T E C H N O L O G Y
P A P E R
F R O M
S E A G AT E
CHEETAH X15-36LP
Economies of Speed
CHEETAH X15-36LP TECHNOLOGY PAPER
CONTENTS
Economies of Speed
Introduction..........................................................................................1
Balancing the System ............................................................................2
High Performance with Fewer Drives .......................................................2
Seek Time and Latency vs. Data Transfer Rate .........................................3
Optimizing Systems for Performance.......................................................5
Performance Benchmark Results ............................................................6
Applications..........................................................................................9
Conclusion..........................................................................................10
Appendix A. Cheetah X15-36LP Specifications and Features......................11
Reliability ...........................................................................................11
Power Requirements............................................................................11
Cheetah X15-36LP Advanced Features .................................................12
Interfaces ...........................................................................................13
Appendix B. Balancing the System...........................................................13
FIGURES
Figure 1. Market Acceptance of 15K Has Been Faster Than That of 10K ......1
Figure 2. Seek Time and Latency ...............................................................4
Figure 3. Time-to-Data Comparison Between 10K and 15K Drives ................5
Figure 4. Disc Drive Short-Stroking ............................................................6
Figure 5. StorageReview.com Performance Comparison—IOMeter .................6
Figure 6. Response Time Comparison—15K vs. 10K ...................................7
Figure 7. StorageReview.com Performance Comparison—WinBench ..............7
Figure 8. NetBench Performance Comparison: 15K vs.10K..........................8
 2002 Seagate Technology LLC. All rights reserved. Printed in USA.
Publication Number: TP-579B, February 2002
Seagate, Seagate Technology and the Seagate logo are registered trademarks of Seagate Technology LLC. Cheetah,
3D Defense System, Seagate Advanced Multidrive System and SeaShell are either registered trademarks or trademarks of Seagate Technology LLC. Other product names are registered trademarks or trademarks of their owners.
Seagate reserves the right to change, without notice, product offerings or specifications.
ECONOMIES OF SPEED
Economies of Speed
Introduction
Businesses are constantly being asked to do more with less—achieve results
faster with fewer resources—provide data to more clients faster and with lower
incremental IT administration cost. Businesses are responding with investment
in increased system processor speeds and increased network bandwidth. And
if similar investments are made in disc drive performance, the clients may find
more satisfactory response times, avoiding potential loss in business opportunity.
Seagate has recognized this issue and has been leading the way in developing
and delivering new highly reliable, state-of-the-art, cost-effective solutions—
solutions that the server and storage industry have come to rely on.
The most recent result of our effort is the industry’s first 15K-RPM disc drive, the
Seagate Cheetah X15. This drive has been recognized by many of the IT industry’s most successful companies as the most cost-effective means of meeting the
demand for data delivery to clients with response-time limits. The increased
demand for “time-to-data,” along with the 15K’s low power and cooling requirements, have driven the market to 15K. In fact, 15K-market acceptance has been
much faster than that of the first 10K drive, as shown in Figure 1.
350,000
300,000
250,000
200,000
150,000
100,000
50,000
0
1st
Qtr
2nd
Qtr
1st 10K-RPM Drive
3rd
Qtr
4th
Qtr
5th
Qtr
1st 15K-RPM Drive
Figure 1. Market Acceptance of 15K Has Been Faster Than That of 10K.
1
CHEETAH X15-36LP TECHNOLOGY PAPER
Seagate is now shipping the second-generation Cheetah X15, the Cheetah X15-36LP.
(Appendix A on page 11 provides a summary of the Cheetah X15-36LP’s specifications and features.)
Economies of Scope and Economies of Scale are terms for the concept that a
company gains certain efficiencies from increases in their scope or scale. This
paper outlines the concept of Economies of Speed and why the Cheetah X15-36LP
is so effective for a large variety of applications.
Balancing the System
Hard disc drives have been highlighted as an example of rapidly improving
technology as the areal density of the drive has doubled year after year, surpassing
many people’s expectations. But when considering the speed of the drive and
how quickly one can get at the drive’s data, the drive performance increases
have not been as impressive, especially compared to other components in the
system (see Appendix B on page 13). This is essentially due to the fact that the
drive’s speed is dictated by mechanical movements of the spinning disc
together with the mechanical positioning of the arm across the disc.
