Aperiodic and Sporadic Jobs. Scheduling Aperiodic and Sporadic Jobs

Similar documents
Mixing Real-Time and Non-Real-Time. CSCE 990: Real-Time Systems. Steve Goddard.

Advanced Queueing Theory. M/G/1 Queueing Systems

Consider a system of 2 simultaneous first order linear equations

Chapter 7 Stead St y- ate Errors

Final Exam : Solutions

Lecture 4 : Backpropagation Algorithm. Prof. Seul Jung ( Intelligent Systems and Emotional Engineering Laboratory) Chungnam National University

Conventional Hot-Wire Anemometer

Ma/CS 6a Class 15: Flows and Bipartite Graphs

Engineering Circuit Analysis 8th Edition Chapter Nine Exercise Solutions

The Variance-Covariance Matrix

8. Queueing systems. Contents. Simple teletraffic model. Pure queueing system

Chapter 9 Transient Response

Poisson process Markov process

10.5 Linear Viscoelasticity and the Laplace Transform

Summary: Solving a Homogeneous System of Two Linear First Order Equations in Two Unknowns

Chapter 13 Laplace Transform Analysis

Frequency Response. Response of an LTI System to Eigenfunction

Homework: Introduction to Motion

Economics 302 (Sec. 001) Intermediate Macroeconomic Theory and Policy (Spring 2011) 3/28/2012. UW Madison

Wave Superposition Principle

Lecture 1: Numerical Integration The Trapezoidal and Simpson s Rule

Safety and Reliability of Embedded Systems. (Sicherheit und Zuverlässigkeit eingebetteter Systeme) Stochastic Reliability Analysis

( r) E (r) Phasor. Function of space only. Fourier series Synthesis equations. Sinusoidal EM Waves. For complex periodic signals

4.1 The Uniform Distribution Def n: A c.r.v. X has a continuous uniform distribution on [a, b] when its pdf is = 1 a x b

Safety and Reliability of Embedded Systems. (Sicherheit und Zuverlässigkeit eingebetteter Systeme) Stochastic Reliability Analysis

CIVL 8/ D Boundary Value Problems - Triangular Elements (T6) 1/8

Institute of Actuaries of India

Heisenberg Model. Sayed Mohammad Mahdi Sadrnezhaad. Supervisor: Prof. Abdollah Langari

14.02 Principles of Macroeconomics Fall 2005 Quiz 3 Solutions

Chapter 5 The Laplace Transform. x(t) input y(t) output Dynamic System

2. The Laplace Transform

Boosting and Ensemble Methods

9. Simple Rules for Monetary Policy

t=0 t>0: + vr - i dvc Continuation

AR(1) Process. The first-order autoregressive process, AR(1) is. where e t is WN(0, σ 2 )

CS 268: Packet Scheduling

(heat loss divided by total enthalpy flux) is of the order of 8-16 times

(,,, ) (,,, ). In addition, there are three other consumers, -2, -1, and 0. Consumer -2 has the utility function

CHAPTER CHAPTER14. Expectations: The Basic Tools. Prepared by: Fernando Quijano and Yvonn Quijano

Time to Recruitment for a Single Grade Manpower System with Two Thresholds, Different Epochs for Inter-Decisions and Exits Having Correlated Wastages

Why Laplace transforms?

Supplementary Figure 1. Experiment and simulation with finite qudit. anharmonicity. (a), Experimental data taken after a 60 ns three-tone pulse.

DEPARTMENT OF ELECTRICAL &ELECTRONICS ENGINEERING SIGNALS AND SYSTEMS. Assoc. Prof. Dr. Burak Kelleci. Spring 2018

UNIT #5 EXPONENTIAL AND LOGARITHMIC FUNCTIONS

EXERCISE - 01 CHECK YOUR GRASP

Microscopic Flow Characteristics Time Headway - Distribution

Two-Dimensional Quantum Harmonic Oscillator

Real-Time Scheduling for Event-Triggered and Time-Triggered Flows in Industrial Wireless Sensor-Actuator Networks

The Mathematics of Harmonic Oscillators

cycle that does not cross any edges (including its own), then it has at least

Grand Canonical Ensemble

NAME: ANSWER KEY DATE: PERIOD. DIRECTIONS: MULTIPLE CHOICE. Choose the letter of the correct answer.

