Progress Report

(1)

Progress Report

2014/06/20

(2)

Things to Do from Last Mee ting

 Journal version for ICPP 2014 pape r

◦Enhance with reviewer comments.

 Prepare ITRI midterm report

◦Late June(?)

 How to reduce cutting stock proble m to knapsack

(3)

Agenda

 Reduce cutting stock problem to kn apsack

 Progress report

(4)

Cutting Stock Problem

 There are a number of rolls of paper o f fixed width waiting to be cut.

 Need different numbers of rolls of var ious-sized width.

 Minimize the waste.

 Use integer linear programming to find how many times each pattern i s used

◦Pattern is a possible combination of cuts.

(5)

Reducible to the Knapsack Problem

 Column generation is a way of solv ing cutting stock problem.

◦Start with only a few patterns and ge nerate new ones as needed.

 Generating a new pattern by maximi zing the reduced cost.

◦Integer knapsack problem.

 any feasible solution for the auxiliary pr oblem corresponds to a feasible cutting pa ttern in the cutting-stock problem.

(6)

Progress Report

 Problem formulation

◦Divide each time interval into 100 ti me slices.

◦Given a set of vCPU and the amount of time slices a vCPU can use on each ph ysical core.

◦Find a way to schedule these vCPU.

 A vCPU cannot appear in two or more physic al core on the same time.

(7)

Example – 3 vCPUs, 2 Physical Co re

… …

vCPU₀ (60, 20)

vCPU₁ (0, 50)

vCPU₂ (20, 30)

t=10 0

t=0

(8)

Current Status

 Consider n physical cores with m v CPUs.

◦n < m

 Greedy seems to be a good solution (?)

 Still working on this problem.

(9)

Things to Do

 Journal version for ICPP 2014 pape r

◦Enhance with reviewer comments.

 Prepare ITRI midterm presentation

◦6/25