Search or ask a question
Pricing
Login
Sign up
Home
Notebook
BETA
Literature Review
Copilot
Citation generator
Paraphraser
AI Detector
Chrome Extension
Talk with us
Use on ChatGPT
All figures (14)
Fig. 1: Power8 NUMA architecture
TABLE VII: Experiment 2: Speedup (B=Node with binding, NB= Node without binding)
Fig. 4: Experiment 3: Number of executions (60 minutes interval)
TABLE VIII: Configurations for different co-scheduling workload combinations; (n-n=NUMA node combination; w1=workload-1; w2=workload-2; w/n=worker per node; c/w-=core per worker; w+e=worker and executor memory; PR=PageRank)
Fig. 3: Experiment 3: Completion time speedup (60 minutes interval; binding vs non-binding(OS default allocation))
Fig. 5: Experiment 3: Context switches per second (60 minutes interval)
Fig. 6: Experiment 3: Amount of remote memory access in GB (60 minutes interval)
TABLE I: Spark configuration parameters
TABLE II: Experiment 1: Evaluated software configurations (wem is worker and executor memory; tma is total memory allocated; and tsw is total Spark workers)
TABLE IV: SVM completion time (seconds) groups
TABLE III: Experiment 1: Best configuration when optimizing for completion time
TABLE V: SQL completion time (seconds) groups
TABLE VI: PageRank completion time (seconds) groups
Fig. 2: Experiment 1: CPU Usage (percentage) and Memory Bandwidth (GB/s) for optimal configuration
Proceedings Article
•
DOI
•
Performance Characterization of Spark Workloads on Shared NUMA Systems
[...]
Shuja-ur-Rehman Baig
1
,
Marcelo Amaral
1
,
Jordà Polo
2
,
David Carrera
1
•
Institutions (2)
Barcelona Supercomputing Center
1
,
Polytechnic University of Catalonia
2
01 Mar 2018
-