Download presentation
Presentation is loading. Please wait.
1
Job Fairness in Queue Scheduling:
Hanoch Levy School of Computer Science Tel Aviv University Alternative name: “Satellite-linked Web Caches” May 2005 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
2
To provide FAIRNESS in waiting/service
Why (ordered) Queues? Why Fairness in Queues? “Not Fair!!!” “This is more Fair…” To provide FAIRNESS in waiting/service Queue = “A Fairness Management Facility” 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
3
Why Fairness in Queues? (2)
Fairness inherent/crucial part of queues Recent studies, Rafaeli et. al. [2003] (experimental psychology): Experiments on humans in queue scenarios Fairness in queue is very important to people Perhaps even more than delay itself! The issue Evaluate (metrics) Queue Fairness Use on: System + scheduling policies Aim: Applications 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
4
The Size vs. Seniority Dilemma
Mr. Short vs. Mrs. Long Is it more fair to serve Short ahead of Long? By how much? Mr. Short Mrs. Long 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
5
Fairness of Job Scheduling in Queueing Systems, H. Levy
Outline Queue Model Job-Based systems, Flow-Based systems + applications Prior work “Requirements” The performance issue: Delay vs. Service The granularity level: How fine… Dealing with stochastics The physical entities: seniority, service requirement, resources The Fairness Measures - Overview + properties: Seniority based fairness Service time based fairness Resource allocation based fairness Mixture based Application perspective References 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
6
Queue Model (single server)
Jobs Service time Arrivals Departures 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
7
History: Queueing Theory and Fairness
Queueing theory: Decades of research Delay of individual Practical Applications: many & diverse Fairness in queues: Many importance statements: Importance of fairness: Larson (1988), Palm (1953), Mann (1969), Whitt (1984), Rothkopf & Rech (1987) Analysis: was Little, now growing(job fairness): Morris & Wang (85) Gordon (87) Avi-Itzhak & Levy (96) Bender, Chakrabarti . Muthukrishnan (98), Wierman & Harchol-Balter (03), Harchol-Balter et. al. (03) Avi-Itzhak, Levy & Raz (05), Raz, Avi-Itzhak & Levy (04, 05a,b), Raz, Levy& Avi-Itzhak (04), Brosh, Levy & Avi-Itzhak (05) Sandmann (05) (more is coming) Exception: Flow Fairness 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
8
History: Queueing Theory and Fairness (2)
We know only little about queue (job) fairness! More complex than measuring individual delay! 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
9
Designing a Metrics - Keep in mind:
To be used by: Researchers Designers / operators Customers (appeal to) Desired properties (“requirements”) Fits application(s) [as many as possible] Based on a “sound” intuitive basis Fit widely accepted intuition in simple cases Yields to analysis Call center 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
10
The performance issues: Delay vs. Service
Job delay (waiting time, sojourn time) Traditional queueing-theory measure The major factor when service is guaranteed What a wait… Job (service) Completion Have the job done Less popular in queueing theory Applies when service is not guaranteed Ticket line for scarce tickets My seat is gone… 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
11
Granularity of Fairness Evaluation
At what granularity level, should fairness be evaluated C1 C2 C0 Individual Scenario System (steady state) All are important Individual, scenario: build confidence (scale of reference) in metrics System : to evaluate systems/policies Note: All exist for individual waiting times 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
12
The Physical Factors (queues)
Seniority (Arrival Time) Size (Service Requirement) Resources 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
13
Size and Seniority preference principles (“requirements”)
11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
14
Size and Seniority preference principles (“requirements”)
Seniority principle Weak: All jobs same service times if ai< aj then more fair to complete service of Ji before Jj Strong: Ji and Jj same service times Service-requirement principle Weak: All jobs same arrival times if si< sj then more fair to complete service of Ji before Jj Strong: Ji and Jj same arrival times 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
15
