Mac debut in Apple's PowerMac G5 line and has been extended the new iMac line. SATA has also been used in newer Xserves, with the notable exception of the Xserve RAID. SATA remains compatible, on the software level, with ATA and, therefore, all the existing firmware, operating systems, and applications that work with ATA will also work with SATA. SATA advantages include point-to-point interconnect that enables full bandwidth for each drive, native command queuing, lower voltage requirements, hot-plug capability, thin cabling, longer cable length, faster transfer rate and CRC error checking. These added features make SATA suitable for cost sensitive server deployments.
SATA II - maximum transfer rate is 300MBytes/sec over a cable up to 1 meter long
SATA-III - maximum transfer rate is 600MBytes/sec (will be available in 2007)
ESATA is an emerging external storage interface extension to SATA, which offers an inexpensive and faster external storage solution and is poised to overtake FireWire disk implementations, so common in the today's Mac environments. Currently, eSATA supports a maximum transfer rate of 150MBytes /sec and has a maximum cable length of 2 meters.
Parallel SCSI, better known as just plain SCSI, is a parallel bus technology that connects a variety of peripherals to one's computer. Parallel SCSI can support up to 16 devices on a single bus. Although it offers faster throughput and comes with more advanced drive features than ATA, its price to performance ratio leaves something to be desired. SCSI's parallel nature makes it hard to deploy in larger environments and the technology has reached its practical performance limit. Ultra 320 is the last line in the evolution of parallel SCSI and supports a maximum transfer rate of 320MBytes/sec over a cable up to 12 meters in length. Although SCSI has dominated the enterprise market for the past two decades, SAS represents a natural evolution of the technology, which is likely to eclipse SCSI in the near term.
SAS is a serial implementation of SCSI that is designed to extend the existing capabilities of parallel SCSI and enables compatibility with Serial ATA. SAS allows users to interconnect either SATA or SAS hard disks within an expander that can be configured to support up to 16,256 mixed drives An essential component of the SAS infrastructure, expanders enable scalable storage with extended distances between devices and support for multiple device attachment to multiple host initiators, thereby affording fault tolerance. The maximum cable length of SAS is 10 meters. SAS also supports smaller storage devices such as 2.5-inch drives. Its point-to-point configuration and highly scalable architecture makes SAS an advanced, high-cost enterprise-class solution for applications that do not require the long cable lengths associated with Fibre technologies.
SAS 300 - maximum transfer rate is 300MBytes/sec
SAS 600 - maximum transfer rate is 600MBytes/sec (will be available in 2007)
SAS 1200 - maximum transfer rate is 1200MBytes/sec (will be available in 2010)
Fibre Channel is a serial bus interface for high-speed storage solutions. It is interoperable with SCSI, IP, ATM, HIPPI and IEEE 802.2 protocols and offers a maximum data transfer rate of 400 MBytes/sec, with a maximum length of 30 meters over copper and 10 kilometers using optical. Fibre Channel offers point-to-point, switched, and loop interfaces. Via arbitrated loop (FC-AL), fibre channel can support up to 127 devices and up to 15,663,104 using switched fabric. Fibre has been increasingly deployed for enterprise applications such as SAN (Storage Area Network), which benefit from Fibre Channel-specific features like long cable length and capacious device support.
FireWire is a high-speed serial bus developed by Apple. It offers simple connectivity for multimedia peripherals such as digital video cameras, iPods and high-performance storage. FireWire can connect up to 63 devices on a single bus and supports hot-swapping. FireWire was originally designed for streaming digital video and, therefore, was not implemented with heavy file serving in mind, with its frequent random reads and writes. FireWire hard drives have been widely deployed as an inexpensive backup solution, particularly in Mac environments. Although cable length is limited to 4.5 meters, distances up to 50 meters can be achieved using plastic optical fibre (POF) and 100 meters with hard polymer or glass-based fiber cabling. Although, not considered an enterprise grade solution, Apple's popularization and support for this technology, have gone a long way to legitimizing its use in the enterprise, as well as, home environments.
FireWire 400 - maximum transfer rate is 400 MBits/sec
FireWire 800 - maximum transfer rate is 800 MBits/sec
USB is designed for a variety of peripheral interconnect applications. A single USB channel can connect up to 127 devices with a maximum cable length of 5 meters. External USB-based hard drives are popular among PC users since FireWire does not typically come standard. Although used in many third party storage solutions, USB-based hard drives not appropriate for server or other high-end applications like digital video and file serving.
USB 1.1 - maximum transfer rate is 12 MBits/sec
USB 2.0 (Hi-speed USB) - maximum transfer rate is 480 MBits/sec
RAID is a method of consolidating multiple drives into Logical Units (LUNs). RAID can be implemented using extant hard drive technologies, including SCSI, SATA, and ATA. Depending on the particular RAID configuration used, benefits will include increased data integrity and redundancy, facilitated disk cluster management, fault-tolerance, and improved throughput. RAID solutions are commonly used in the enterprise and have now gained a foothold in the high-end consumer market.
