skip to main content

Search for: All records

Award ID contains: 1717834

Note: When clicking on a Digital Object Identifier (DOI) number, you will be taken to an external site maintained by the publisher. Some full text articles may not yet be available without a charge during the embargo (administrative interval).
What is a DOI Number?

Some links on this page may take you to non-federal websites. Their policies may differ from this site.

  1. Deep learning (DL) is a popular technique for building models from large quantities of data such as pictures, videos, messages generated from edges devices at rapid pace all over the world. It is often infeasible to migrate large quantities of data from the edges to centralized data center(s) over WANs for training due to privacy, cost, and performance reasons. At the same time, training large DL models on edge devices is infeasible due to their limited resources. An attractive alternative for DL training distributed data is to use micro-clouds---small-scale clouds deployed near edge devices in multiple locations. However, micro-clouds present the challenges of both computation and network resource heterogeneity as well as dynamism. In this paper, we introduce DLion, a new and generic decentralized distributed DL system designed to address the key challenges in micro-cloud environments, in order to reduce overall training time and improve model accuracy. We present three key techniques in DLion: (1) Weighted dynamic batching to maximize data parallelism for dealing with heterogeneous and dynamic compute capacity, (2) Per-link prioritized gradient exchange to reduce communication overhead for model updates based on available network capacity, and (3) Direct knowledge transfer to improve model accuracy by merging the bestmore »performing model parameters. We build a prototype of DLion on top of TensorFlow and show that DLion achieves up to 4.2X speedup in an Amazon GPU cluster, and up to 2X speed up and 26% higher model accuracy in a CPU cluster over four state-of-the-art distributed DL systems.« less
  2. Adaptability is critical for stream processing systems to ensure stable, low-latency, and high-throughput processing of long-running queries. Such adaptability is particularly challenging for wide-area stream processing due to the highly dynamic nature of the wide-area environment, which includes unpredictable workload patterns, variable network bandwidth, occurrence of stragglers, and failures. Unfortunately, existing adaptation techniques typically achieve these performance goals by compromising the quality/accuracy of the results, and they are often application-dependent. In this work, we rethink the adaptability property of wide-area stream processing systems and propose a resource-aware adaptation framework, called WASP. WASP adapts queries through a combination of multiple techniques: task re-assignment, operator scaling, and query re-planning, and applies them in a WAN-aware manner. It is able to automatically determine which adaptation action to take depending on the type of queries, dynamics, and optimization goals. We have implemented a WASP prototype on Apache Flink. Experimental evaluation with the YSB benchmark and a real Twitter trace shows that WASP can handle various dynamics without compromising the quality of the results.