I found this article from Seagate outlining the challenges and rewards of their new Cheetah Disk Drive running at 10,000 RPM. Obviously, Seagate feels this program is very important to their effort to remain the leader in Enterprise.
So, what's the point? This is the same program that is the first to announce using Innovex's flex circuit HIF. Therefore, you can see that Seagate is thouroughly behind Innovex's technology.
Anyway, here's the article: idema.org
and this is the site for very good source of information on strage in general.
idema.org
CHEETAH: THE CHALLENGES AND BENEFITS
By Dave B. Anderson, Seagate Technology
Over the past several years, a few milestone developments have improved the performance of disk storage by step functions. The 5400-RPM 5.25" drive and the 7200-RPM 3.5" drive each in turn established a new standard for high performance storage. The introduction of the first 10,000-RPM drive promises to offer the next major improvement.
The Challenges of 10,000 RPM
When it was decided to develop the 10,000-RPM disk drive, it was obvious that several challenges had to be overcome. The two biggest were reading and writing at higher transfer rates and managing the spindle speed.
" Reading & Writing: Data rates increase, for a given BPI, in proportion to the spindle speed. This presents new challenges for both LSI design and head/media integration. The movement to MR heads introduced additional challenges. While a thin film head benefits from higher linear velocity (it receives a stronger signal), the signal from an MR head is insensitive to velocity yet the noise continues to increase as a function of higher speed. Therefore, a higher amplitude signal must be achieved in order to compensate for the increased noise.
" Spindle Speed: The development of the 7200-RPM Barracuda spindle revealed that not only power dissipation but also lubricant life, vibration, and acoustics had to be well managed. At 10,000-RPM, the bearings run about 10ø C hotter. The higher temperatures cause the lubricant to lose viscosity and to accelerate oxidation. The bearings also travel about 39% farther each year making for wear concerns.
Higher Transfer Rates
To meet the needs for higher effective amplitude, a dual stripe MR head was employed. Because the differential characteristics of the head produces a stronger read signal, the signal to noise ratio was greatly improved thus enhancing the margin when reading data.
Managing 10,000-RPM
Experience developing the 7200-RPM underpinned the development of the Cheetah in two ways: life testing the components and the development of analytical models and analysis techniques. Determining the necessity and the best approach to life testing of the components led to the development of analytical models and analysis techniques. These, in turn, were adopted to assess the effects of 10,000-RPM.
Since no one had previously attempted a 7200-RPM disk spindle, there was a question of whether acceptable spin motors were available. This was not the only complication. Even if this could be developed, a major hurdle would be to convince the customers that there was no danger of the spindle motor wearing out due to the higher RPM. However, by the time this product was shipping, the dependability of its spindle had been demonstrated convincingly. Most importantly, a great deal was learned about what it took to validate a new RPM design point. This was directly applied in developing the plan for assessing 10,000-RPM spindles and components.
In fact, the spindle motor used in the 7200-RPM turned out to be fine for the higher RPM. It had been used in commercial applications up to 20,000-RPM - though not with the same tolerance requirements. While other 7200 components proved to be inadequate for 10,000-RPM, the spindle used was more than suitable for the task.
Life Testing Components
Nothing proved more valuable during 7200-RPM development than a rigorous program established for life testing components, particularly spindles and bearings. This testing revealed several changes needed for a 7200-RPM motor to last five years.
Life testing at 10,000 revealed that while many of the 7200 components would work in this environment, others, including bearings and lubricants would have to be changed significantly. In some cases, however, no modeling or analysis could have revealed the needed changes.
The results from evaluating lubricants is a good example. It is well known that key to its usefulness is the high temperature viscosity of the bearing grease. Early testing indicated the grease being used for 7200 was not adequate. From the greases evaluated, the ten most promising candidates were submitted to life testing. This empirical work disclosed an interesting fact - due to other characteristics, the grease with the highest viscosity at high temperatures was not the best choice for the 10,000-RPM disk spindle.
Life testing in most cases involves running identical sets of devices at several different accelerated temperatures so that trends can be extrapolated. For example, in one set of tests each combination of spindle, bearing and grease was run at 50, 60 and 70 degrees Centigrade with bearing temperatures reaching over 85 degrees. This was done with the 7200-RPM to develop a wear model which field experience proved to be quite accurate. The same process laid the foundation for the 10,000-RPM wear model.
An interesting aspect of the spindle testing is the inspection process. For one series of tests, two spindles were removed each month and torn down to look for wear symptoms. The inspection process included:
ÿ
<Picture>
Figure 1: Acoustics vs accelerated run.
ÿ
1. Visual assessment of wear
2. X-ray for surface and subsurface fatigue and defects
3. Surface characterization
4. Measuring run current - looking for increased friction
5. Measurement of non-repeatable runout - looking for any increase
6. Acoustic measurements
7. Lubricant chemical analysis, looking for weight loss, antioxidant presence, and amount of oxidation.
The final results of the acoustic analysis with the final configuration spindle are shown in Figure 1. Acoustic measurements are a method used to assess any change in the bearing friction. An usual increase is an early indicator of a wear condition. The other phases of the inspection process were measured against similar criteria. Figure 1 shows that the Cheetah spindle has no worse acoustic increase than the field proven Barracuda 2LP (several million have been shipped without having a spindle wear issue).
