Distributed executions of iterative machine learning (ML) algorithms can suffer significant performance losses due to stragglers. The regular (e.g., per iteration) barriers used in the traditional BSP approach cause every transient slowdown of any worker thread to delay all others. This project describes a scalable, efficient solution to the straggler problem for this important class of ML algorithms, combining a more flexible synchronization model with dynamic peer-to-peer re-assignment of work among workers. Experiments with real straggler behavior observed on Amazon EC2, as well as injected straggler behavior stress tests, confirm the significance of the problem and the effectiveness of the solution, as implemented in a framework called FlexRR. Using FlexRR, we consistently observe near-ideal run-times (relative to no performance jitter) across all real and injected straggler behaviors tested.
Matrix Factorization Running on two classes of AWS EC2 Machines |
LDA Running on two classes of AWS EC2 Machines |
FlexRR resolving emulated straggler patterns |
FACULTY
Greg Ganger
Phil Gibbons
Garth Gibson
Eric Xing
GRAD STUDENTS
Aaron Harlap
Henggang Cui
We thank the members and companies of the PDL Consortium: Amazon, Bloomberg, Datadog, Google, Honda, Intel Corporation, IBM, Jane Street, Meta, Microsoft Research, Oracle Corporation, Pure Storage, Salesforce, Samsung Semiconductor Inc., Two Sigma, and Western Digital for their interest, insights, feedback, and support.