Docsity
Docsity

Prepare for your exams
Prepare for your exams

Study with the several resources on Docsity


Earn points to download
Earn points to download

Earn points by helping other students or get them with a premium plan


Guidelines and tips
Guidelines and tips

Performance Metrics - Computer Systems Performance Evaluation - Lecture Slides, Slides of Computer Science

Some concept of Computer Systems Performance Evaluation are Measurement and Statistics, Performance Evaluation, Performance Metrics, Queueing Lingo, Software Performance Engineering. Main points of this lecture are: Performance Metrics, Equations, Performance, Later Section, System, Departures, Arrivals, Observed, Arrivals Observed, Departures Observed

Typology: Slides

2012/2013
On special offer
30 Points
Discount

Limited-time offer


Uploaded on 04/27/2013

divyaa
divyaa 🇮🇳

4.4

(59)

71 documents

1 / 12

Toggle sidebar

This page cannot be seen from the preview

Don't miss anything!

bg1
Performance
Engineering
Performance Metrics
Docsity.com
pf3
pf4
pf5
pf8
pf9
pfa
Discount

On special offer

Partial preview of the text

Download Performance Metrics - Computer Systems Performance Evaluation - Lecture Slides and more Slides Computer Science in PDF only on Docsity!

Performance

Engineering

Performance Metrics

All About Numbers

Arrivals ------> RESOURCE ------> Departures

B is the length of time that the resource was observed to be BUSY.

Note that the pictures may be equivalent; a system may have only one resource. A request

is presented to a system ( in which case it might be called a "job") or is presented to a

particular resource.

Metric Description Formula Units

Arrival Rate How many requests arrive per time.

Y = A / T (^) Requests/second

Throughput (Departure Rate)

How many requests complete per time.

X = C / T Requests/second

Utilization Fraction of time the resource is in use.

U = B / T (^) Number in range 0 - 1.

Service Requirement

Time needed by a resource to handle a request.

S = B / C Seconds

Requests in system Average number of requests in the system.

N = W / T Dimensionless

Residence time Average time a request spends in the system.

R = W / C Seconds

UTILIZATION LAW U = X S

Proof: B / T = C / T * B / C

Example:

The server of dinners in the Cafeteria keeps busy 75% of the time between 12:00 and 1:00. During this time 90 people are served a dinner. How long does it take to serve each customer?

Requests

Time ( seconds )

Y = A / T Arrival rate =

X = C / T Throughput = R = W / C Residence Time = U = B / T Utilization = S = B / C Service time/request = N = W / T Average number of =

requests in the system.

T Time interval =

A Arrivals =

C Completions =

B Time spent servicing =

requests

W Accumulated time =

within system

Example:

One day I was at the Weston Toll Booth on the Mass Pike. There were 10 collectors

working; when I joined the line, there were 8 cars ahead of me, and when my turn

came it took me 12 seconds to pay my toll.

What information can be derived from the above paragraph? What information can NOT

be deduced?

Example:

Recently in "Newsweek" it was reported that the current U.S. divorce rate is 4.6 per 1000

adults. It was also reported that half of all adults are currently married.

Develop a simple model/picture in order to make some sense of these numbers.

Some more practice problems:

Example:

Packets arrive at a gateway at a rate of 125 packets/second. The gateway takes an

average of 2 milliseconds to forward them.

What is the Throughput, Service Time, and Utilization.

Example:

A system has a CPU, Disk A, and Disk B. The throughputs of these devices are XCPU = 35.48,

XA = 15.70, X B = 19.6. We also know that the Queue lengths are QCPU = 8.88, Q A =

3.19, QB = 1.40.

What are the device response times for these devices?

Example:

On the campus, there are a number of graduate students who have varying degree and

course requirements. Build a model using duration of time until their degree, as well as

the fraction of time in school.

What is the service time, residence time, of a graduate student? What is the throughput

of the CS Graduate School?