SingleDriveVsRaid0

Also see this much more in-depth look (cache).

At any given point in time, all one needs to do is look at topic list in the SR Discussion Community to see a frenzy of enthusiasts pursuing increased drive performance through striping, or RAID 0. Many readers are pouring lots of money into such equipment and spending a great deal of time and effort configuring their arrays.

Details on RAID 0 may be found in this section of the SR Reference Guide. There are many misunderstandings based on the fact that with an even number of drives, when properly configured, RAID 0 will offer the combined capacity of all drives and almost the combined sequential transfer rate of all drives. This assumes that all drives in the array are identical. If this is not the case, each drive will generally be treated as if it is the same size as the smallest drive in the array, and the same speed as the slowest drive.

Does RAID 0 truly offer the combined space and performance of all drives?

Capacity will indeed appear as one large, combined drive. Bear in mind however, that data striped across multiple drives is much more vulnerable to loss as a physical failure of even one drive results in the loss of all data on the entire array. If capacity is the goal, it is almost always better to run the two drives as separate units.

What about performance? This, we suspect, is the primary reason why so many users doggedly pursue the RAID 0 "holy grail." This inevitably leads to dissapointment by those that notice little or no performance gain.

Theory states that RAID 0 increases the sequential transfer rate, but how much does this really effect performance in contemporary desktop machines? As often indicated in StorageReview's forums, the answer is: Not much. STR simply does not significantly impact performance of typical desktop applications.

There are certain uncommon situations where RAID 0 can significantly improve system performance. For example, editing of large audio or video files is sometimes limited by the maximum sequential transfer rate of the hard drives, but it is far more common for the processor to be limiting factor. Generally, if you frequently make simple edits to large media files, RAID 0 can potentially improve your productivity. Examples of "simple" edits might include removing portions of an uncompressed audio file, or combining two video files. Most "filters" in media editing software are not bottlenecked by disk speed. Note that the audio and video editing that I am referring to does not include encoding digital music (such as Ogg/MP3 compression) or encoding/transcoding video (such as converting a DVD video file to DivX, etc.).
These tasks are CPU limited, so RAID 0 would have little or no effect. In most cases, a CPU upgrade or RAM upgrade will provide a much more tangible benefit to these tasks.
One other example of a situation where RAID 0 might improve performance substantially is in certain games. Usually, first person shooters are not part of this list.
So, how do you tell what games might benefit significantly? It isn''t easy unless you have some knowledge of how the game works internally. For example, Interplay's "Baldur's Gate" series, as well as games based on the same engine (Planescape: Torment, Icewind Dale to name a few) benefit from RAID 0 because the levels that they load are essentially huge bitmaps. Because the hard drive is reading one large data file in a linear fashion (as opposed to the heads having to move rapidly back and forth to access many different files), sequential transfer rate is all but the only factor effecting loading speeds. Some other games benefit to a lesser degree.
As stated above, first person shooters rarely benefit from RAID 0.__ Frame rates will almost certainly not improve, as they are determined by your video card and processor above all else. In fact, theoretically your FPS frame rate may decrease, since many low-cost RAID controllers (anything made by Highpoint at the tiem of this writing, and most cards from Promise) implement RAID in software, so the process of splitting and combining data across your drives is done by your CPU, which could better be utilized by your game. That said, the CPU overhead of RAID0 is minimal on high-performance processors.

It is important to note that many RAID 1 (mirroring) controllers can offer the same benefits to hard drive performance (for reads, which are more common than writes), while actually improving data integrity. RAID 1+0 (for use with 4 or more drives) offers further substantial benefits. The disadvantage, of course, is that you do not get the combined space of all of your drives, due to data redundancy. The actual available disk space depends on the RAID level. Please check SR's Reference Guide (cache) for more information on the many RAID levels that exist today.

So what's a "real world" speed increase of typical Windows and Linux applications? It is difficult to put a solid number on this figure, because of the diversity of software out there, but it is reasonable to assume a 0% - 15% overall disk performance increase moving from a single disk drive to two in RAID 0, with rapidly diminishing returns as you add more drives.

