|
|
How much power is needed for a billion-thread high-throughput server? |
Zhiwei XU() |
Institute of Computing Technology, Chinese Academy of Sciences, Beijing 100190, China |
|
|
Abstract With the advent of Internet services, big data and cloud computing, high-throughput computing has generated much research interest, especially on high-throughput cloud servers. However, three basic questions are still not satisfactorily answered: (1) What are the basic metrics (what throughput and high-throughput of what)? (2) What are the main factors most beneficial to increasing throughput? (3) Are there any fundamental constraints and how high can the throughput go? This article addresses these issues by utilizing the fifty-year progress in Little’s law, to reveal three fundamental relations among the seven basic quantities of throughput (λ), number of active threads (L), waiting time (W), system power (P), thread energy (E), Watts per thread ω, threads per Joule θ. In addition to Little’s law L = λW, we obtain P = λE and λ = Lωθ, under reasonable assumptions. These equations help give a first order estimation of performance and power consumption targets for billion-thread cloud servers.
|
Keywords
high-throughput computing
billion-thread servers
power consumption
waiting time and latency
performance formulation
Little’s law
|
Corresponding Author(s):
XU Zhiwei,Email:zxu@ict.ac.cn
|
Issue Date: 01 August 2012
|
|
1 |
Barroso L, Hoelzle U. The datacenter as a computer: an introduction to the design of warehouse-scale machines. Synthesis Lectures on Computer Architecture , 2009, 4(1): 1-108 doi: 10.2200/S00193ED1V01Y200905CAC006
|
2 |
Cloudstone. http://radlab.cs.berkeley.edu/wiki/projects/cloudstone
|
3 |
Little J. Little’s law as viewed on its 50th anniversary. Operational Research , 2011, 59(3): 536-549 doi: 10.1287/opre.1110.0940
|
4 |
Little J, Graves S. Little’s law. In: Chhajed D, Lowe T J, eds. Building Intuition: Insights from Basic Operations Management Models and Principles. New York: Springer Science and Business Media LLC, 2008
|
5 |
Garland M, Kirk D. Understanding throughput-oriented architectures. Communications of the ACM , 2010, 53(11): 58-66 doi: 10.1145/1839676.1839694
|
6 |
Hanlon C. A conversation with john hennessy and david patterson. ACM Queue , 2006-2007, 4(10): 14-22 doi: 10.1145/1189276.1189286
|
7 |
Brumelle S L. On the relation between customer and time averages in queues. Journal of Applied Probability , 1971, 8(3): 508-520 doi: 10.2307/3212174
|
8 |
Heyman D, Stidham S J. The relation between customer and time averages in queues. Operational Research , 1980, 28(4): 983-994 doi: 10.1287/opre.28.4.983
|
9 |
Glanz J. Google details, and defends, its use of electricity. The New York Times , 2011
|
10 |
High-throughput computing. http://research.cs.wisc.edu/condor/htc. html. see also http://en.wikipedia.org/wiki/condor_high-throughput_computing_system and http://en.wikipedia.org/wiki/high-throughput_computing
|
11 |
Many-task computing. http://en.wikipedia.org/wiki/Many-task_computing
|
12 |
Little J. A proof for the queuing formula: L = λW. Operational Research , 1961, 9(3): 383-387 doi: 10.1287/opre.9.3.383
|
13 |
El-Taha M, Stidham S. Sample-Path Analysis of Queueing Systems. Springer Netherlands , 1999
|
|
Viewed |
|
|
|
Full text
|
|
|
|
|
Abstract
|
|
|
|
|
Cited |
|
|
|
|
|
Shared |
|
|
|
|
|
Discussed |
|
|
|
|