Abstract
Capacity management, whether it involves servers in a data center, or human staff in a call center, or doctors in a hospital, is largely about balancing a resource-delay tradeoff. On the one hand, one would like to turn off servers when not in use (or send home staff that are idle) to save on resources. On the other hand, one wants to avoid the considerable setup time required to turn an ''off'' server back ''on.'' This paper aims to understand the delay component of this tradeoff, namely, what is the effect of setup time on average delay in a multi-server system?
Surprisingly little is known about the effect of setup times on delay. While there has been some work on studying the M/M/k with Exponentially-distributed setup times, these works provide only iterative methods for computing mean delay, giving little insight as to how delay is affected by k, by load, and by the setup time. Furthermore, setup time in practice is much better modeled by a Deterministic random variable, and, as this paper shows, the scaling effect of a Deterministic setup time is nothing like that of an Exponentially-distributed setup time.
This paper provides the first analysis of the M/M/k with Deterministic setup times. We prove a lower bound on the effect of setup on delay, where our bound is highly accurate for the common case where the setup time is much higher than the job service time. Our result is a relatively simple algebraic formula which provides insights on how delay scales with the input parameters. Our proof uses a combination of renewal theory, martingale arguments and novel probabilistic arguments, providing strong intuition on the transient behavior of a system that turns servers on and off.
- J. R. Artalejo, A. Economou, and M. J. Lopez-Herrero. Analysis of a Multiserver Queue with Setup Times. Queueing Syst., 51(1):53--76, 2005.Google ScholarDigital Library
- S. Asmussen. Applied Probability and Queues, volume 2. Springer, 2003.Google Scholar
- W. Bischof. Analysis of M/G/1-Queues with Setup Times and Vacations under Six Different Service Disciplines. Queueing Syst., 39(4):265--301, 2001.Google ScholarDigital Library
- G. Choudhury. On a batch arrival Poisson queue with a random setup time and vacation period. Comp. & Oper. Res., 25(12):1013--1026, 1998.Google ScholarDigital Library
- A. Gandhi and M. Harchol-Balter. How Data Center Size Impacts the Effectiveness of Dynamic Power Management. In Proc. Ann. Allerton Conf. Communication, Control and Computing, pages 1164--1169, Urbana-Champaign, IL, September 2011.Google ScholarCross Ref
- A. Gandhi and M. Harchol-Balter. M/G/k with staggered setup. Oper. Res. Lett., 41(4):317--320, 2013.Google ScholarCross Ref
- A. Gandhi, V. Gupta, M. Harchol-Balter, and M. Kozuch. Optimality analysis of energy-performance trade-off for server farm management. In Proc. Int. Symp. Computer Performance, Modeling, Measurements and Evaluation (IFIP Performance), Namur, Belgium, November 2010.Google ScholarDigital Library
- A. Gandhi, M. Harchol-Balter, and I. Adan. Server farms with setup costs. Performance Evaluation, 67(11):1123--1138, 2010.Google ScholarDigital Library
- A. Gandhi, M. Harchol-Balter, and M. Kozuch. The case for sleep states in servers. In SOSP Workshop on Power-Aware Computing and Systems (HotPower), pages 1--5, Cascais, Portugal, October 2011.Google ScholarDigital Library
- A. Gandhi, M. Harchol-Balter, and M. Kozuch. Are sleep states effective in data centers? In Int. Conf. Green Computing (IGCC), pages 1--10, San Jose, CA, 2012.Google ScholarDigital Library
- A. Gandhi, M. Harchol-Balter, R. Raghunathan, and M. A. Kozuch. AutoScale: Dynamic, Robust Capacity Management for Multi-Tier Data Centers. ACM Trans. Comput. Syst., 30(4):1--26, 2012.Google ScholarDigital Library
- A. Gandhi, S. Doroudi, M. Harchol-Balter, and A. Scheller-Wolf. Exact analysis of the M/M/k/setup class of Markov chains via Recursive Renewal Reward. In Queueing Syst., pages 153--166, 2013.Google ScholarDigital Library
- Q.-M. He and E. Jewkes. Flow time in the MAP/G/1 queue with customer batching and setup times. Stochastic Models, 11(4):691--711, 1995.Google ScholarCross Ref
- E. Hyytiä, D. Down, P. Lassila, and S. Aalto. Dynamic Control of Running Servers. In Int. Conf. Measurement, Modelling and Evaluation of Comput. Systems, pages 127--141, Erlangen, Germany, 2018. Springer.Google Scholar
- A. Kara. Energy Consumption in Data Centers with Deterministic Setup Times. Master's thesis, Middle East Technical University, 2017.Google Scholar
- M. Mao and M. Humphrey. A Performance Study on the VM Startup Time in the Cloud. In IEEE Int. Conf. Cloud Computing (CLOUD), pages 423--430, Honolulu, HI, 2012.Google ScholarDigital Library
- D. Mukherjee and A. Stolyar. Join Idle Queue with Service Elasticity: Large-Scale Asymptotics of a Nonmonotone System. Stoch. Syst., 9(4):338--358, 2019.Google ScholarCross Ref
- D. Mukherjee, S. Dhara, S. C. Borst, and J. S. van Leeuwaarden. Optimal Service Elasticity in Large-Scale Distributed Systems. Proc. ACM SIGMETRICS Int. Conf. Measurement and Modeling of Computer Systems, 1:1--28, 2017.Google ScholarDigital Library
- J. Pender and T. Phung-Duc. A law of large numbers for m/m/c/delayoff-setup queues with nonstationary arrivals. In Int. Conf. on Analytical and Stochastic Modeling Techniques and Applications, pages 253--268, Cardiff, UK, 2016. Springer.Google ScholarCross Ref
- T. Phung-Duc. Exact solutions for M/M/c/setup queues. Telecommun. Syst., 64(2):309--324, 2017.Google ScholarDigital Library
- K. Rzadca, P. Findeisen, J. Swiderski, P. Zych, P. Broniek, J. Kusmierek, P. Nowak, B. Strack, P. Witusowski, S. Hand, et al. Autopilot: workload autoscaling at Google. In Proc. European Conf. Computer Systems (EuroSys), pages 1--16, Heraklion, Crete, Greece, 2020.Google Scholar
- P. D. Welch. On a Generalized M/G/1 Queuing Process in Which the First Customer of Each Busy Period Receives Exceptional Service. Oper. Res., 12(5):736--752, 1964.Google ScholarDigital Library
Index Terms
- The M/M/k with Deterministic Setup Times
Recommendations
The M/M/k with Deterministic Setup Times
SIGMETRICS '23: Abstract Proceedings of the 2023 ACM SIGMETRICS International Conference on Measurement and Modeling of Computer SystemsCapacity management, whether it involves servers in a data center, or human staff in a call center, or doctors in a hospital, is largely about balancing a resource-delay tradeoff. On the one hand, one would like to turn off servers when not in use (or ...
The M/M/k with Deterministic Setup Times
SIGMETRICS '23Capacity management, whether it involves servers in a data center, or human staff in a call center, or doctors in a hospital, is largely about balancing a resource-delay tradeoff. On the one hand, one would like to turn off servers when not in use (or ...
Priority Queues with Setup Times
<P>Queueing analysis is conducted for M/G/1-type systems with multiple classes of service requests that need a setup time prior to each busy period. We consider a variety of service disciplines: FCFS first-come, first-served, LCFS last-come, first-...
Comments