TY - GEN
T1 - Finishing flows quickly with preemptive scheduling
AU - Hong, Chi Yao
AU - Caesar, Matthew
AU - Godfrey, P. Brighten
PY - 2012
Y1 - 2012
N2 - Today's data centers face extreme challenges in providing low latency. However, fair sharing, a principle commonly adopted in current congestion control protocols, is far from optimal for satisfying latency requirements. We propose Preemptive Distributed Quick (PDQ) flow scheduling, a protocol designed to complete flows quickly and meet flow deadlines. PDQ enables flow preemption to approximate a range of scheduling disciplines. For example, PDQ can emulate a shortest job first, algorithm to give priority to the short flows by pausing the contending flows. PDQ borrows ideas from centralized scheduling disciplines and implements them in a fully distributed manner, making it scalable to today's data centers. Further, we develop a multipath version of PDQ to exploit path diversity. Through extensive packet-level and flow-level simulation, we demonstrate that PDQ significantly outperforms TCP, ItCP and D3 in data center environments. VVe further show that PDQ is stable, resilient to packet loss, and preserves nearly all its performance gains even given inaccurate flow information.
AB - Today's data centers face extreme challenges in providing low latency. However, fair sharing, a principle commonly adopted in current congestion control protocols, is far from optimal for satisfying latency requirements. We propose Preemptive Distributed Quick (PDQ) flow scheduling, a protocol designed to complete flows quickly and meet flow deadlines. PDQ enables flow preemption to approximate a range of scheduling disciplines. For example, PDQ can emulate a shortest job first, algorithm to give priority to the short flows by pausing the contending flows. PDQ borrows ideas from centralized scheduling disciplines and implements them in a fully distributed manner, making it scalable to today's data centers. Further, we develop a multipath version of PDQ to exploit path diversity. Through extensive packet-level and flow-level simulation, we demonstrate that PDQ significantly outperforms TCP, ItCP and D3 in data center environments. VVe further show that PDQ is stable, resilient to packet loss, and preserves nearly all its performance gains even given inaccurate flow information.
KW - Data center
KW - Deadline
KW - Flow scheduling
UR - http://www.scopus.com/inward/record.url?scp=84894564139&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84894564139&partnerID=8YFLogxK
U2 - 10.1145/2377677.2377710
DO - 10.1145/2377677.2377710
M3 - Conference contribution
AN - SCOPUS:84894564139
SN - 9781450314190
T3 - Computer Communication Review
SP - 127
EP - 138
BT - Proceedings of the ACM SIGCOMM 2012 and Best Papers of the Co-located Workshops
T2 - Annual Conference of the ACM Special Interest Group on Data Communication on the Applications, Technologies, Architectures, and Protocols for Computer Communication, ACM SIGCOMM 2012
Y2 - 13 August 2012 through 17 August 2012
ER -