Transcript System Management

Understanding Operating Systems
Fifth Edition
Chapter 12
System Management
Evaluating an Operating System
• To evaluate an OS, you need to understand:
–
–
–
–
Its design goals and history
How it communicates with users
How its resources are managed
What tradeoffs were made to achieve goals
• Operating system’s strengths and weaknesses need
to be weighed in relation to:
• Users of the OS
• Hardware on which the OS will run
• Purpose
Understanding Operating Systems, Fifth Edition
2
Cooperation Among Components
• Performance dependency
– One resource depends on other system resources
• System improvement
– Requires extensive analysis of the needs of the
• System’s resources, requirements, managers, users
• System change results
– Trade one problem for another
• Consider entire system performance
– Not just individual components
Understanding Operating Systems, Fifth Edition
3
Role of Memory Management
• Consider actual operating environment
– Before memory-related changes
• Tradeoff
– Memory use versus CPU overhead
– As the memory management algorithm complexity
increases, the CPU overhead increases, which could
affect overall system performance
• But for some operating systems, adding more
memory can result in a remarkable improvement in
performance
Understanding Operating Systems, Fifth Edition
4
Role of Processor Management
– A multiprogramming system may be desirable if you
wish to increase CPU utilization but this
– Requires synchronization
• Memory manager, processor manager, and I/O devices
– Tradeoffs to consider
• Obtain better CPU utilization versus increased
overhead, slower response time, decreased throughput
Understanding Operating Systems, Fifth Edition
5
Role of Processor Management
(continued)
• Problems
– System could reach saturation point
• If the CPU is fully utilized and still accepting additional
jobs
• Will lead to higher overhead and less time to run
programs
– Under heavy loads
• CPU time required to manage I/O queues could
dramatically increase the time required to run jobs
– With long queues at channels, control units, and I/O
devices, the CPU could be idle waiting for processes
to finish I/O
Understanding Operating Systems, Fifth Edition
6
Role of Device Management
• I/O device utilization improvement techniques
include:
– buffering, rescheduling I/O requests
Tradeoffs involved:
• Increased CPU overhead
• Additional memory space used
Understanding Operating Systems, Fifth Edition
7
Role of Device Management
(continued)
• Buffering
– CPU matches slower I/O device speed (and vice
versa)
– Requires memory space (buffers)
– Tradeoff
• Less multiprogramming versus better I/O device use
• Rescheduling requests
– Seeks to optimize I/O request time by reordering I/O
queues
– Overhead is involved, so the CPU and I/O device
speeds must be weighed against the time it would
take to execute the reordering algorithm
Understanding Operating Systems, Fifth Edition
8
Role of Device Management
(continued)
Understanding Operating Systems, Fifth Edition
9
Role of Device Management
(continued)
• Example: without reordering
– CPU 1 and disk drive A
• Access track 1, track 9, track 1, track 9
• Arm already located at track 1
Understanding Operating Systems, Fifth Edition
10
Role of Device Management
(continued)
• Example: after reordering
– Arm performs both accesses on Track 1 before
traveling Track 9 (35 ms)
Understanding Operating Systems, Fifth Edition
11
Role of Device Management
(continued)
• Reordering requests not always warranted
– Example: CPU 1 and much faster disk drive C
• Without reordering: access time = 5 + 5 + 5 = 15 ms
• With reordering: access time = 5 + 30 = 35 ms
• Reordering algorithm
– Always on or always off
– Requires reconfiguration by system admin to change
– Initial setting
• Determined by evaluating system on average
Understanding Operating Systems, Fifth Edition
12
Role of File Management
• The way in which files are stored on secondary storage can affect
system performance
• Important considerations:
– File organization
• Example: If a file is stored non-contiguously that has several
sections residing in widely separated cylinders of a disk pack,
sequentially accessing all of its records could be slow
• This requires compaction (defragmentation). This takes CPU time
and makes the files unavailable to users during compaction.
– Volume directory location
• Affects retrieval time
• Different schemes offer different flexibility but the tradeoff for file
flexibility is CPU overhead
• File management is closely related to device on which files are
stored
Understanding Operating Systems, Fifth Edition
13
Role of File Management (continued)
• File management related to device where files
stored
Understanding Operating Systems, Fifth Edition
14
Role of Network Management
• Determines message priorities
• Tries to selects most efficient communication paths
over multiple data communication lines
• The network manager allows a network admin to
monitor the use of individual computers and shared
hardware, and ensure software license agreement
compliance
• Simplifies updating data files and programs on
networked computers by coordinating changes
through a server instead of making changes on
individual computers.