How Scheduling policies meet the principles (are fair by principle)
FCFS LCFS ROS (random) Shortest job first longest job first Seniority + - Service requirement 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
16
The Size vs. Seniority Dilemma
Mr. Short vs. Mrs. Long Is it more fair to serve Short ahead of Long? By how much? Mr. Short Mrs. Long 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
17
Review of Measures (job based)
Seniority (Arrival Time) Seniority based Service-requirement based Size (Service Requirement) Resource based Resources Mixtures 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
18
Approach 1: Order (seniority) Fairness
(Arrival Time) Gordon (87) – Ph.D MIT Deals with skips and slips: Performance measure of overtaking to quantify the level of “social justice” B skips A (A experiences a slip): A arrives before B, B leaves before A Considers several Markovian models Analyses distribution of #slips, #skips Measure: No explicit suggestion Lets try: E[#skips ] - E [ # slips] Problem: E[#skips ] - E [ # slips]=0 So perhaps: E[#skips ] B A 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
19
Approach 1: Order (seniority) Fairness: results
(Arrival Time) Systems analyzed : 2 M/M/1 in parallel 2 M/M/1’s with one SJQ “smart” customer Multi server system: M/M/m Infinite server system: M/G/Inf Analysis: distribution of SKIP and SLIP Results: AVG(#skips) = AVG (#slips) Most systems: Dist(#skips) != Dist (#slips) Smart customer: increases # skips, decreases # slips M/U/Inf with symmetric service time dist: Dist(#skips) = Dist (#slips) [only system found] B A 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
20
Approach 1: Order (seniority) Fairness: How do I use it?
(Arrival Time) Compute: Dist(#skips), Dist (#slips) Analysis or simulation How to use for fairness measure – open issue. B A 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
21
Measure 2: “service-time” Fairness
Size (Service Requirement) Bender, Chakrabarti . Muthukrishnan (1998), Harchol-Balter et. al. (2003) Wierman & Harchol-Balter (2003): Propose a Fairness Criterion Emphasis on service requirement Slowdown: for job of size x compute E[T(x)/x] If the slowdown is lower then 1/(1-ρ) for all x - FAIR Classification of a large variety of policies: Always fair: fair for all dist. + loads Always unfair: unfair for all dist. + loads How my type is doing on average 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
22
Fairness of Job Scheduling in Queueing Systems, H. Levy
Size (Service Req.) Compute E[T(x)] (use queueing theory or simulation), divide by x and check rules You end up with a criterion. Can create a measure. We are trying to use SRM to analyze the CSDS 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
23
Measure 3: Resource Allocation Fairness (RAQFM)
Resources Raz, Levy, Avi-Itzhak (04) Aim at the dilemma between size and seniority Focus on: fair share of resources Ideal: At t, each customer deserves 1/N(t) of system resources (N(t)= # customers(t)) Compare warranted service with granted service Warranted service (customer i) Granted service: Individual Discrimination: System unfairness measure: Summary statistics (var) of 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
24
Resource allocation Fairness: Results
Resources Adhering to principles: STRONG Seniority service principles – YES WEAK Service-time service principle – YES STRONG Service-time service principle – NOT 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
25
Resource allocation Fairness: Results (2)
Resources PS most fair Reacts to both s.time and seniority FCFS > LCFS (seniority dominant) FCFS < PLCFS (s. time dominant) 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
26
Resource allocation Fairness: Results (3)
Resources Multiple server more fair than single server with same combined rate Single Queue more fair than Multi-Queue Jockeying increases fairness Jockeying from head more fair than jockeying from tail Prioritizing short jobs is more fair in most cases, but not if the jobs are almost as large as the rest of the population 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
27
Resource Allocation Fairness: Applicability
Resources Good for: S. times and A. times arbitrary Issue is: Waiting times What a wait… Applications: Waiting lines where resources guaranteed Call centers (non-scarce resources) Web services Supermarkets Airport services We are trying to use SRM to analyze the CSDS 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
28
Resource Allocation Fairness: How Do I use it
