Scheduling Jobs on Grid Processors Joan Boyar Lene M. Favrholdt - - PowerPoint PPT Presentation

scheduling jobs on grid processors
SMART_READER_LITE
LIVE PREVIEW

Scheduling Jobs on Grid Processors Joan Boyar Lene M. Favrholdt - - PowerPoint PPT Presentation

Scheduling Jobs on Grid Processors Joan Boyar Lene M. Favrholdt Department of Mathematics and Computer Science University of Southern Denmark, Odense NIST 2006 p. 1/21 The Grid Grid computing: wide area distributed computing A New


slide-1
SLIDE 1

Scheduling Jobs on Grid Processors

Joan Boyar Lene M. Favrholdt Department of Mathematics and Computer Science University of Southern Denmark, Odense

NIST 2006 – p. 1/21

slide-2
SLIDE 2

The Grid

Grid computing: wide area distributed computing “A New Infrastructure for 21st Century Science” built on the Internet analogous to electical power grid source and location of processors invisible request resources (processors with memory) pay for resources used

NIST 2006 – p. 2/21

slide-3
SLIDE 3

Grid Scheduling Problem

Jobs: J1, J2,..., Jn given initially job Ji has requirement pi Processors: P1, P2,..., Pk arrive online processor Pj has capacity cj Goal: Minimize total capacity of processors used

NIST 2006 – p. 3/21

slide-4
SLIDE 4

Grid Scheduling Problem

Jobs: J1, J2,..., Jn given initially job Ji has requirement pi Processors: P1, P2,..., Pk arrive online processor Pj has capacity cj Goal: Minimize total capacity of processors used ———————————————————————- Bin Packing Problem [G. Zhang ’97] Items: sizes ∈ {1, 2, ..., B}: s1, s2,..., sn Bins: sizes ∈ {1, 2, ..., B}: b1, b2,..., bk arrive on-line pack current bin before next arrives Goal: Minimize total size of bins used Restriction: Must use bin if any remaining item fits

NIST 2006 – p. 4/21

slide-5
SLIDE 5

Competitive Ratio

A is c-competitive if for any input seq. I, A(I) ≤ c · OPT(I) + b. ր

  • ptimal off-line algorithm

տ

constant

The competitive ratio of A is CRA = inf {c | A is c-competitive} .

NIST 2006 – p. 5/21

slide-6
SLIDE 6

Grid Scheduling Algorithms

FFI — First-Fit Increasing FFD — First-Fit Decreasing searches entire list of items FFDα (1/2 < α ≤ 1) try FFD for each item size B, B − 1, ..., 1 stop looking if bin filled to ≥ α

α ≤ 1/2: FFDα same as FFD α < 3/4: FFDα “same” as FFD on identical bins α > 3/4: can be worse than FFD on identical bins

NIST 2006 – p. 6/21

slide-7
SLIDE 7

FFI — First-Fit Increasing

B = 40.

Item sizes: 4 × [11], 4 × [20] Bin sizes: 4 × [20], 4 × [11], 4 × [39] Result: Asymptotically, FFI uses 2 times what OPT (FFD) uses.

NIST 2006 – p. 7/21

slide-8
SLIDE 8

FFD — First-Fit Decreasing

B = 16.

Input sizes: [12], 2 × [8], 4 × [6], 8 × [5] Bin sizes: [16], 2 × [12], 4 × [10], [12], 6 × [9] Result: FFD uses ≈ 2 times what OPT uses. [G. Zhang]

NIST 2006 – p. 8/21

slide-9
SLIDE 9

FFD2/3 — First-Fit Decreasing2/3

B = 16.

Input sizes: [12], 2 × [8], 4 × [6], 8 × [5] Bin sizes: [16], 2 × [12], 4 × [10], [12], 6 × [9] Partial result: FFD2/3 treats items [12], [8], [8] as FFD. But not [6].

NIST 2006 – p. 9/21

slide-10
SLIDE 10

FFD2/3 — First-Fit Decreasing2/3

B = 16.