A. Inventory model. Why are we interested in it? What do we really study in such cases.

Gauge Theories. Elementary Particle Physics Strong Interaction Fenomenology. Diego Bettoni Academic year

PFC Predictive Functional Control

1 Finite Automata and Regular Expressions

On the Speed of Heat Wave. Mihály Makai

innovations shocks white noise

u x v x dx u x v x v x u x dx d u x v x u x v x dx u x v x dx Integration by Parts Formula

Double Slits in Space and Time

Problem 1: Consider the following stationary data generation process for a random variable y t. e t ~ N(0,1) i.i.d.

EE243 Advanced Electromagnetic Theory Lec # 10: Poynting s Theorem, Time- Harmonic EM Fields

Neutron electric dipole moment on the lattice

Vertical Sound Waves

Laplace Transforms recap for ccts

A L A BA M A L A W R E V IE W

14.02 Principles of Macroeconomics Problem Set 5 Fall 2005

Theoretical Seismology

Discussion 06 Solutions

Cmd> data<-matread("hwprobs.dat","exmpl8.2") This is the data shown in Table 8.5, the copper and diet factors.

Chapter 8 Theories of Systems

Section 6.1. Question: 2. Let H be a subgroup of a group G. Then H operates on G by left multiplication. Describe the orbits for this operation.

Chapter 12 Introduction To The Laplace Transform

Jonathan Turner Exam 2-10/28/03

2.1. Differential Equations and Solutions #3, 4, 17, 20, 24, 35

Alpha and beta decay equation practice

5. An object moving along an x-coordinate axis with its scale measured in meters has a velocity of 6t

COHORT MBA. Exponential function. MATH review (part2) by Lucian Mitroiu. The LOG and EXP functions. Properties: e e. lim.

SOLUTIONS. 1. Consider two continuous random variables X and Y with joint p.d.f. f ( x, y ) = = = 15. Stepanov Dalpiaz

(1) Then we could wave our hands over this and it would become:

FAULT TOLERANT SYSTEMS

Chapter 8: Electron Configurations and Periodicity

(ΔM s ) > (Δ M D ) PARITY CONDITIONS IN INTERNATIONAL FINANCE AND CURRENCY FORECASTING INFLATION ARBITRAGE AND THE LAW OF ONE PRICE

Transfer function and the Laplace transformation

Charging of capacitor through inductor and resistor

Some Transcendental Elements in Positive Characteristic

Economics 201b Spring 2010 Solutions to Problem Set 3 John Zhu

Random Access Techniques: ALOHA (cont.)

P a g e 5 1 of R e p o r t P B 4 / 0 9

Title: Long-Term Fiscal Indicators: Sustainability versus Terminal Debt Constraints. 1

Control Systems. Transient and Steady State Response.

P a g e 3 6 of R e p o r t P B 4 / 0 9

Well-publicized worms Worm propagation curve Scanning strategies (uniform, permutation, hitlist, subnet) Three factors define worm spread:

CHAPTER 33: PARTICLE PHYSICS

Lucas Test is based on Euler s theorem which states that if n is any integer and a is coprime to n, then a φ(n) 1modn.

46. Let y = ln r. Then dy = dr, and so. = [ sin (ln r) cos (ln r)

FL/VAL ~RA1::1. Professor INTERVI of. Professor It Fr recru. sor Social,, first of all, was. Sys SDC? Yes, as a. was a. assumee.

