skip to main content
US FlagAn official website of the United States government
dot gov icon
Official websites use .gov
A .gov website belongs to an official government organization in the United States.
https lock icon
Secure .gov websites use HTTPS
A lock ( lock ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites.

Explore Research Products in the PAR
It may take a few hours for recently added research products to appear in PAR search results.


Title: Characterizing On-Chip Traffic Patterns in General-Purpose GPUs: A Deep Learning Approach
Architectural optimizations in general-purpose graphics processing units (GPGPUs) often exploit workload characteristics to reduce power and latency while improving performance. This paper finds, however, that prevailing assumptions about GPGPU traffic pattern characterization are inaccurate. These assumptions must therefore be re-evaluated, and more appropriate new patterns must be identified. This paper proposes a methodology to classify GPGPU traffic patterns, combining a convolutional neural network (CNN) for feature extraction and a t-distributed stochastic neighbor embedding (t-SNE) algorithm to determine traffic pattern clusters. A traffic pattern dataset is generated from common GPGPU benchmarks, transformed using heat mapping, and iteratively refined to ensure appropriate and highly accurate labels. The proposed classification model achieves 98.8% validation accuracy and 94.24% test accuracy. Furthermore, traffic in 96.6% of examined kernels can be classified into the eight identified traffic pattern categories.  more » « less
Award ID(s):
1750047
PAR ID:
10210355
Author(s) / Creator(s):
; ; ;
Date Published:
Journal Name:
IEEE 37th International Conference on Computer Design (ICCD)
Page Range / eLocation ID:
56 to 64
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. ; ; ; (, Proceedings of the ACM on Measurement and Analysis of Computing Systems)
    null (Ed.)
    As Graphics Processing Units (GPUs) are becoming a de facto solution for accelerating a wide range of applications, their reliable operation is becoming increasingly important. One of the major challenges in the domain of GPU reliability is to accurately measure GPGPU application error resilience. This challenge stems from the fact that a typical GPGPU application spawns a huge number of threads and then utilizes a large amount of potentially unreliable compute and memory resources available on the GPUs. As the number of possible fault locations can be in the billions, evaluating every fault and examining its effect on theapplication error resilience is impractical. Application resilience is evaluated via extensive fault injection campaigns based on sampling of an extensive fault site space. Typically, the larger the input of the GPGPU application, the longer the experimental campaign. In this work, we devise a methodology, SUGAR (Speeding Up GPGPU Application Resilience Estimation with input sizing), that dramatically speeds up the evaluation of GPGPU application error resilience by judicious input sizing. We show how analyzing a small fraction of the input is sufficient to estimate the application resilience with high accuracy and dramatically reduce the duration of experimentation. Key of our estimation methodology is the discovery of repeating patterns as a function of the input size. Using the well-established fact that error resilience in GPGPU applications is mostly determined by the dynamic instruction count at the thread level, we identify the patterns that allow us to accurately predict application error resilience for arbitrarily large inputs. For the cases that we examine in this paper, this new resilience estimation mechanism provides significant speedups (up to 1336 times) and 97.0 on the average, while keeping estimation errors to less than 1%. 
    more » « less
  2. ; ; (, 17th USENIX Symposium on Networked Systems Design and Implementation (NSDI 20))
    When designing, understanding, or optimizing a computer network, it is often useful to identify and rank common patterns in its usage over time. Often referred to as a network traffic pattern, identifying the patterns in which the network spends most of its time can help ease network operators' tasks considerably. Despite this, extracting traffic patterns from a network is, unfortunately, a difficult and highly manual process. In this paper, we introduce tpprof, a profiler for network traffic patterns. tpprof is built around two novel abstractions: (1) network states, which capture an approximate snapshot of network link utilization and (2) traffic pattern sub-sequences, which represent a finite-state automaton over a sequence of network states. Around these abstractions, we introduce novel techniques to extract these abstractions, a robust tool to analyze them, and a system for alerting operators of their presence in a running network. 
    more » « less
  3. ; ; (, Proceedings of the VLDB Endowment)
    Analyzing database access logs is a key part of performance tuning, intrusion detection, benchmark development, and many other database administration tasks. Unfortunately, it is common for production databases to deal with millions or more queries each day, so these logs must be summarized before they can be used. Designing an appropriate summary encoding requires trading off between conciseness and information content. For example: simple workload sampling may miss rare, but high impact queries. In this paper, we present LogR, a lossy log compression scheme suitable for use in many automated log analytics tools, as well as for human inspection. We formalize and analyze the space/fidelity trade-off in the context of a broader family of “pattern” and “pattern mixture” log encodings to which LogR belongs. We show through a series of experiments that LogR compressed encodings can be created efficiently, come with provable information-theoretic bounds on their accuracy, and outperform state-of-art log summarization strategies. 
    more » « less
  4. ; ; ; ; ; (, Proceedings of the ACM on Measurement and Analysis of Computing Systems)
    Network management often relies on machine learning to make predictions about performance and security from network traffic. Often, the representation of the traffic is as important as the choice of the model. The features that the model relies on, and the representation of those features, ultimately determine model accuracy, as well as where and whether the model can be deployed in practice. Thus, the design and evaluation of these models ultimately requires understanding not only model accuracy but also the systems costs associated with deploying the model in an operational network. Towards this goal, this paper develops a new framework and system that enables a joint evaluation of both the conventional notions of machine learning performance (e.g., model accuracy) and the systems-level costs of different representations of network traffic. We highlight these two dimensions for two practical network management tasks, video streaming quality inference and malware detection, to demonstrate the importance of exploring different representations to find the appropriate operating point. We demonstrate the benefit of exploring a range of representations of network traffic and present Traffic Refinery, a proof-of-concept implementation that both monitors network traffic at 10~Gbps and transforms traffic in real time to produce a variety of feature representations for machine learning. Traffic Refinery both highlights this design space and makes it possible to explore different representations for learning, balancing systems costs related to feature extraction and model training against model accuracy. 
    more » « less
  5. ; (, IEEE INFOCOM 2020 - IEEE Conference on Computer Communications)
    null (Ed.)
    In this paper, we consider the problem of scheduling real-time traffic in wireless networks under a conflict-graph interference model and single-hop traffic. The objective is to guarantee that at least a certain fraction of packets of each link are delivered within their deadlines, which is referred to as delivery ratio. This problem has been studied before under restrictive frame-based traffic models, or greedy maximal scheduling schemes like LDF (Largest-Deficit First) that can lead to poor delivery ratio for general traffic patterns. In this paper, we pursue a different approach through randomization over the choice of maximal links that can transmit at each time. We design randomized policies in collocated networks, multipartite networks, and general networks, that can achieve delivery ratios much higher than what is achievable by LDF. Further, our results apply to traffic (arrival and deadline) processes that evolve as positive recurrent Markov chains. Hence, this work is an improvement with respect to both efficiency and traffic assumptions compared to the past work. We further present extensive simulation results over various traffic patterns and interference graphs to illustrate the gains of our randomized policies over LDF variants. 
    more » « less
  • Citation Formats
  • MLA
  • APA
  • Chicago
  • BibTeX
  • Save / Share this Record
  • Send to Email