Transcript Unix Systems Performance Tuning

Unix Systems Performance
Tuning
Project of COSC 513
Name: Qinghui Mu
Instructor: Prof. Anvari
Performance Tuning
• Performance tuning is a process of
observing the operations of a system and
making adjustments to different
components based on those observations.
As a result we get an overall more efficient
system.
What is system performance?
• In a broader sense, system performance
refers to how well the computer resources
accomplish the work they are designed to
do. The performance of any computer
system may be defined by two criteria:
Response time
Throughput
Key factors
● Hardware
● Operating system
● Application software
● Users
● Changes over time
Managing system performance
• Monitoring usage of system resources
• Selecting tools to measure system performance
• Diagnosing problems from the results of
measurement
• Tuning the operating system and application
parameters
• Upgrading the hardware resources of the
system
• Planning for the optimal performance
Bottlenecks
• A resource is a bottleneck, if the size of a
request exceeds the available resource. In
other words, a bottleneck is a limitation of
system performance due to the
inadequacy of a hardware or software
component, or of the system’s
organization.
Two ways to solve a bottleneck
• increasing the size of available resource
• decreasing the size of the request
Guidelines in tuning a real system
•
•
•
•
•
•
Never tune at random
Tune one area at a time
Change only one parameter at a time
Always use at least two tools
Experience is the best tool
Know when to say stop
Tuning steps
Step 1: Get current measurements
Step 2: Find current bottleneck
Step 3: Remove it
Step 4: Make new measurements
Step 5: If required performance is not
achieved go to step 1
Step 6: Stop for the time being.
System Resources
•
•
•
•
•
•
CPU
Memory
Disks and disk controllers
Busses
Networking
Operating system
Test Tools
•
•
•
•
Tools for CPU Utilization
Tools for Memory Utilization
Tools for Disk Utilization
Tools for Network Utilization
Tools for CPU Utilization - 1
• mpstat: mpstat reports per-processor
statistics in tabular form. Each row of the
table represents the activity of one
processor. The first table summarizes all
activity since boot; each subsequent table
summarizes activity for the preceding
interval. All values are rates (events per
second) unless otherwise noted.
Tools for CPU Utilization - 2
• iostat: The iostat utility iteratively reports
terminal, disk and other I/O activities, as
well as CPU utilisation. The –c option of
iostat is useful to report the percentage of
time the CPU has spent in user mode, in
system mode, waiting for I/O and idling.
Tools for CPU Utilization - 3
• Sar: System Activity reports are also a
good tool to see the resources utilization
of a system. sar can be used to analyze all
the system resources. The –u option gives
the CPU utilization. Again constant high
percentage under %usr column means
system is CPU bound.
Tools for CPU Utilization - 4
• Top: Top is a good free-ware tool available
on net. This tool shows the top processes
on a system using system resources. This
includes total system-wide memory and
CPU usage and break-up of these
resources being used by every process.
Tools for CPU Utilization – 5
• Use vmstat 5 to confirm the CPU bottleneck.
Things to look for is procs column, which shows
how many processes are running, waiting for I/O
or other activity like lock or sleep. The r column
includes processes that are running on the CPU.
r>0 means that adding CPU power can be
profitable. Where as CPU column has similar
columns us sy and id and has same meaning as
in system activity report (sar). If there is no idle
time that means system is CPU bound.
Tools for Memory Utilization - 1
• vmstat: vmstat can be used to see the
virtual memory usage. If there is any
paging or swapping being done by the
system. This is a good tool to see if a
system is memory bound.
Tools for Memory Utilization - 2
• sar: System Activity Report can be used to see if
swapping is being done by the system. Use sar
–w to see if there is any swapping going on.
Some swapping and paging is normal but too
many swap-in and swap-outs can be an
indicator of low memory. sar –g also reports
number of pages per second scanned by the
paging daemon (pgscan/s). If this value is high,
the page daemon is spending a lot of time
checking for the free memory.
Tools for Memory Utilization - 3
• The best RAM shortage indicator is the
scan rate (sr) output from vmastat. A scan
rate of above 200 pages per second for
long periods of time indicates memory
shortage. Excessive paging (pi and po) is
also an indicator of low memory.
Tools for Disk Utilization - 1
• Use iostat –x . The wait column shows the
queue length, that is, the average number of
transactions waiting for service. The %w column
shows the percentage of time the queue is not
empty; that is, the percent of time there are
transactions waiting for service. If a system
spends a good percentage of time for I/O, this
could indicate a disk bottleneck. High %b and
svc_t are also an indication of disk I/O
bottleneck. Careful configuration, thatis,
spreading the load across devices can improve
response time.
Tools for Disk Utilization - 2
• vmstat
• Whenever there are blocked process, all
CPU idle time is treated as wait for I/O
time. If number of blocked processes; that
is, the number of process under b column,
exceed the number of processes in run
queue, under r column. This shows disk
sub-system bottleneck.
Tools for Network Utilization - 1
• ntop: This is a tool with the similar
interface as top. It provides detailed
network utilization for each computer on
local network. We can write filter to see
traffic for specific hosts. Detailed network
traffic statistics available from this tool.
Tools for Network Utilization - 2
• netstat: netstat with –i option gives the
incoming packets, outgoing packets,
collisions and errors for each network
interface on the system. So this can also
be used to see the traffic being generated
on the system.
Viewing and setting Tuning
Parameters in Solaris
• sysdef: provides a list of approximately 60 of
the most commonly used tuning parameters
• adb: The kernel degugger, enables to look and
alter the valuesof tuning parameters. adb and
crash utility allow the inspection of the running
system
 ndd: Used to look at and change the settings of
network device driver and protocol stack.