Among the new technologies evaluated were hydrodynamic bearings (HDB). They seemed to have some desirable traits, such as lower acoustics. However, the life testing had shown some definite problems. While considerable progress has been made in HDBs, at the time a component decision had to be made they were not suitable for this application.
ÿ
Modeling and Analysis of 10,000-RPM Mechanics
Several models were developed for analyzing the effect of higher RPM during testing of the Barracuda HDA. After some preliminary testing, modifying these for 10,000-RPM was not a difficult task. This proved invaluable for saving design time.
A secondary benefit of the 7200-RPM spindle design effort was that it led to the development of a strategic relationship with lubricant vendors. Forging a strong relationship with these vendors facilitated much of the design changes for 10,000-RPM. This relationship proved particularly helpful because it not only provided the opportunity to obtain early access to the latest lubricants, but it also opened up the laboratories to a more in-depth exchange of information. Changing the composition of several of the grease components, especially the soap and antioxidant, turned out to be critical to the success of the 10,000-RPM design.
The Result: Cheetah
Early responses from customers confirm that the 10,000-RPM drive will set a new standard for high performance magnetic storage. While this article has focused primarily on the challenges specific to the 10,000-RPM, there are other innovations in this drive. The most interesting, from a performance perspective, is a technique for improving seek performance. During actuator movement, the servo system is alternately reading positioning information off two different surfaces. This doubles the servo bandwidth by having twice as many samples to employ in positioning. This feature benefits the acceleration and deceleration profiles of the drive. This, combined with a new, very short suspension, makes for a drive that has an excellent latency and transfer rate as well as the fastest seek performance.
The understandable presumption is that a 10,000-RPM drive will be less reliable than a drive with a lower rotation speed. This is, in fact, baseless. What we have learned through past designs, such as the Barracuda, is that by introducing new technology - even when it consumes more power - can actually result, with a thorough and creative design effort, in improved reliability. There is every reason to expect that a 10,000-RPM drive will have as good a reliability record as the best 7200-RPM drives, if not better (Table 1).
ÿ
------------------------------------------------------------------------
IMPROVEMENT
ÿ
Low Profile HDA: 51%
Half-Height HDA: 24%
Acoustics are the same as the latest version of the 7200-RPM drives (4.3 bels sound pressure). Test data and bearing life analysis show Cheetah life equivalent to or better than Barracuda. Vibration and shock sensitivity are less.
ÿ
ÿ
PERFORMANCE BENEFITS OF 10,000-RPM
ÿ
Focused on benefits that pertain specifically to spinning at 10,000-RPM. There are two that obviously accrue:
Increased Data Transfer Rate
Reduced Latency
ÿ
ÿ
To complete a read-modify-write sequence, one or more sectors must be read and rewritten on two different disks, the parity drive, and the data drive. Each of these involves an average latency plus another full rotation in order to rewrite to the original location. Here is a comparison of just latency times for this operation.
ÿ
ÿ
Locate Data
Rewrite
Total
Improvement
10,000-RPM
2.99
6
9.99
28%
7200-RPM
4.17
8.33
12.50
ÿ
Table 1: Comparison of 7200-RPM and 10,000-RPM.
------------------------------------------------------------------------
Which leads us to answering the question on everyone's mind. Where will the customers see the advantage of this higher performance?
" Increased Data Transfer Rate: Video processing gets lots of headlines but more mundane work, like data mining, can equally profit from this high data rate. For instance, a workload of random 64K transfers of typical data mining show about a 30% advantage over the 7200-RPM drives. The longer transfers of video applications, 256K and above, will reveal even more dramatic differences.
" Reduced Latency: Almost every I/O operation involved some amount of latency, the rotation time waiting for the desired sector to arrive under the read/write heads. In operations with little or no seeking, the benefit will be more pronounced. All applications will see some benefit from the improved latency. But there is one application where the latency really stands out: RAID 5 disk arrays. RAID 5 is the preferred array organization for business applications. One performance penalty associated with this parity arrangement is the updating of a small amount of information. (A small amount being less than the amount of data comprising an entire parity stripe). The updating requires reading the old data, using the old and new data to update the parity information, and writing the new data.ÿ
The Market for 10,000-RPM
When the first 7200-RPM drive was shipped, it was expected that only relatively small segments of each systems market workstations, servers, and mainframes would adopt it. However, it turned out that essentially every computer manufacturer placed 7200-RPM drives into their product line. In some cases, such as desktop PCs, it was only a small percentage of the systems. In others, like file servers, it quickly became the standard for storage. Overall the 7200-RPM drive dominated the mainstream high performance (SCSI) market within two generations of its first appearing.
What does that say about the prospects for 10,000-RPM and who will benefit from its performance advantage? That is easy - Everybody!
Dave B. Anderson is the director of systems storage architecture at Seagate Technology's disk drive division located in Bloomington, Minnesota. |