Due to the slow speed of drives, relative to the other components of the system,
most enterprise systems require large numbers of drives so that the drives are
not a bottleneck in the system. The infrastructure to support these drives is a
significant cost to the system. The alternative of designing the system with
extra cache to try to hide the drives’ slower performance during peak system
use is also very expensive.
High Performance with Fewer Drives
Because the Cheetah X15-36LP has much faster time-to-data than the newest
10K-RPM drives, it delivers up to 40 percent more IOPS (Input/Output
operations Per Second) than 10K-RPM drives for the majority of performance
applications. This means that to support a certain level of IOPS and maintain
a certain response time for a group of clients, fewer 15K-RPM drives are
required than would be if the workload were handled by 10K-RPM drives.
Economies of Speed is the concept that for a majority of performance applications,
the faster the drive is, the fewer drives are needed to deliver a performance level.
What does fewer drives translate to for the IT manager?
• Greater reliability because the system will have a lower spindle count.
Refer to Appendix A for discussion of Cheetah X15’s proven reliability. The
requirement for fewer drives increases system reliability even more.
2
ECONOMIES OF SPEED
• Lower overall system cost. Fewer drives means fewer HDAs, less cabling
and lower overall drive cost.
• Less space. Fewer drives means less space required to hold the drives.
This translates into denser performance; in other words, more IOPS/ft2
or more IOPS/ft3.
• Less total system power and cooling requirement because fewer drives
are required. Refer to Appendix A for discussion of Cheetah X15’s
extremely low power and cooling requirements that are on par with 10K
drives. The need for fewer drives increases the overall system power and
cooling requirements even more.
• Lower IT administration costs. As the costs of implementing and managing
complex storage solutions become more heavily weighted in people,
integration and operational expenses, having fewer drives to manage
translates into savings of storage-management labor cost.
Cheetah X15-36LP is the leader in lowering total cost of ownership (TCO)—
a critically important priority for organizations faced with managing rapidly
increasing volumes of information. Seagate customers have demonstrated the
fact that the 15K-RPM drive allows fewer drives to be used:
“Customers often have to configure a large number of disc drives
to achieve the performance levels required by their databases.
The Cheetah X15s have enabled us to achieve those same levels
of performance utilizing 42 percent fewer disc drives.”
—Compaq’s Mike Nikolaiev, Transaction Performance Council
(TPC) Director
“The 36-Gbyte 15K Cheetah drive offers a 33 percent improvement
in IOPS (input/output operations per second) over 10K-RPM
drives. Alternately, fewer 15K drives can be used in a system to
maintain the same overall performance, which reduces system
cost and manageability issues while increasing reliability…
”
—Hewlett-Packard’s Steve Young, Director of Product Marketing
Let’s look at how the 15K achieves this and the applications that will benefit
from the Economies of Speed.
Seek Time and Latency vs. Data-Transfer Rate
The vast majority of performance applications benefits more from faster seek
times and latency than from faster transfer rates. These applications are transferring small blocks of data that are located in a variety of positions within
the drive. A classical example of these types of applications is transaction
3
CHEETAH X15-36LP TECHNOLOGY PAPER
processing. A specific example of transaction processing is “online transaction
processing,” for which a typical pattern consists of two-thirds random reads
and one-third random writes, transferring data that is on average around 8 Kbytes
in length.
Disc drive access time is the time it takes for the head in the drive to get to
the location on the disc where it transfers the data. In these applications,
access time is much greater than data-transfer time, the time the drive uses to
actually read or write the data. The major components of access time are seek
time and latency. That is why seek time and latency play such a big role in
determining drive performance. For an illustration of seek time and latency,
refer to Figure 2.
Figure 2. Seek Time and Latency
Seek Time: After the drive receives a command, the head moves across the
disc (seeks) to the track that the data is on. This is the seek time. Due to the
Cheetah X15-36LP’s small 2.5-inch disc diameter, the total seek distance is
short, allowing a very fast seek time of 3.6 msec.
Latency: After the head is on the right track on the disc, the head waits on that
track until the desired sector (section of the disc for the data transfer) of the
track reaches the head. The time it takes for the data to reach the head is the
latency time.
The latency is determined by how fast the disc is spinning. The faster the disc
is spinning, the shorter the latency time because the data will reach the head
faster. For a 15K-RPM disc, the total time the disc takes to make one revolution
is 4 msec. For a 10K-RPM disc, the total time is 6 msec.