Resources Derive variance of discrimination at steady state Use Queueing Theory + methodology developed Good for variety of Markovian systems Large systems need approximations (more research) Waiting lines where resources guaranteed Use simulation We are trying to use SRM to analyze the CSDS 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
29
What fits? Go by the application
Service critical Airline reservation, Call center, “Line for bread” Order fairness Delay sensitive, seniority-blind (and careless) customers Computer systems (?) “service time” fairness Delay sensitive, variable service time Call centers, Supermarkets, Banks, Computer systems Resource allocation fairness Seniority (Arrival Time) Size (Service Requirement) Resources Mixture 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
30
How People Perceive Queue Fairness
Rafaeli et. al. (2005): Queue experiments on humans FCFS is FAIR (seniority) Violation of FCFS == unfair Regardless if close to the person asked or away of it. Size-issue – not addressed yet. Multi-server: Multi-Queue (MQ) less fair than single queue. One queue in MQ is shorter less fair. If people PAY for the short Q (first-class) Fair. 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
31
Fairness of Job Scheduling in Queueing Systems, H. Levy
Concluding remarks Fairness in Queues is important Measures must Fit applications Agree with one’s intuition / be consistent Researcher, designer, customer Yield to analysis Research of subject in it’s early stages Much more to study: Scheduling policies Weights Multiple queues /servers Complex structures Relations between measures Other measures… 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
32
Closing Words: Why Study Fairness in Queues (5 reasons)
The Fundamental: (“Q = fairness”) Isn’t fairness why we have queues (for human services) in the first place? The Scientific evidence: (Rafaeli et. al. [2003]) Fairness in queues important to people / perhaps > delay The Inductive reasoning: WFQ: 10’s of papers – fairness on O(1) microsecond jobs!! The Challenge: We know very little on Queue fairness The problem is harder than Q delay! The Practical: Relax your nerves… 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
33
Fairness of Job Scheduling in Queueing Systems, H. Levy
THANK YOU 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
34
Dealing with Stochastics: Actual measures vs. Expected values (2)
Harder to compute Expected values: May be misleading “Like having salaries given in random (with their mean be the real salary)”. How do they treat me now… How do they treat short jobs on avg… 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
35
Dealing with Stochastics: The Black Box question
Customer does not see other customers “Close approach”: don’t see (arrivals) don’t care. Only size matters: Is my size class doing ok? “Open approach”: Customer is still interested in Fairness even though he cannot check it directly How do they treat me now… How do they treat short jobs on avg… 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
36
Fairness of Job Scheduling in Queueing Systems, H. Levy
Measure 4: Mixture Mixture Sandmann (05) Mixture: seniority and size Accounts for 2 parameters: #Overtaking (=#slips) [seniority] #Large jobs [Size] X is a large job for C if: i) Upon C’s arrival X has no-less residual service than C. ii) X departs before C. Fairness measure: 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
37
Measure 4 - Mixture : results
Adhering to principles: STRONG Seniority service principles – YES STRONG Service-time service principle – YES Difficulties: Asymmetry in sensitivity to size + seniority E.g.: X arrives with 100 sec s.time, and waits an hour. Y arrives 1 hour later with 99 sec s.time. It is as fair to serve in both orders. Same if 10 hours. 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
38
Measure 4 - Mixture : How do I use it
Analysis of skips – see Gordon (87) – Markovian – can do. Analysis of Large – not provided yet (perhaps in the making). 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
39
Job-Based vs. Flow-Based +Applications (repeat)
Server Q Jobs Server Q Jobs FLOW BASED Customer = Flow Applications: Networking:network level equipment: Routers, gateways, load-balancers JOB BASED Customer = Job Applications: Networking: Application level equipment: Web server, file server Supermarket, Bank, public office + alike Call center Computer system 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
40
