Analysis of SRPT scheduling: investigating unfairness

By Nikhil Bansal, Mor Harchol-Balter. Measurement and Modeling of Computer Systems, 2001

388 citations recorded by the source corpus.

Subfields: Computer Networks and Communications, Electrical and Electronic Engineering, Management Information Systems

View Paper: https://www.semanticscholar.org/paper/44b9d4fa5cdf82d87527e44c6b4c249d95a6bd3b

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Networks and Communications

In Electrical and Electronic Engineering

In Management Information Systems

Citations in the corpus

Improving Online Algorithms via ML Predictions (2024, background): “(As observed in prior work [7, 8, 2], job size distributions in a number of settings are well-modeled by a Pareto distribution with α close to 1.)”
Wireless downlink data channels: user performance and cell dimensioning (2003, background): “These allocations, which are known to outperform the processor sharing discipline for a heavy-tailed service distribution, have been proposed to be implemented in Web servers in a wireline context [3, 19] and more recently in scheduling algorithms in a wireless context [13].”
Performance Modeling and Design of Computer Systems: Queueing Theory in Action (2013, methodology): “2 (All-Can-Win [12]) Given an M/G/1, if ρ < (1)2 , then, ∀x, E [T (x)] ≤ E [T (x)] .”
Size-based scheduling to improve web performance (2003, methodology): “For a mathematical formalization of the above argument, in the case of an M/G/1 queue, we refer the reader to [Bansal and Harchol-Balter 2001].”
pFabric: minimal near-optimal datacenter transport (2013, background): “[8]), under realistic heavy-tailed traffic distributions, SRPT actually improves the majority of flows (even the large flows) compared to TCP’s fair sharing.”
The war between mice and elephants (2001, background): “We then argue that, to achieve transmission rates comparable to that of long connections, short TCP connections should experience much less packet drops, which requires preferential treatment at the bottleneck link queues.”
Finishing flows quickly with preemptive scheduling (2012, methodology): “Assuming a M/G/1 queueing model with heavy-tailed flow distributions; see [4].”