Wave Phenomena Physics 15c

The Matrix Exponential

Instructors Solution for Assignment 3 Chapter 3: Time Domain Analysis of LTIC Systems

Transcription:

CPSC-663: Ral-Tm Sym Arodc and Soradc Job Schdulng Arodc and Soradc Job Dfnon Comaron o radonal chdulng of aynchronou vn Pollng Srvr Dfrrabl Srvr Soradc Srvr Gnralzd Procor Sharng Conan Ulzaon Srvr Toal Bandwdh Srvr Prmv Wghd Far Quung Schdulng Arodc and Soradc Job Whn varaon n nr-rla m and xcuon m ar mall: can ra a a rodc a T=(, ), and chdul accordngly. Wha abou oradc job? can arrv a any m xcuon m vary wdly dadln ar unown a ror? Gvn: n rodc a T,, T = (, ),, T n rory-drvn chdulng algorhm W wan o drmn whn o xcu arodc and oradc job,.., oradc job: accanc chdulng of accd job arodcjob: chdul job o coml ASAP.

CPSC-663: Ral-Tm Sym Arodc and Soradc Job Prory Quu for Prodc/Soradc/Arodc Job rjc Soradc Job Accanc T Prodc Job Procor Arodc Job Schdulng Algorhm, Ovrvw Schdulng Arodc (Soradc) Job: Non-Ral-Tm Sym Ral-Tm Sym Bacground Slac Salng Inrru-Drvn Polld Bandwdh Prrvng 2

CPSC-663: Ral-Tm Sym Arodc and Soradc Job Bacground/Inrru-Drvn v. Slac Salng Bacground: Arodc job quu ha alway low rory among all quu. Prodc a and accd job alway md dadln. Sml o mlmn. Excuon of arodc job may b unduly dlayd. Inrru-Drvn: Ron m a hor a obl. Prodc a may m om dadln. Slac Salng: Poon xcuon of rodc a only whn af o do o: Wll-ud for cloc-drvn nvronmn. Wha abou rory-drvn nvronmn? (qu comlcad) Examl Bacground: T T 2 = = ( 3, ) ( 0, 4 ) A : r = 0., = 2. Inrru-Drvn: Slac Salng: 3

CPSC-663: Ral-Tm Sym Arodc and Soradc Job Polld Excuon v. Bandwdh Prrvng Srvr Pollng rvr (, ): chduld a rodc a. : Pollr rady for xcuon vry m un. : Ur bound on xcuon m. Trmnology: (Excuon) budg: Rlnhmn: budg o a bgnnng of rod. Pollr conum budg a ra whl xcung arodc job. Pollr xhau budg whnvr ollr fnd arodc quu my. Whnvr h budg xhaud, h chdulr rmov h ollr from rodc quu unl rlnhd. Bandwdh-rrvng rvr algorhm: Imrov uon ollng aroach U rodc rvr Ar dfnd by conumon and rlnhmn rul. Examl: Pollng Srvr Ra-Monoonc: PS=(3, ) A : r = 2.8, =.7 T =(φ=2, 3.5,.5) T 2 =(φ=0, 6.5, 0.5) budg 4

CPSC-663: Ral-Tm Sym Arodc and Soradc Job Dfrrabl Srvr Rul: Conumon: Excuon budg conumd only whn rvr xcu. Rlnhmn: Excuon budg of rvr o a ach mull of. Prrv budg whn no arodc job rady. Any budg hld ror o rlnhmn lo (no cumulaon). Examl: Dfrrabl Srvr wh RM Ra-Monoonc: DS=(3, ) A : r = 2.8, =.7 T =(φ=2, 3.5,.5) T 2 =(φ=0, 6.5, 0.5) budg 5

CPSC-663: Ral-Tm Sym Arodc and Soradc Job Examl: Dfrrabl Srvr wh EDF EDF: DS=(3,) A : r = 2.8, =.7 T =(φ=2,3.5,.5) T 2 =(φ=0,6.5,0.5) budg Combnaon of Dfrabl Srvr wh Bacground Srvr DS=(3,) A : r = 2.8, =.7 rv n bacground! T =(φ=2,3.5,.5) T 2 =(φ=0,6.5,0.5) budg 6

