Compaq ProLiant 1000 I/O Performance Tuning of Compaq Servers - Page 15

RAID levels and performance

Page 15 highlights

I/O Performance Tuning of Compaq Servers 15 All of these delays are associated with positioning the drive head above the requested data. The delay caused by the swinging of the head to any given cylinder is called seek time. In some cases, the head may only have to seek from one track to the next; other cases may require the head to move across the entire data area. By averaging the time required for the head to move between any two random tracks, we can obtain average seek time. Lower average seek times provide faster data access. Once the drive head has completed the seek to the requested cylinder, it is likely that the drive will have to wait for the platter to rotate the data under the drive head. The delay introduced by waiting for the data to rotate to the head is called rotational latency. Since the likelihood of having to complete a full rotation in order to position the head over the data is just as great as having the head land immediately on the requested data, we can define the average latency as the time required to rotate the disk through one half of a revolution. By combining the two physical performance factors, we obtain the measurement of average access time. Average access time is simply the average seek time added to the average latency. The average, non-cached request that the drives must perform will be subject to delays equal to this average access time. While there is no way to eliminate the physical delays of hard drives, there are a few tools that can help reduce their effect. Drive technology is constantly improving, and doing so at an ever-increasing pace. By using newer drives, you can take advantage of this technology allowing • Faster and more accurate head movement - New servomotors allow faster placement of the drive head, thereby reducing the average seek time. • Higher velocity platters. - Higher velocity platters reduce latency by moving more data under the head faster. Higher speed platters also enable the drive to read more data in shorter periods of time. • Greater data density - Combined with faster platters, greater data density allows the drive to retrieve more data per track and more tracks per surface. The table below shows comparative measures of physical delays in various drives. Table 5. Relative Performance of Drive Technologies Rotational Speed 3,600 rpm 5,400 rpm 7,200 rpm 10,000 rpm Average Latency 8.3 ms 5.5 ms 4.2 ms 3 ms Average Seek 13 ms 11 ms 7.1 ms 5.4 ms Average Access Time 21.3 ms 16.5 ms 11.3 ms 8.4 ms RAID levels and performance Another tool to help reduce data access delays is RAID technology. When implemented wisely, RAID can provide phenomenal performance for your drive system and at the same time protect your server from data loss. Published in 1987, RAID combines many smaller disks to yield performance that is greater than a single drive of the same capacity. In terms of seek time and latency, RAID cannot change the physical characteristics of the drive. What RAID can do is distribute the workload so that each drive in an array has fewer requests to ECG044.0399

  • 1
  • 2
  • 3
  • 4
  • 5
  • 6
  • 7
  • 8
  • 9
  • 10
  • 11
  • 12
  • 13
  • 14
  • 15
  • 16
  • 17
  • 18
  • 19
  • 20
  • 21
  • 22
  • 23
  • 24
  • 25
  • 26
  • 27
  • 28
  • 29
  • 30
  • 31
  • 32
  • 33
  • 34
  • 35
  • 36
  • 37
  • 38

I/O Performance Tuning of Compaq Servers
15
ECG044.0399
All of these delays are associated with positioning the drive head above the requested data. The
delay caused by the swinging of the head to any given cylinder is called
seek time
. In some cases,
the head may only have to seek from one track to the next; other cases may require the head to
move across the entire data area. By averaging the time required for the head to move between
any two random tracks, we can obtain
average seek time
. Lower average seek times provide
faster data access.
Once the drive head has completed the seek to the requested cylinder, it is likely that the drive
will have to wait for the platter to rotate the data under the drive head. The delay introduced by
waiting for the data to rotate to the head is called
rotational latency
. Since the likelihood of
having to complete a full rotation in order to position the head over the data is just as great as
having the head land immediately on the requested data, we can define the
average latency
as the
time required to rotate the disk through one half of a revolution.
By combining the two physical performance factors, we obtain the measurement of
average
access time.
Average access time is simply the average seek time added to the average latency.
The average, non-cached request that the drives must perform will be subject to delays equal to
this average access time. While there is no way to eliminate the physical delays of hard drives,
there are a few tools that can help reduce their effect.
Drive technology is constantly improving, and doing so at an ever-increasing pace. By using
newer drives, you can take advantage of this technology allowing
Faster and more accurate head movement
– New servomotors allow faster placement of the
drive head, thereby reducing the average seek time.
Higher velocity platters. – Higher velocity platters reduce latency by moving more data under
the head faster. Higher speed platters also enable the drive to read more data in shorter
periods of time.
Greater data density – Combined with faster platters, greater data density allows the drive to
retrieve more data per track and more tracks per surface.
The table below shows comparative measures of physical delays in various drives.
Table 5.
Relative Performance of Drive Technologies
Rotational Speed
Average Latency
Average Seek
Average Access Time
3,600 rpm
8.3 ms
13 ms
21.3 ms
5,400 rpm
5.5 ms
11 ms
16.5 ms
7,200 rpm
4.2 ms
7.1 ms
11.3 ms
10,000 rpm
3 ms
5.4 ms
8.4 ms
RAID levels and performance
Another tool to help reduce data access delays is RAID technology. When implemented wisely,
RAID can provide phenomenal performance for your drive system and at the same time protect
your server from data loss. Published in 1987, RAID combines many smaller disks to yield
performance that is greater than a single drive of the same capacity.
In terms of seek time and latency, RAID cannot change the physical characteristics of the drive.
What RAID can do is distribute the workload so that each drive in an array has fewer requests to