4
ECONOMIES OF SPEED
When the head reaches the track, the data could be immediately under the
head (actual latency = 0 msec), or it could have just passed the head, in
which case the head must wait a full revolution (4.0 msec). Therefore, on
average, the latency will be 2.0 msec for a 15K drive.
The Cheetah X15-36LP’s seek time plus latency is 5.6 msec, 2.1 msec faster
than that of the 10K Cheetah 73LP 36-Gbyte drive.
Data Transfer Time: For the majority of performance applications, the remaining
time the drive takes to transfer the data, determined by the data-transfer rate,
is very small compared to the access time, and therefore has relatively little
impact on the drive. It is the 15K’s access time that has assured its dominance
in the performance marketplace.
10K
2.00
4.7
Total
Time
to Data
(msec)
5.95
2.99
0.14
0.2
3.6
Data Transfer Time*
15K
Latency
0.14
Command Time
Average
Seek Time
0.2
Figure 3 compares the time to data between 10K and 15K drives.
8.03
10K increases
time to data
by 35%
Time to Data (msec)
*8K block transfer size
Figure 3. Time-to-Data Comparison Between 10K and 15K Drives
Optimizing Systems for Performance
To ensure that the disc drive is not a bottleneck in the system, the drive’s
access time must be minimized. Access time for 10K-RPM drives can be minimized by storing data only on the outermost part of the disc, so that the arm
does not need to seek very far to get to the right track. This is referred to as
destroking, or short-stroking, the disc.
In Figure 4 on page 6, data is stored on the dark-shaded outer diameter of the
disc. The lighter shade is available data storage that remains unused to keep
the seek time to a minimum. Minimizing the seek time compensates for the
longer latency and seek time of the 10K-RPM drive. The result is a very
expensive cost per usable gigabyte and cost per transaction to the end-user.
5
CHEETAH X15-36LP TECHNOLOGY PAPER
The Cheetah X15-36LP decreases the
access time without giving up the capacity
on the disc. The result is that fewer drives
are needed of a given capacity to meet
the performance requirements of the system, if those drives are Cheetah X1536LP drives instead of 10K drives. This is
the concept behind Economies of Speed.
Example: Assume that you want to store
360 Gbytes of data and want to minimize
Figure 4: Disc Drive Short-Stroking access time to that data. If you store this
data on ten 15K drives, the average seek
time plus latency is 5.6 msec. To maintain the 5.6 msec seek and latency
time on a set of 10K drives, the seek time must be maintained at an average
of 2.6 msec because the 10K drives’ latency is 1 msec higher than that of
15K drives. This means the 10K drive must be short-stroked to drop its
4.7-msec seek time to the 2.6-msec seek time limit, reducing the capacity for
each 10K drive. Therefore, additional 10K drives will be required to store the
360 Gbytes of data.
Performance Benchmark Results
How do Cheetah X15-36LP’s faster access times translate in terms of performance
benchmarks? Here we take a look at popular performance benchmarks, comparing
the Cheetah X15-36LP with the fastest 10K drives introduced in 2001.
StorageReview.com, a third-party performance evaluator of disc drives, published
the results in Figure 5. In each of StorageReview.com’s IOMeter patterns—
File Server, Database and Workstation—the 15K-RPM drive outperformed the
10K-RPM drive by 45 percent.
15K delivers
45% more
IOPS than 10K
IOMeter
StorageReview.com 7/19/01
400
IOPS
300
200
100
0
File Server
10K 73LP
Database
Workstation
Cheetah X15-36LP
Figure 5. StorageReview.com Performance Comparison—IOMeter
6
ECONOMIES OF SPEED
Response Time (msec)
Response Time (msec)
Response Time (msec)
The Cheetah X15-36LP enables support of a higher level of IOPS, enabling
the system to support more clients and have a faster response time for each
client. As businesses push an ever-increasing amount of information onto
Intranets and the Internet, the need to improve response times is critical.
Figure 6 illustrates the faster response time of the Cheetah X15-36LP over
the newest 10K-RPM drives.