For more concrete numbers, read on.
While formal, ongoing RAID tests currently fall outside the existing purview of SR, consider an array that we constructed in our third-generation testbed using a Promise FastTrak SX2000 in conjunction with a pair of 200 GB Maxtor DiamondMax Plus 9 drives. We''ve run StorageReview.com's Desktop DriveMarks to demonstrate what kind of application-level increases one may expect. Those who are unfamiliar with and those who doubt the veracity of the Desktop DriveMarks are strongly advised to re-read our methodology article to put these results in proper perspective.



Unsurprisingly, the dual-drive RAID 0 solution delivers double the sequential transfer rate of a single unit. The SR Office, High-End, and Gaming DriveMarks, however, all climb by less than 10%. Also consider the fact that the RAID array boasts double the capacity of the single drive: as a result, some of that performance increase we see between the single drive and the RAID array simply comes from the larger capacity and resultant shorter actuator travel distances. Is this worth twice the cost (plus the cost of the controller, unless you use pure software RAID)?

A notable exception arises within the SR Bootup DriveMark 2002. This particular test features higher-than-normal queue depths for a single-user machine, reflecting optimizations that Microsoft has made to the Windows XP bootup process. This high load allows an array's two or more independent actuators to service queued requests and improves overall performance. Further, XP tracks the order of requests during the boot process and does its best to reorder data found on a drive to facilitate sequential reads as a system starts up. Since the Bootup DriveMark 2002 trace was captured from a system that had been restarted and defragmented many times, this individual test reflects the transfer rate advantage that one achieves through RAID 0. Therefore, if the primary purpose of one's machine is to start Windows XP, RAID 0 offers a noteable decrease in boot time. That said, Windows XP boots up fairly quickly on most modern systems, so one wonders how many reboots it will take to compensate you for the time spend installing and configuring your RAID 0 array.

Again, RAID 0 does have its advantages in a handful of key applications and uses where data files are huge and/or data requests are highly sequential in nature. Data requests are not highly sequential, however, in typical desktop productivity and most gaming usage patterns, the most often cited in "Help me build my RAID 0!" posts.

The point? Dont assume RAID 0 offers increased performance for all or even most applications... and dont assume that transfer rates reflect application-level performance.

So RAID 0 probably won't help much. But will it hurt?

Despite the information above, many are still interested in pursuing RAID 0. Certainly, there are sometimes good reasons to do so. So the question then becomes: are there reasons AVOID RAID 0, purely from a performance perspective?

In regards to the specific question of whether a pair of drives in RAID 0 can degrade performance as compared to a single identical drive, other than the unlikely CPU example above the answer is "probably not." Some may say that access time can be increased (worse) due to a specific read (or write) requiring both heads to seek to their portion of the data, creating the result that average access time will be slower. However, this only really applies when accessing large numbers of small files as larger files will benefit from the improved STR. Unless the configured stripe size is too small, most of the small reads will access only one drive.

That said, badly written firmware on the controller can degrade performance, but this is a controller problem, not a problem with RAID technology itself, and will become increasingly rare as consumer-level RAID implementations continue to mature. One example of this problem is the occasional drive incompatibility. Some controllers have provided poor performance with certain brands or models of drive, though these problems are usually corrected quickly. Check the forums to see if a controller you are considering has known problems of this nature.

Finally, it is sometimes beneficial to use the identical drives individually rather than in a RAID array because it is then possible for each drive to individually serve a different process. For example, one drive could handle paging to disk, and the other could handle other disk I/O needed by applications. Using a media example, one drive could read audio/video and the other could write the modified result. This would be faster than having a RAID array do both reads and writes, because individual drives could read/write linearly, eliminating the need to move the actuators back and forth between the area of disk being written to and the area being read from.

To summarize, RAID 0 offers generally minimal performance gains, significantly increased risk of data loss, and greater cost. That said, it offers the ability to have one large partition using the combined space of your identical drives, and there are situations where the benefit of the benefits outweight the disadvantages. It is your computer: The choice is up to you.

Addendum: In June of 2004 SR took a much more in depth look at the effects of RAID on both SATA and SCSI arrays (cache). The basic gist - regardless of interface, controller, or drive used the conclusions above remain.

If you have further questions, feel free to ask in StorageReview's public forums (cache)