Solving Large-Scale Granular Resource Allocation Problems Efficiently with POP

被引:25
|
作者
Narayanan, Deepak [1 ]
Kazhamiaka, Fiodar [1 ]
Abuzaid, Firas [1 ]
Kraft, Peter [1 ]
Agrawal, Akshay [1 ]
Kandula, Srikanth [2 ]
Boyd, Stephen [1 ]
Zaharia, Matei [1 ]
机构
[1] Stanford Univ, Stanford, CA 94305 USA
[2] Microsoft Res, Redmond, WA USA
关键词
Resource scheduling; optimization problems in computer systems; cluster scheduling; traffic engineering; load balancing;
D O I
10.1145/3477132.3483588
中图分类号
TP31 [计算机软件];
学科分类号
081202 ; 0835 ;
摘要
Resource allocation problems in many computer systems can be formulated as mathematical optimization problems. However, finding exact solutions to these problems using off-the-shelf solvers is often intractable for large problem sizes with tight SLAs, leading system designers to rely on cheap, heuristic algorithms. We observe, however, that many allocation problems are granular: they consist of a large number of clients and resources, each client requests a small fraction of the total number of resources, and clients can interchangeably use different resources. For these problems, we propose an alternative approach that reuses the original optimization problem formulation and leads to better allocations than domain-specific heuristics. Our technique, Partitioned Optimization Problems (POP), randomly splits the problem into smaller problems (with a subset of the clients and resources in the system) and coalesces the resulting sub-allocations into a global allocation for all clients. We provide theoretical and empirical evidence as to why random partitioning works well. In our experiments, POP achieves allocations within 1.5% of the optimal with orders-of-magnitude improvements in runtime compared to existing systems for cluster scheduling, traffic engineering, and load balancing.
引用
收藏
页码:521 / 537
页数:17
相关论文
共 50 条
  • [1] Efficiently solving very large-scale routing problems
    Arnold, Florian
    Gendreau, Michel
    Sorensen, Kenneth
    COMPUTERS & OPERATIONS RESEARCH, 2019, 107 : 32 - 42
  • [2] SOLVING LARGE-SCALE MATCHING PROBLEMS EFFICIENTLY - A NEW PRIMAL MATCHING APPROACH
    DERIGS, U
    NETWORKS, 1986, 16 (01) : 1 - 16
  • [3] A random coordinate descent method for large-scale resource allocation problems
    Necoara, I.
    2012 IEEE 51ST ANNUAL CONFERENCE ON DECISION AND CONTROL (CDC), 2012, : 4474 - 4479
  • [4] Solving large-scale control problems
    Benner, P
    IEEE CONTROL SYSTEMS MAGAZINE, 2004, 24 (01): : 44 - 59
  • [5] The Optimal Allocation of Large-scale Water Resource
    Xu Chen-guang
    Li Yue-peng
    PROGRESS IN CIVIL ENGINEERING, PTS 1-4, 2012, 170-173 : 1883 - 1886
  • [6] RISK ALLOCATION IN LARGE-SCALE RESOURCE VENTURES
    SIEBERT, H
    KYKLOS, 1987, 40 (04) : 476 - 495
  • [7] Towards Solving Large-scale Expensive Optimization Problems Efficiently Using Coordinate Descent Algorithm
    Rahnamayan, Shahryar
    Mousavirad, Seyed Jalaleddin
    2020 IEEE INTERNATIONAL CONFERENCE ON SYSTEMS, MAN, AND CYBERNETICS (SMC), 2020, : 2506 - 2513
  • [8] SOLVING LARGE-SCALE TOUR SCHEDULING PROBLEMS
    JARRAH, AIZ
    BARD, JF
    DESILVA, AH
    MANAGEMENT SCIENCE, 1994, 40 (09) : 1124 - 1144
  • [9] Solving large-scale optimization problems with EFCOSS
    Bischof, CH
    Bücker, HM
    Lang, B
    Rasch, A
    ADVANCES IN ENGINEERING SOFTWARE, 2003, 34 (10) : 633 - 639
  • [10] SOLVING (LARGE-SCALE) MATCHING PROBLEMS COMBINATORIALLY
    DERIGS, U
    METZ, A
    MATHEMATICAL PROGRAMMING, 1991, 50 (01) : 113 - 121