Response Time
IOMeter, 8K OLTP, Queue Depth = 1
Write Cache On
Write Cache Off
10
8
6
4
2
0
Competitor 10K 73LP
36 Gbytes
Cheetah 10K 73LP
36 Gbytes
Competitor 10K 73LP
36 Gbytes
Cheetah 10K 73LP
Cheetah X15-36LP
Response Time
IOMeter, 8K OLTP, Queue Depth = 4
30
20
10
0
Competitor 10K 73LP
36 Gbytes
Cheetah 10K 73LP
36 Gbytes
Competitor 10K 73LP
36 Gbytes
Cheetah 10K 73LP
Cheetah X15-36LP
Response Time
IOMeter, 8K OLTP, Queue Depth = 16
90
75
60
45
30
15
0
Competitor 10K 73LP
36 Gbytes
Cheetah 10K 73LP
36 Gbytes
Competitor 10K 73LP
36 Gbytes
Cheetah 10K 73LP
Cheetah X15-36LP
Figure 6. Response Time Comparison—15K vs. 10K
The Cheetah X15-36LP also demonstrates its superior performance with the
WinBench performance benchmarks. Figure 7 shows the results as measured
by third-party StorageReview.com.
WinBench
Business Disk 99
WinBench
26%
higher
score
High-End Disk 99
StorageReview.com 7/19/01
12,000
35,000
10,000
30,000
Mbytes/sec
Mbytes/sec
StorageReview.com 7/19/01
8,000
6,000
4,000
2,000
28%
higher
score
25,000
20,000
15,000
10,000
5,000
0
0
10K Cheetah 73LP
15K Cheetah X15-36LP
10K Cheetah 73LP
15K Cheetah X15-36LP
Figure 7. StorageReview.com Performance Comparison—WinBench
7
CHEETAH X15-36LP TECHNOLOGY PAPER
Let us now look at the difference in performance at the server level. NetBench
is the current Ziff Davis benchmark measuring the performance of file servers.
NetBench does this by using a network of clients to generate file I/O requests
for the server under test. Various levels of client activity are used. NetBench
measures the throughput and response times resulting from these requests
and generates a data set detailing file server performance. Plotting this data
set yields a chart with network throughput (units of Mbits/second) on the y-axis
and client load (number of clients) on the x-axis. The chart shows a steep
climb in network throughput as client load increases. At these initial client
levels, much of the file I/O activity is handled from the available system cache
in the server or host bus adapter. At some point, throughput peaks and begins
a slow decay as client load is increased. It is in this region of the curve that
differences in drives are obvious.
Netbench 7.0 — 4 Drive RAID 0+1
PIII Xeon 500 / Windows NT 4.0 / Compaq RAID Adapter
Throughput (Mbits/sec)
100
80
60
The fastest 10K drive
reaches 85 Mbits/sec
with 12 clients.
40
Cheetah X15-36LP
reaches 86 Mbits/sec
with 40 clients.
20
0
1
4
8
12
16
20
24
28
32
36
40
44
48
52
56
60
52
56
60
Number of Clients
Netbench 7.0 — 4 Drive RAID 5
PIII Xeon 500 / Windows NT 4.0 / Compaq RAID Adapter
Throughput (Mbits/sec)
100
75
50
The fastest 10K drive
reaches 60 Mbits/sec
with 8 clients.
25
Cheetah X15-36LP
reaches 62 Mbits/sec
with 52 clients.
0
1
4
8
12
16
20
24 28 32 36 40
Number of Clients
Cheetah 36XL
Cheetah 73LP
Competitor 10K 73LP
Competitor 10K 73LP
44
48
Cheetah X15-36LP
Figure 8. NetBench Performance Comparison: 15K vs.10K
8
ECONOMIES OF SPEED
The top chart in Figure 8 reflects RAID 0+1 setup. RAID 0+1 is a combination
of RAID 0 (striping) and RAID 1 (mirroring), where the data is both striped
across a set of drives and the set of drives is mirrored onto another set of
drives. The second chart reflects classic RAID 5. As is reflected in these
charts, the Cheetah X15-36LP exceeds 10K performance while supporting
many times the number of clients that 10K supports.
Incorporating the Cheetah X15-36LP into the system may reduce the need to
over-design the system with larger amounts of expensive cache to cover peak
periods of system use. The Cheetah X15-36LP may also increase consistency
of response time when caching algorithms no longer can handle the load or
the cache is saturated, or where there is diverse client activity such that the
system cannot rely on what is in the cache to ensure adequate response times.
Incorporating the Cheetah X15-36LP may be the more cost-efficient means of
building a performance-balanced system.
Applications
The Cheetah X15-36LP fills the need for exceptionally high reliability and
faster access times in the vast majority of performance applications. The
Cheetah X15 drives are incorporated in large numbers in a wide variety of
hardware platforms:
• Midrange servers
• High-end enterprise servers
• Workstations
• SANs, etc.