CPSC-663: Ral-Tm Sym Arodc and Soradc Job Why no Incra h Budg? DS=(3,) T =(3.5,.5) T 2 =(6.5,0.5) Schdulably for Sac-Prory Sym (DS ha hgh rory) Lmma: In a ac-rory rodc ym wh D <=, wh a dfrabl rvr T DS (, ) wh hgh rory, a crcal nan for T han whn: () r,c = 0 for om job J,c n T. (2) job of hghr-rory a ar rlad a m 0. (3) budg of (bacloggd) rvr a m 0. (4) nx rlnhmn m 0. Inuvly: Low-rory a uffr from a bac-o-bac h by h dfrabl rvr. 7

CPSC-663: Ral-Tm Sym Arodc and Soradc Job 8 Tm-Dmand Analy Schdulabl Ulzaon: Gnrally, no nown chdulabl ulzaon. Only xcon: < < 2 < < n < 2 =D ra-monoonc chdulng n > For h ca, h chdulabl ulzaon w = = ) ( = 2 ) ( ) ( / n DS RM u u n n U Dfrabl Srvr and Arbrary Sac Prory Problm: Any budg ha no conumd a nd of rvr rod lo. Maxmum amoun of m DS can conum dnd on Rla m of all rodc job (wh rc o rlnhmn m) Excuon m of all a. Ur bound on m dmand for lowr-rory a han DS: Mull dfrabl rvr: Tm dmand for a wh rory lowr han m DS : w = ) ( q m q q q w = =,,, ) (

CPSC-663: Ral-Tm Sym Arodc and Soradc Job Ung h Schdulabl Ulzaon: Aum ha T ha lowr rory han rvr. T DS (, ) bhav l a rodc a (, ), xc ha may xcu for a mo addonal m un durng h nrval (r,c,r,c D ). = u u b U X ( ) Examl: chdulng algorhm T T T T DS 2 3 ( 3, 06. ) ( 4, 08. ) ( 5, 05. ) ( 7, 4. ) T : T 2 : T 3 : no affcd by T DS 3 2 = = u u = 0.66 0.7797 = U 2 u u = 0.843 0.757 = U 3 no! RM (3) RM (4) Schdulably for Dadln-Drvn Sym Lmma: A rodc a T n a ym of n ndndn, rmv rodc a chdulabl wh a DS wh rod, xcuon m, and ulzaon u, accordng o h EDF algorhm f n = mn( D, u ) D 9

CPSC-663: Ral-Tm Sym Arodc and Soradc Job 0 Proof Proof: L b h dadln of om Job J,c. L - b h la on n m bfor whr hr rocor dl, or wa xcung a lowr-rory a (dadln afr ).... - - If J c m dadln a m, oal amoun of rocor m conumd by Dfrrabl Srvr durng nrval ( -, ] boundd by : J c m dadln hr! Proof (II) Proof: Tm conumd by dfrrabl rvr: Tm conumd by Ta T : W ud floor nad of clng bcau la nvocaon ha dadln afr. W m dadln f w don hav nough m o fnh by m : Dvd by ( - ) and go from hr. ) ( ) ( DS u w = = w / ) ( ) ( = = < n u ) ( ) (

CPSC-663: Ral-Tm Sym Arodc and Soradc Job Soradc Srvr Problm wh Dfrabl Srvr: T DS (, ) may dlay lowr-rory job longr han rodc a T(, ). Soradc Srvr (SS): Nvr u mor m han h rodc a T(, ) wh am aramr. If o, w can ra T SS ju a a rodc a. Soradc Srvr n Sac-Prory Sym Noaon: T : Ta ym wh n a. T SS : Soradc rvr, arbrary rory. T H : Sub of T wh hghr rory han T SS. r : La rlnhmn m. f : Fr nan afr r a whch rvr bgn o xcu. : Effcv rlnhmn m. Th chdulr drmn bad on hory and nx rlnhmn m o.