RAID 0 - also known as striping, splits data evenly across all drives and lacks parity
information for redundancy. It increases performance in read and write operations, but introduces an
increased rate of failure with each disk added. RAID 0 is recommended for applications that are not
data critical but require high speed such as audio and video streaming and editing, graphic design
and high-end gaming systems.
RAID 1 - aka mirroring creates an exact copy of the data on two or more disks. It
increases performance in read operations but imposes a small penalty for write operations. RAID 1's
inefficient use of the disk space is considered its main drawback. It is well suited, however, to
data critical applications in which superior reliability is required.
RAID 0+1 - a combination of RAID 0 and RAID 1, a mirror of stripes. Benefits include
increased performance in read and write operations while maintaining data redundancy. Disadvantages
include high cost and limited scalability. Performance is severely degraded during recovery, since
all disks in an array must participate in the rebuilding process. RAID 0+1 is recommended for
imaging applications and general fileserver tasks that require high performance but do not demand
maximum redundancy,
RAID 3 - byte-level striping with a dedicated parity disk. Its random read/write
performance is poor due to byte-level striping, parity calculation overhead and the bottleneck of a
dedicated parity drive. RAID 3 is useful for applications that deal with large files such as video
editing and prepress, which benefit from increased performance with redundancy.
RAID 5 - data and parity information are striped across all drives. It is the most
popular of the RAID configurations because it provides a good balance between availability,
capacity, data protection and performance relative to other RAID implementations. RAID 5 is often
employed in transaction processing applications such as web, e-mail, and news servers, where large
numbers of users access information concurrently.
RAID 10 - a combination of RAID 1 and RAID 0, a stripe of mirrors. It is very costly
to deploy and suffers from limited scalability. Unique advantages include rapid data transfer, with
its striped array, and superior data redundancy. RAID 10 is used in database solutions that require
high performance and fault tolerance.
RAID 50 - block striping with the distributed parity of RAID 5 combined with the block striping of RAID 0. This combination achieves fast transfer rates via distributed parity and is very costly to implement. RAID 50 is only recommended for data critical applications that require high fault tolerance with fast read/write operations.
Emmanuel Stein has been an avid Mac user since 1984 and has honed his cross-platform skills while working at France Telecom, Time Magazine and Reed-Elsevier. He has recently started his own Mac-centric consulting company, MacVerse, which offers implementation, system administration and development services geared towards the enterprise market. As a diehard GNU/Linux geek, he enjoys hacking open source software and experimenting with new open source projects on OS X. You may reach him at macverse@mac.com
Now that you have a better idea as to what all the different technologies are with their strengths and weaknesses, which do you use? Unfortunately, there is no one easy answer, but we thought we'd give you one example of a way to take advantage of these technologies.
Here at MacTech Magazine, we're always bringing in new equipment and replacing old stuff. It's part of the magazine process. Recently, we decommissioned an older Power Mac G4 tower machine, which was running perfectly fine, and was wondering what to do with it. As a 450MHz G4, it wasn't fast enough to be a workstation, but it was plenty fast to be aserver ... depending on the type.
We chose to turn it into another file server. We installed Mac OS X Server on it, gave it 768MB of RAM, and then started to look and see what disks we could put into it. The problem with the older machines is that they, many times, don't support drives larger than 128GB, and we weren't going to invest in drives that small.
Instead of investing in old technology, we decided to combine the best of new and old. In this case, we went with SATA drives from Maxtor. Why SATA? Well, for starters, they break the 128GB barrier as a general rule.
And, these Maxtor drives are designed to live for a minimum of 5 years. Meaning that even though these are going in an older machine, we're investing in technology that will be useful for a while. We can use these drives for something else down the road. This is one of the best reasons to go with SATA instead of Ultra ATA.
Specifically, we went with the Maxtor DiamondMax 10. These Serial ATA drives feature a 16MB buffer, at 7200 RPM, and a data transfer rate of 150 MBps. At time of writing, street prices are between $110 and $160 for these drives making them a cost effective solution.
Yep, none of the older ones do. No worries. You can get an inexpensive SATA PCI card for around $50. In our case, we used one from Sonnet, and attached two SATA drives to the one cord. Don't forget, both the interface cables and the power cables are different than ATA. You'll need a power adapter cable, and SATA interface cables if they don't come with your card.
Maxtor has some other cool stuff as well. The Maxtor Shared Storage drive is a budget conscious way to share files. It's basically, a low cost file server that you attach to your network, is automatically configured (and can be further configured in a web interface) and can even help you share two USB printers. They come in 200 and 300 GB versions.
Maxtor also has some great "OneTouch" devices which you can use as either an external drive, or as an easy way to do a "one touch" backup. The Maxtor OneTouch II, FireWire 800 Edition actually sports a triple interface: FireWire 800, FireWire 400 and USB 2.0. Its Mac bootable feature turns a drive into a bootable system drive. It includes a simple user interface and automated back-up software so you can designate a portion of drive for high-performance storage and the other portion for backing up critical data. Two capacities: 200 GB (MSRP: $229.95); 300 GB(MSRP: $319.95)
by Neil Ticktin, publisher of MacTech Magazine.