Examples of these include Compaq’s ProLiant servers, AlphaServer systems
and StorageWorks enterprise storage solutions; HP’s SureStore E FC 60 Disk
Array and SureStore E SC10 and HVD10 Disk Systems; LSI’s MetaStor
storage systems; and XIOtech’s MAGNITUDE SAN.
Within these, the Cheetah X15 benefits a broad array of applications requiring
faster access times:
• Database queries and transaction updates (transactional by definition)
• Data mining/warehousing (allows quick manipulation of data)
• File and print (helps ensure fast response time as clients are added)
• E-mail (helps handle time to copy and updates individual messages of
copied e-mails)
9
CHEETAH X15-36LP TECHNOLOGY PAPER
• Internet (aside from caching on the front end, the Cheetah X15-36LP
allows for the e-commerce transactions on the back end, similar to a
database application, but without the ability to schedule peak use.)
• A/V editing
Conclusion
The Cheetah X15-36LP delivers experience, reliability and performance. It is
designed by the same team that delivered the first-generation Cheetah X15
that achieved:
• World-record performance
• Highest reliability rating in the industry, supported by an excellent field
reliability track record
• High-volume adoption by major OEMs, distributors and resellers
The Cheetah X15-36LP is poised to build on that success. This second-generation 15K-RPM drive brings increased capacity and performance and lower
acoustics, all with exceptionally low power.
The Cheetah X15-36LP achieves up to 40 percent more IOPS than 10K-RPM
drives. The Cheetah X15-36LP achieves 10K-RPM’s high IOPS and rapid
response time, but with much fewer drives. These Economies of Speed
translate into a more successful solution for the customer.
10
ECONOMIES OF SPEED
Appendix A. Cheetah X15-36LP
Specifications and Features
Cheetah X15-36LP
36 Gbytes
Interfaces:
Ultra160 160 Mbytes/sec
ST336752 LW/LC
2 Gbyte FC 200 Mbytes/sec
ST336752 FC
ST336732 LW/LC
Ultra320 320 Mbytes/sec
36.7
Capacity (Gbytes)
8/4
Heads/Discs
3.6
Seek: Avg Read (msec)
4.2
Avg Write (msec)
0.3
Adjacent Track Read (msec)
0.4
Adjacent Track Write (msec)
Transfer Rate: Internal (Mbits/sec)
522–709
Internal Formatted (Mbytes/sec)
51–69
Sustained (Mbytes/sec)
42–58
8
Cache Buffer (Mbytes)
12 / 13
Power: Idle SCSI/FC (watts)
16 / 16.3
Peak Opr. SCSI/FC (watts)
Acoustics Idle (bels)
3.7
250
2 msec Nonop Shock (Gs)
1,200,000 / 0.73%
MTBF (hours) / AFR
18 Gbytes
ST336752 LW/LC
ST336752 FC
ST336732 LW/LC
18.4
4/2
3.6
4.2
0.3
0.4
522–709
51–69
42–58
8
10.3 / 11
13.7 / 13.7
3.5
250
1,200,000 / 0.73%
Seagate’s FC experience makes
it the #1 choice of customers.
Ultra320: Packetization & QAS
Popular capacities; with fewer
moving parts for high reliability
Enhanced with Seagate’s
advanced caching and
command queuing algorithms
Enhanced with Seagate’s
Nonlinear Optimized Formatting
and SAMS
High standard cache
Very low power and cooling
requirements
Sound barrier technology
Over 1,000 Gs with SeaShell™
Highest Reliability Rating
Reliability
The highest reliability rating in the industry is underpinned with proven reliability
leadership based on the first-generation Seagate Cheetah X15. The Cheetah
X15 has led by example, establishing a standard of excellence in reliability
and providing customer confidence.
CRT and RDT Testing: The Cheetah X15 18LP CRT and RDT data is as good as its
contemporary generation 10K and twice as good as the previous generation 10K.
Field Data: The Cheetah X15 18LP AFR data indicates that it’s similar to its
contemporary generation 10K-AFR data.
Reliability and Spindle Speed: Reliability data over previous generations shows
that there is no correlation between increasing spindle speed and reliability.
The Cheetah X15-36LP has the highest reliability rating of any RPM drive in
the industry. The Cheetah X15-36LP incorporates the most advanced elements
of the 3D Defense System, enabling an MTBF of 1,200,000 hours and an
AFR of 0.73 percent.