CPSC-663: Ral-Tm Sym Arodc and Soradc Job Sml Soradc Srvr Conumon Rul: Th rvr xcuon budg conumd a h ra of on a any m afr r unl h budg xhaud whnvr h followng wo condon ar ru. Whn h condon ar no ru, h rvr hold budg: C: Th rvr xcung. C2: Th rvr ha xcud nc r and undd a h m, and T H dl. Rlnhmn Rul: R: Th xcuon budg o and h currn m r rcordd nally whn h ym bgn xcuon and ach m whn h budg rlnhd. R2: Th nx budg rlnhmn m drmnd a m f whn h rvr fr bgn o xcu nc r. A m f, o h la m nan a whch a lowr-rory a xcu n ( r, f ), and o r f T H buy hroughou h nrval. Th nx rlnhmn m a. R3: Th nx rlnhmn occur a h nx rlnhmn m, xc undr h followng condon whn h rlnhmn may b don oonr or lar. (a) If h nx rlnhmn m arlr han f, h budg rlnhd a oon a xhaud. (b) Th budg rlnhd a m whnvr h ym T ha bn dl bfor and a rodc job rlad a. Sml Soradc Srvr: Examl T = (3, 0.5) T 2 = (4,.0) T 3 = (9, 4.5) T S = (5,.5) T T 2 A (r=3, =) A 2 (r=7, =2) A 3 (r=5.5, =2) T S T 3.5.0 0.5 Budg 2

CPSC-663: Ral-Tm Sym Arodc and Soradc Job A Suaon whr Rul 3a Al T H rvr T L a a f Informal Proof of Corrcn Corrcn : Th rvr nvr dmand mor m n any nrval han corrondng rodc a T = (, ). For now: T ha no bn dl, and Rul R3(b) ha nvr ald. W how ha rvr mula Ta T =(, ). For h, w vw rlnhmn m a nomnal rla m of rvr job. Rul C: Each rvr job nvr xcu for mor han budg.. Rul C2: Budg of dl oradc rvr dcra a f rvr wa xcung. Each rvr job only xcu a m whn a job of T would. On h ohr hand: C2 man ha rvr hold on o budg whn Job n T H xcung. (obvouly corrc) Srvr ha no xcud nc r. (Acual rla m can b lar han nomnal rla m.) 3

CPSC-663: Ral-Tm Sym Arodc and Soradc Job Informal Proof of Corrcn (con.) Rul R2 and R3(a) ma ur ha nx rlnhmn m alway m un lar han ffcv rla m. Th nx ffcv rla m nvr arlr han nx rlnhmn m. R2: Ma ffcv rla m a arly a obl. R3(a): Emula uaon whr job n T a mor m o coml han on rod. R3(b): Alcabl only whn buy nrval of rodc a ym nd, and a nw on ar. Bhavor of a n old buy rod do no affc nw buy rod. Th condon alrady accound for n chdulably analy of T and T. Emulang Gnralzd Procor Sharng Gnralzd Procor Sharng (b-by-b Round Robn): Tmng olaon. Emula GPS by (for xaml) Conan Ulzaon Srvr Toal Bandwdh Srvr Wghd Far-Quung Srucur: Run rvr algorhm on o of EDF chdulr. 4

CPSC-663: Ral-Tm Sym Arodc and Soradc Job Schdulng Soradc Job wh EDF Dfnon: Dny of oradc job J wh rla m r, maxmum xcuon m and dadln d : dny = /(d -r ). Thorm: A ym of ndndn, rmabl oradc job chdulabl accordng o EDF f h oal dny of all acv job n h ym no grar han a all m. Schdulng Soradc Job wh EDF (con) Thorm no ncary! Examl: dny 0.5.0.5 5