Understanding Operating Systems, Fifth Edition
15
Measuring System Performance
• Total system performance is the efficiency with
which computer system meets goals
• System efficiency
– Not easily measured
– Affected by three components
• User programs, operating system programs, hardware
• System performance
– Very subjective
– Difficult to quantify
– When quantifiable
• Not an absolute measure
Understanding Operating Systems, Fifth Edition
16
Measurement Tools
• Measures/metrics of system performance:
–
–
–
–
–
–
–
Throughput
Capacity
Response time
Turnaround time
Resource utilization
Availability
Reliability
Understanding Operating Systems, Fifth Edition
17
Measurement Tools (continued)
• Throughput
• Composite measure
– Indicates system productivity as a whole
– Measured under steady-state conditions
– Example: quantities
• Number of jobs processed per day
• Number of online transactions handled per hour
– Measures work volume handled by system unit
Understanding Operating Systems, Fifth Edition
18
Measurement Tools (continued)
• Throughput bottlenecks
– Capacity
– Maximum throughput level
• Resources saturated
• Processes not passed along
– Main memory over-committed
• Multiprogramming level reaches peak point. Leads to
thrashing (memory manager continuously swaps pages
between main memory and secondary storage and the CPU
does not make much progress in executing jobs because it is
too busy swapping pages).
• Monitored by hardware or software
• Bottleneck detection
– Monitor queues at each resource
Understanding Operating Systems, Fifth Edition
19
Measurement Tools (continued)
• Response time
– Online interactive user
– Interval required to process user request
• From when user presses key to send message until
system indicates receipt of message
• Turnaround time
– Batch job response time
• Time from job submission until output returned to user
Understanding Operating Systems, Fifth Edition
20
Measurement Tools (continued)
• Dependencies: Whether in an online or batch
system, this measure (response time or turnaround
time) depends on both the:
– Workload handled by system at time of request
– Type of job or request being submitted
• Include
– Average values and variance
Understanding Operating Systems, Fifth Edition
21
Measurement Tools (continued)
• Resource utilization
– How much unit contributing to overall operation
– Percentage of time resource actually in use
• Example: CPU busy 60 percent of time?
– Helps analyst determine
• Whether there is balance among system units
• Whether the system is I/O-bound or CPU-bound
Understanding Operating Systems, Fifth Edition
22
Measurement Tools (continued)
• Availability
– Indicates likelihood a resource will be available for
use when a user needs it
• Influenced by:
– Mean time between failures (MTBF)
• Average time unit is operational it before breaks down
– Mean time to repair (MTTR)
• Average time needed to fix failed unit and put back in
service
MTBF
Availabili ty(A) 
Availability (A) = MTBF  MTTR
Understanding Operating Systems, Fifth Edition
23
Measurement Tools (continued)
• Reliability
– Measures probability unit will not fail during given time
period
– Function of MTBF
R(t )  e
 (1 MTBF )( t )
Understanding Operating Systems, Fifth Edition
24
Measurement Tools (continued)
• Performance measures
– Avoid taking in isolation from system workload
• Overall system performance
– Varies with time
– Important to define actual working environment
• Before making generalizations
Understanding Operating Systems, Fifth Edition
25
Feedback Loops
• Monitor system resource utilization for adjustments
– Prevents processor time spent on overhead
– More time executing jobs
• Feedback loop types
– Negative feedback loop
– Positive feedback loop
Understanding Operating Systems, Fifth Edition
26
Feedback Loops (continued)
• Negative feedback loop
– Process arrival rate decreased when system too
congested
• Stabilized system
• Queue lengths close to estimated mean values
• Positive feedback loop
– Arrival rate increased when system underutilized
• Paged virtual memory systems use this
• Implementation more difficult (than negative loops)
Understanding Operating Systems, Fifth Edition
27
Feedback Loops (continued)
Understanding Operating Systems, Fifth Edition
28
Feedback Loops (continued)
Understanding Operating Systems, Fifth Edition
29
Monitoring
• Hardware monitors
– More expensive
– Minimum impact on system
• Outside and attached electronically
– Examples: counters, clocks, comparative elements
• Software monitors
– Relatively inexpensive
– Distortion of analysis results
• Software monitor becomes part of system
– Developed for each specific system
– Difficult to move from system to system
Understanding Operating Systems, Fifth Edition
30
Monitoring (continued)
• Early systems performance measurements
– Monitored CPU speed
• Today’s measurements
– Other hardware units, operating system, compilers,
other system software
• Measurements made in variety of ways
– Real programs: production programs
• Run with different configurations of CPUs, operating
systems, other components
• Results called benchmarks
– Using simulation models
Understanding Operating Systems, Fifth Edition
31
Monitoring (continued)
• Benchmarks
– Demonstrate specific advantages
• New CPU, operating system, compiler, or piece of
hardware
– Useful when comparing systems experiencing
extensive changes
– Results dependent upon:
• System’s workload
• System’s design and implementation
• Specific requirements of applications loaded on system
Understanding Operating Systems, Fifth Edition
32