Power Requirements
The Cheetah X15-36LP has exceptionally low power and cooling requirements,
comparable to 10K-RPM drives. In fact, the Cheetah X15-36LP’s overall maximum average operating current is lower than the 10K’s last two generations.
This is shown in the two Peak Operating Power tables for both SCSI and Fibre
Channel interfaces on page 12.
11
CHEETAH X15-36LP TECHNOLOGY PAPER
Peak Operating Power
Fibre Channel
19
18
17
16
15
14
13
Watts
Watts
Peak Operating Power
SCSI LVD
15K 18LP
15K 36LP
10K 18LP
10K 36LP
10K 73LP
19
18
17
16
15
14
13
15K 18LP
Seagate 15K vs. 10K
15K 36LP
10K 18LP
10K 36LP
10K 73LP
Seagate 15K vs. 10K
In addition, as with other Seagate Enterprise disc drives, the Cheetah X15-36LP
has a temperature sensor that sends a warning in advance to the host when
the temperature exceeds limits either specified by the product manual or by
the customer.
Cheetah X15-36LP Advanced Features
In addition to the fast access time specifications of the Cheetah X15-36LP,
Seagate has designed into this drive advanced technology that ensures this
drive will maintain maximum performance in a variety of applications.
Performance is maximized with the Seagate advanced caching and command
queuing algorithms, superior adjacent-track seek performance, Nonlinear
Optimized Formatting and SAMS, the Seagate Advanced Multidrive System™.
Nonlinear Optimized Formatting
Seagate takes advantage of its superior adjacent track seek performance in
formatting its high-performance drives by using a patent-pending design with
a layout that optimizes the trade-off between one-track seeks and head switches.
It allows for superior sustained transfer rates.
Zero Acceleration Path Technology
Our Seagate-exclusive patent-pending Zero Acceleration Path technology
allows higher track densities by employing algorithms using prewritten track
position information to maximize head track following accuracy. It allows for the
highest 15K-RPM TPI in the industry, which translates into shorter seek
distances and allows for a lower disc and head count, bringing greater reliability.
SAMS
The Seagate Advanced Multidrive System (SAMS), is a set of designs that
keeps performance on track by reducing the effects of rotational vibration (RV)
generated from multiple drives operating simultaneously in a cabinet. RV can
hamper performance by making it difficult for the head to stay on track. When
the head is not on track, the disc must make one complete rotation before the
head can get the data it missed.
12
ECONOMIES OF SPEED
SAMS maintains performance in two ways. First, SAMS helps minimize the
amount of RV energy emitted from the Seagate drive, minimizing the amount
of RV in the cabinet. Second, SAMS reduces the impact that the cabinet’s
RV has on the heads and discs in the Seagate drive. An example of a SAMS
design element is the Cheetah X15-36LP’s Inertia Weights, which add mass
to limit drive-chassis RV energy transfer, maintaining peak performance from
Seagate while competitor performance suffers.
Interfaces
The Cheetah X15-36LP offers Ultra160 SCSI, 2-Gbit-per-second Fibre Channel
and Ultra320 interfaces, which will be available when the infrastructure is in
place. The performance differences gained by the Ultra320 version should be
negligible in all cases except for only huge multidrive arrays servicing applications that require high transfer rates. A single 15K-RPM drive won’t come
close to saturating an Ultra160 channel.
“In conclusion, the X15-36LP delivers mightily on all fronts: the
lowest access times, the highest transfer rates, the highest
WinBench 99 scores, and the highest IOMeter scores that we’ve
yet recorded. Simply amazing. Make no mistake about it, the
Cheetah X15-36LP is the fastest hard drive one can buy for any
and all applications. Period.”
—StorageReview.com, in testing the Seagate
Cheetah X15-36LP, July 19, 2001
Appendix B. Balancing the System
’87
’99
Increase
CPU
1 MIPS
700 MIPS
700×
Memory
100 microsec
100 nanosec
1,000×
Drives
60 msec
6 msec
10×
Source: Compaq TPC Labs
Because of the slow speed of disc drives relative to the other components in
the system, most enterprise solutions require large numbers of disc drives. The
infrastructure to support these discs is a significant cost of the system.
13
Seagate Technology LLC
920 Disc Drive, Scotts Valley, California 95066, USA
Publication Number: TP-579B, Printed in USA