CPSC-663: Ral-Tm Sym Arodc and Soradc Job Schdulng Soradc Job wh EDF (con) Soradc a S a ram of oradc job S, S 2, S 3,... Excuon m of S j j. Prod j m bwn nvocaon of S j and S (j). Inananou ulzaon of oradc job S j : j / j. Inananou ulzaon of oradc a S : u = max j ( j / j ). Corollary: A ym of n ndndn, rmabl oradc a, whch uch ha h rlav dadln of vry job qual o rod, chdulabl on a rocor accordng o h EDF algorhm f h oal nananou ulzaon qual or l o. Conan Ulzaon Srvr Algorhm A conan ulzaon rvr mula a oradc a wh a conan nananou ulzaon. Conumon rul: A rvr conum budg only whn xcu. Rlnhmn rul (aum: rvr allocad ulzaon u ): R Inally, := 0 and d :=0. R2 Whn an arodc job wh xcuon m arrv a m o an my arodc job quu, (a) f < d, do nohng. (b) f >= d, :=, and d := /u. R3 A h dadln d of h rvr, (a) f h rvr bacloggd, := and d := d /u (b) f h rvr dl, do nohng. 6

CPSC-663: Ral-Tm Sym Arodc and Soradc Job Conan Ulzaon Srvr: Examl T = (3, 0.5) T 2 = (4,.0) T 3 = (9, 4.5) T T 2 T 3 A (r=3, =) A 2 (r=6.9, =2) A 3 (r=5.5, =2) T CU (u =25%) 2.0.0 Budg Wha abou Unnown Excuon Tm? Aumon for conan ulzaon rvr: xcuon m of arodc job ar nown uon arrval. Rrcv. Pobl oluon: Agn fxd bandwdh o rvr: fxd budg fxd rod /u Uon job comlon of job wh xcuon m <, rduc currn dadln of rvr by ( -)/u bfor rlnhng agan. For xcuon m >, u mor han on rvr rod. 7

CPSC-663: Ral-Tm Sym Arodc and Soradc Job Problm wh Conan Ulzaon Srvr: Unud Caacy T = (3, 0.5) T 2 = (4,.0) T 3 = (9, 4.5) T T 2 T 3 A (r=3, =) A 2 (r=6.9, =2) A 3 (r=4, =2) T S 2.0.0 Budg! d=5 Toal Bandwdh Srvr Allow rvr o u bacground m. Conumon rul: A rvr conum budg only whn xcu. Rlnhmn rul: R Inally, := 0 and d :=0. R2 Whn an arodc job wh xcuon m arrv a m o an my arodc job quu, d := max(d,) /u, and :=. R3 Uon comlon of h currn arodc job, rmov job from quu. (a) f h rvr bacloggd, d := d /u and := ; (b) f h rvr dl, do nohng. 8

CPSC-663: Ral-Tm Sym Arodc and Soradc Job Unud Caacy Elmnad wh Toal Bandwdh Srvr T = (3, 0.5) T 2 = (4,.0) T 3 = (9, 4.5) T T 2 T 3 A (r=3, =) A 2 (r=6.9, =2) A 3 (r=4, =2) T S 2.0.0 Budg d=5 Corrcn of Toal Bandwdh Srvr Conan Ulzaon Srvr corrc. How do Toal Bandwdh Srvr affc rodc a dffrnly? Only nrng ca: Budg of Toal Bandwdh Srvr rlnchd a m bfor dadln. Nw dadln d = d /u. How do h affc h xcuon of rodc a? Ca : Currn rodc job J,c ha dadln bfor d xcuon of rodc job no affcd. Ca 2: Currn rodc job J,c ha dadln afr d Ca 2.: Currn rodc job J,c rady bfor m xcuon m dmandd by Toal Bandwdh Srvr from r, o d am a for Conan Ulzaon Srvr. Ca 2.2: Currn rodc job J,c rady afr m xcuon m dmandd by Toal Bandwdh Srvr from r, o d l han ha of Conan Ulzaon Srvr. 9