Input sizes: [12], 2 × [8], 4 × [6], 8 × [5] Bin sizes: [16], 2 × [12], 4 × [10], [12], 6 × [9] Result: Items of size 5 paired in bins of size 10.

NIST 2006 – p. 10/21

slide-11
SLIDE 11

FFD2/3 — First-Fit Decreasing2/3

B = 60.

Input sizes: n × [40], 2n × [30] Bin sizes: n × [60], n × [40], n × [59] Result: FFD2/3 uses n × 159.

NIST 2006 – p. 11/21

slide-12
SLIDE 12

FFD3/4 — First-Fit Decreasing3/4

B = 60.

Input sizes: n × [40], 2n × [30] Bin sizes: n × [60], n × [40], n × [59] Result: FFD3/4 uses n × 100. CRFFDα ≥ 2+α

1+α.

NIST 2006 – p. 12/21

slide-13
SLIDE 13

FFD2/3 — First-Fit Decreasing2/3

B = 120.

Input sizes: 2n × [60], 6n × [29] Bin sizes: 2n × [88], 6n × [57], n × [120] Result: FFD2/3 uses n × 518.

NIST 2006 – p. 13/21

slide-14
SLIDE 14

FFD3/4 — First-Fit Decreasing3/4

B = 120.

Input sizes: 2n × [60], 6n × [29] Bin sizes: 2n × [88], 6n × [57], n × [120] Result: FFD3/4 uses n × 416. CRFFD2/3 ≥ 2(3s−2)+6(2s−3)

2(3s−2)+4s

≈ 1.8.

NIST 2006 – p. 14/21

slide-15
SLIDE 15

Competitive Ratio — Results

CRFFI = CRFFD = 2. [G. Zhang] For α ≤ r−1

r , 3r 2r−1 ≤ CRFFDα.

1.8 ≤ CRFFD2/3 ≤ 13/7 ≈ 1.857.

CRA ≤ 2 for any “reasonable” A. [G. Zhang] CRA ≥ 5/4 for any deterministic A.

NIST 2006 – p. 15/21

slide-16
SLIDE 16

Relative Worst Order Ratio

AW(I): A’s performance on worst permutation of I, i.e., AW(I) = maxσ

  • A(σ(I))}.
  • AW(I) = BW(I)
  • BW(K)
  • AW(K)
  • BW(J)
  • AW(J)
  • AW(N) = BW(N)
  • BW(O)
  • AW(O)
  • AW(L)
  • BW(L)
  • BW(M)
  • AW(M)

[Boyar,Favrholdt: CIAC 03]

If AW(I) ≥ BW(I) − b for all I, WRA,B = inf {c | AW(I) ≤ c · BW(I) + b for all I}.

NIST 2006 – p. 16/21

slide-17
SLIDE 17

Relative Worst Order Ratio

Competitive Ratio: CRA = max

I

A(I)

OPT(I) Relative Worst Order Ratio: WRA = max

I

maxσ

  • A(σ(I))
  • maxσ
  • B(σ(I))
  • NIST 2006 – p. 17/21
slide-18
SLIDE 18

Relative Worst Order Ratio — Results

FFD is better than FFI FFDα is better than FFI FFD and FFDα are incomparable

NIST 2006 – p. 18/21

slide-19
SLIDE 19

Open Problems

Best α for FFDα? Exact competitive ratio of FFDα? Other algorithms?

NIST 2006 – p. 19/21

slide-20
SLIDE 20

Paging Results w. RWOR

New algorithm RLRU - better than LRU LRU better than FWF Look-ahead helps

NIST 2006 – p. 20/21

slide-21
SLIDE 21

Other Results w. RWOR

Bin Packing:

Worst-Fit better than Next-Fit.

Dual Bin Packing:

First-Fit better than Worst-Fit.

Bin Coloring:

Greedy better than keeping only one open bin.

Scheduling – minimizing makespan on two related machines:

Post-Greedy better than using only fast machine.

Proportional Price Seat Reservation:

First-Fit better than Worst-Fit.

NIST 2006 – p. 21/21