TY - GEN
T1 - SNC-meister
T2 - 7th ACM Symposium on Cloud Computing, SoCC 2016
AU - Zhu, Timothy
AU - Berger, Daniel S.
AU - Harchol-Balter, Mor
N1 - Funding Information:
This research is supported in part by Intel as part of the Intel Science and Technology Center for Cloud Computing (ISTC-CC), by a Google Faculty Research Award 2015/16, and by the National Science Foundation under awards CMMI-1538204, CMMI-1334194, CSR-1116282, and XPS-1629444. We also thank the member companies of the PDL Consortium for their interest, insights, feedback, and support.
Publisher Copyright:
© 2016 ACM.
PY - 2016/10/5
Y1 - 2016/10/5
N2 - Meeting tail latency Service Level Objectives (SLOs) in shared cloud networks is both important and challenging. One primary challenge is determining limits on the multitenancy such that SLOs are met. Doing so involves estimating latency, which is difficult, especially when tenants exhibit bursty behavior as is common in production environments. Nevertheless, recent papers in the past two years (Silo, QJump, and PriorityMeister) show techniques for calculating latency based on a branch of mathematical modeling called Deterministic Network Calculus (DNC). The DNC theory is designed for adversarial worst-case conditions, which is sometimes necessary, but is often overly conservative. Typical tenants do not require strict worst-case guarantees, but are only looking for SLOs at lower percentiles (e.g., 99th, 99.9th). This paper describes SNC-Meister, a new admission control system for tail latency SLOs. SNC-Meister improves upon the state-of-the-art DNC-based systems by using a new theory, Stochastic Network Calculus (SNC), which is designed for tail latency percentiles. Focusing on tail latency percentiles, rather than the adversarial worst-case DNC latency, allows SNC-Meister to pack together many more tenants: in experiments with production traces, SNC-Meister supports 75% more tenants than the state-of-the-art.
AB - Meeting tail latency Service Level Objectives (SLOs) in shared cloud networks is both important and challenging. One primary challenge is determining limits on the multitenancy such that SLOs are met. Doing so involves estimating latency, which is difficult, especially when tenants exhibit bursty behavior as is common in production environments. Nevertheless, recent papers in the past two years (Silo, QJump, and PriorityMeister) show techniques for calculating latency based on a branch of mathematical modeling called Deterministic Network Calculus (DNC). The DNC theory is designed for adversarial worst-case conditions, which is sometimes necessary, but is often overly conservative. Typical tenants do not require strict worst-case guarantees, but are only looking for SLOs at lower percentiles (e.g., 99th, 99.9th). This paper describes SNC-Meister, a new admission control system for tail latency SLOs. SNC-Meister improves upon the state-of-the-art DNC-based systems by using a new theory, Stochastic Network Calculus (SNC), which is designed for tail latency percentiles. Focusing on tail latency percentiles, rather than the adversarial worst-case DNC latency, allows SNC-Meister to pack together many more tenants: in experiments with production traces, SNC-Meister supports 75% more tenants than the state-of-the-art.
UR - http://www.scopus.com/inward/record.url?scp=84995554091&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84995554091&partnerID=8YFLogxK
U2 - 10.1145/2987550.2987585
DO - 10.1145/2987550.2987585
M3 - Conference contribution
AN - SCOPUS:84995554091
T3 - Proceedings of the 7th ACM Symposium on Cloud Computing, SoCC 2016
SP - 374
EP - 387
BT - Proceedings of the 7th ACM Symposium on Cloud Computing, SoCC 2016
A2 - Diao, Yanlei
A2 - Aguilera, Marcos K.
A2 - Cooper, Brian
A2 - Diao, Yanlei
PB - Association for Computing Machinery, Inc
Y2 - 5 October 2016 through 7 October 2016
ER -