16th Symposium on Computer Architecture and High Performance Computing
DOI: 10.1109/sbac-pad.2004.12
|View full text |Cite
|
Sign up to set email alerts
|

Characterizing the Dynamic Behavior of Workload Execution in SVM systems

Abstract: 1The overhead associated with software management of shared virtual memory (SVM) systems can seriously impact overall system performance. One way to remedy this situation is to design more efficient SVM consistency protocols.In this paper we study a number of parallel workload characteristics that can negatively impact the performance of SVM systems. We attempt to quantify the sources of performance loss in some parallel workloads. Our goal is to better understand these characteristics, enabling us to develop … Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
1
1

Citation Types

1
2
0

Publication Types

Select...
1
1

Relationship

1
1

Authors

Journals

citations
Cited by 2 publications
(3 citation statements)
references
References 10 publications
(12 reference statements)
1
2
0
Order By: Relevance
“…In other words, how more efficient consistency protocols can be designed using the knowledge obtained from workload characterizations. In this sense, this paper join and extends the work performed in [7,8,24].…”
supporting
confidence: 64%
“…In other words, how more efficient consistency protocols can be designed using the knowledge obtained from workload characterizations. In this sense, this paper join and extends the work performed in [7,8,24].…”
supporting
confidence: 64%
“…A previous characterization work [13] induced us to design the HLRC-DU protocol, which is based on the HLRC protocol. In the proposed protocol the writer node sends write updates instead of write notices when it detects diffs smaller than an experimental threshold, which is used to avoid the network becomes a bottleneck as well as to reduce asynchronous communication.…”
Section: Discussionmentioning
confidence: 99%
“…In a previous work [13], we performed a characterization study of the discussed phenomena on several parallel workloads. We concluded that i) asynchronous communication occurs in bursts, producing critical section dilation; ii) due to the large granularity supported by SVM systems sharing pattern transformation appears, considerably increasing the amount of asynchronous page requests, which is one metric directly related with performance losses.…”
Section: Motivationmentioning
confidence: 99%