Paper by: Chris Ruemmler and John Wikes
Presentation by: Timothy Goldberg, Daniel Sink, Erin
Collins, and Tony Luaders
 Disk Drive performance improvements at 7-10%
 Compared to microprocessors at 40-60% or disk storage
capacities at 60-80% (annually)
 Simulation models to compare alternative approaches
 High quality disk drive model
 Error factor 14 times smaller
 Introduction
 Characteristics of Modern Disk Drives
 Recording Components
 Positioning Components
 Disk Controller
 Modeling Disk Drives
Characteristics of Modern Disk
 Non-removable magnetic disk drives
 Contain a mechanism and controller
 Recording Components: rotation disks and heads
 Positioning Components: moves heads into correct
position with track-following system
 Emphasis on features that could be important when
creating a disk drive model
Recording Components
 Smaller disks:
 Less surface area for data
 Less power consumption
 Can spin faster
 Smaller seek distances
 Increased storage density:
 Better linear recording density, maximum rate of flux changes
 Packing separate tracks of data more closely together
 May contain from 1 to 12 platters
 Stack rotates in lockstep
Recording Components
 Spindle rotation speed:
 Higher spin speed increases transfer rates, shortens rotation
 Higher power consumption, requires better bearings
 Each platter surface has a disk head
 Responsible for recording (writing)
 And sensing (reading) magnetic flux variation
 Single Read-Write data channel
 Can be switched between the heads
 Responsible for encoding and decoding data stream into or
from a series of magnetic phase changes stored on the disk
Disk Drive
Positioning Components
 Data surfaces are set up to store data in tracks
 Modern disks have about 2,000 cylinders and are 3.5
 Cylinder is a single stack of tracks at a common distance from
the spindle
 To access the data stored on a track, the disk arms must
rotate all the disks to get the desired track to the disk head.
 This system ensures that the track is reached even with
 External vibrations, shocks, and disk flaws (non circular
 The speed of head movement
 Faster seeking requires more power
 Half the seek time requires 4x power
 Seek is composed of:
 Speedup (arm moves until at half seek distance)
 Coast (for long seeks, max velocity)
 Slowdown (rest close to desired track)
 Settle (puts disk head on desired location)
Track Following
 Fine-tuning the head position at the end of the seek
and keeping the head on the desired track
 Determines if head is correctly aligned by using
positioning information on the disk at manufacturing
 Performs head switches
 When the controller switches its data channel from one
surface to the next in the same cylinder
Data layout
 A disk appears to its client computer as a linear vector of
addressable blocks which are mapped to physical sectors
on the disk.
 Using this method, the disk can hide bad sectors and do
low-level performance optimizations.
 Zoning: tracks are longer at the outside of a platter than at
the inside.
 Maximize storage capacity
 Track skewing: faster sequential access across track
 Allows data to be read or written at nearly full media speed
 Sparing: stores a list of flaws in the desk surface to be
The Disk Controller
 Mediates access to the mechanism
 Runs the track-following system
 Transfers data between the disk drive and the client
 Manages an embedded cache
caching of requests
 Speed-matching buffer can be extended to include some
form of caching for both reads and writes.
 Caches in disk drives are relatively small because of space
 Read-ahead: faster than seeking if the cache gets a hit
 Write caching: saves cache information
 Cache is volatile, losing its contents if power to the drive is
 Command queuing: allows for multiple outstanding
requests at the same time
 Disk controller determines the best execution order, subject
to additional host constraints.
Modeling Disk Drives
The Simulator
• Based in C++ using a version of the AT&T tasking
• The Basic ideas are readily applicable to other
simulation environments
• The disk drive is modeled as two tasks and some
additional control structures
• Task one models the mechanism, including the head
and platter (rotation) positions.
• Task two, the direct memory access engine (DMA),
models the SCSI bus interface and its transfer engine.
The Simulator
• The cache object buffers requests between two tasks
and is used to manage the asynchronous interactions
between the bus interface and the disk mechanism
• The simulator can process about 2,000 I/Os per
second on an HP9000 Series 800 Model H50 system
• This allows 1 million requests to be serviced in
approximately 10 minutes
• Took week long samples from a longer trace series of
HP-UX (Unix) computer systems.
• A metric to evaluate the models used a time
distribution curve for the real drive and the model
output and use the root mean square of the
horizontal distance between these two curves.
No modeling
•Uses a constant fixed
time for each I/O
•A demerit factor that is
35% of the average I/O
•This model is not good
A simple model
•A better model requires:
•A seek time linear with the
•No head-settle effects or
head-switching costs
•A rotational delay
•A fixed controller overhead
•A transfer time linear with
the length of request
•demerit of 15% of a mean
I/O time
Modeling head-positioning
•Determined which
track and cylinder the
request started on and
where it ended
•Added a fix cost of 2.5
ms for each head and
track switch
•Demerit of 6.2% of a
mean I/O time
Modeling rotation position
•Calculate rotational
latency by keeping track
of rotational position of
the disk
•Account for spare
•A demerit of 2.6% of
mean I/O time
Modeling data caching
•Uses both read-ahead
and immediate
•Large disparity due to
•50% of request are
completed in 3ms or less
•Demerit of 112% is not
Modeling data caching
•Added aggressive readahead and immediate
reporting to the model
•Demerit is now only
5.7% of the mean I/O
Model summary
•Careful modeling is neither
too difficult nor too costly
•A good model needs careful
calibration and tuning
•These features and others
may become particularly
important when a workload
has large data transfers

similar documents