A word on flow-based measures
Deal with flows (of packets) Interested mainly in throughput Literature: Fair bandwidth allocation (network) MinMax fairness (Jaffe (81)) Proportional Fairness (Kelly (97)) Fair Scheduling Weighted Fair Queueing (WFQ) Demers, Keshav and Shenker (1990), Greenberg and Madras (1992), Parekh (1992), Parekh and Gallager (1993), (1994), Golestani (1994), Rexford, Greenberg and Bonomi (1996), Bennet and Zhang (1996), others. 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
41
Measure 5: PS proximity WFQ/RFB literature
Scheduling fairness measures Worst Case deviation from PS (extreme values) Relative Fairness Bound (Golestani (94)) RFB: Can we apply it to jobs (problematic): Not “continuously active” Try on L, S: assume FCFS (L arrives before): RFB=L Try on L, S: assume LCFS (L arrives after): RFB=L General M/G/1 (FCFS, LCFS): infinity! 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
42
Measure 5: PS proximity WFQ/RFB literature
Absolute Fairness Bound (AFB) (Greenberg and Madras (1992) and Keshav (1997)) Maximum (time) discrepancy between schedule and PS Applying to jobs: Try PS completion discrepancy of job LCFS = FCFS = infinity! Most non-PS based (non-WFQ): infinity! (SJF, LJF, SRPT..) Good for very precise PS imitations 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
43
Measure 1: Order (seniority) Fairness
(Arrival Time) Avi-Itzhak & Levy (96, 04) Axioms (for G/D/1): what happens to unfairness measure when interchanging customers: P1: Monotonicity in seniority difference of interchanged neighbors P2: Reversibility of neighbor interchange P3: Independence on position and time P4: Fairness change is not affected by customers not interchanged P4G: interchange of non-neighbors 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
44
Order Fairness: results
Seniority (Arrival Time) ai - Arrival time of customer i Di - Waiting displacement of customer i C > 0, arbitrary constant Expected fairness per customer Claim: Under FCFS: FCFS most fair (LCFS least) Thm : Let (W, W*) be the steady state waiting time under (policy, FIFO), then: 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
45
Order Fairness: results (2)
Seniority (Arrival Time) Adhering to principles: Strong seniority service principle: YES Service-time service principle: NOT 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
46
Order Fairness: Properties & Applicability
Seniority (Arrival Time) Good for: S. times identical S. times don’t matter Issue is: Job completion Applications: Scarce-ticket lines Some call-centers My seat is gone... FCFS is most fair (LCFS least) Intuition & concepts: Peoples’ strong belief in order fairness We are trying to use SRM to analyze the CSDS 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
47
Measure 1: Order (seniority) Fairness: How do I use it?
(Arrival Time) Compute Simulation Compute Var[W] for steady state Remark: range of variance 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
48
“Service-time” Fairness: Results
Size (Service Req.) Classification (not measure) of a large variety of policies Any preemptive size based policy is always unfair (all loads & all service dists). All non-size based non-preemptive policies are always unfair for service time dist defined on neighborhood of zero (short jobs discriminated). Age based policies are always unfair FCFS is always UNFAIR LCFS (preemptive) +PS are always FAIR 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
49
“Service-time” Fairness: Results (2)
Size (Service Req.) Adhering to principles: Seniority service principles – NOT Service-time service principle –Open question 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
50
S. time Fairness: Properties & Applicability
Size (Service Req.) Good for: A. times identical A. times not known / don’t see the queue Your size is always the same Issue is: “wash” seniority by averaging. Advantage: relatively simple analysis Applications: Computer systems (?) We are trying to use SRM to analyze the CSDS 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
51
S. time Fairness: How do I use it
Size (Service Req.) Compute E[T(x)] (use queueing theory or simulation), divide by x and check rules You end up with a criterion. Can create a measure. We are trying to use SRM to analyze the CSDS 11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
52
Fairness of Job Scheduling in Queueing Systems, H. Levy
11/20/2018 Fairness of Job Scheduling in Queueing Systems, H. Levy
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.