CPSC-663: Ral-Tm Sym Arodc and Soradc Job Nonrmabl Poron Nonrmabl oron hr rduc chdulabl ulzaon or nroduc ardn. Dfnon: b max (n) maxmum xcuon m of nonrmabl oron of rodc a and job xcud by rvr. ffcv xcuon m of job xcud by rvr: rao of job xcuon m and rvr z. D mn mnmum of all rlav dadln of rodc a and ffcv xcuon m of job xcud by all rvr n h ym. Corollary: Whn a ym of rodc a chduld wh on or mor oal bandwdh and conan ulzaon rvr on h EDF ba, vry rodc a and vry rvr m dadln f h um of h oal dny of h rodc a and h oal z of all rvr no grar han -b max (n)/d mn. Farn and Sarvaon Toal Bandwdh Srvr no far Examl: TB and TB 2 ach of z 0.5 If boh rvr nvr dl, rvc aroxmaly qually hard among rvr. Wh dlng rvr, h no alway h ca. 0 TB bacloggd TB 2 dl lo of hor job arrv for TB 2 dadln for TB >= 2 2 Procor m allocad farly durng (0,2), bu no durng (, 2) 20

CPSC-663: Ral-Tm Sym Arodc and Soradc Job Farn & Sarvaon Dfnon (Farn): w (, 2 ) = oal aand rocor m for Srvr durng m nrval (, 2 ). w (, 2 )/u = normalzd rvc. Schdulr far durng nrval (, 2 ) f normalzd rvc aand by all rvr do no dffr by mor han a farn hrhold FR. Idally, FR zro: w (, ) u 2 = w (, ) = u ( ) 2 2 w (, ) u j 2 j Elmnang Sarvaon Problm wh Toal Bandwdh rvr: Whn rocng m avalabl, allow o ndfnly u dadln no h fuur. Conan Ulzaon rvr dadln clo : d - <=,max / u, whr,max max. xcuon m of job rvd by Srvr. Rlnhmn rul for arvaon-fr Conan Ulzaon / Bacground rvr: R - R3 : Sam a Conan Ulzaon rvr. R4 : Whnvr buy nrval nd, rlnh budg of all bacloggd rvr. No: Bacground m no drbud o bacloggd rvr accordng o hr z => arvaon lmnad, bu no unfarn. 2

CPSC-663: Ral-Tm Sym Arodc and Soradc Job Prmv Wghd Far-Quung Algorhm Rlnhmn rul mlar o Toal Bandwdh rvr; xc for comuaon of dadln a ach rlnhmn m. WFQ algorhm bound farn. Rlnhmn rul of WFQ rvr ma mula GPS rvr wh am z. u =/4 5 0 5 2.2 u 2 =/8 u 3 =/4 3 25.2 u 4 =3/8 8 3 Vrual Tm: Enquu job n ordr of fnh numbr: numbr of round for GPS rvr o xhau budg. Rul for WFQ Schdulng Rul: Agn ror n ordr of ncrang fnh numbr. Conumon Rul: WFQ rvr conum budg only whn xcu. Inalzaon Rul: I: Whn ym dl, FN = 0, U b = 0, - = 0. Budg of all rvr ar zro. I2: Whn fr job arrv a m wh xcuon m a om rvr FQ whn ym dl: (a) - :=, and U b := U b u, and (b) budg of FQ o and fnh numbr fn := /u. Rul for udang Fnh Tm durng Sym Buy Inrval: R: Whn job arrv a quu FQ whl FQ dl (a) ncrmn ym fnh numbr FN := FN (- - )/U b (b) - :=, and U b := U b u, and (c) budg of FQ o and fnh numbr fn := FN /u, nquu rvr R2: Whnvr FQ coml job (a) f rvr rman bacloggd, rvr budg o and ncrmn fnh numbr: fn := fn /u. (b) f rvr bcom dl, uda Ub and FN a follow: FN := FN (- - )/U b, - :=, and U b := U b -u. 22

CPSC-663: Ral-Tm Sym Arodc and Soradc Job Schdulng Soradc Ta n EDF Sym Toal dny of rodc a. A long a oal dny of oradc job do no xcd -, all dadln can b manand by EDF. Accanc T: Manan l of m nrval and hr dn: Accanc hn add nw dny: dny of nw job 23