inferring the topology and traffic load of parallel programs in a vm environment
DESCRIPTION
Inferring the Topology and Traffic Load of Parallel Programs in a VM environment. Ashish Gupta Peter Dinda Department of Computer Science Northwestern University. Overview. Motivation behind parallel programs in a VM environment Goal: To infer the communication behavior - PowerPoint PPT PresentationTRANSCRIPT
![Page 1: Inferring the Topology and Traffic Load of Parallel Programs in a VM environment](https://reader036.vdocuments.us/reader036/viewer/2022062501/56815cb3550346895dcab166/html5/thumbnails/1.jpg)
Inferring the Topology and Traffic Load of Parallel Programs in a VM
environment
Ashish GuptaPeter Dinda
Department of Computer ScienceNorthwestern University
![Page 2: Inferring the Topology and Traffic Load of Parallel Programs in a VM environment](https://reader036.vdocuments.us/reader036/viewer/2022062501/56815cb3550346895dcab166/html5/thumbnails/2.jpg)
Overview• Motivation behind parallel programs in a
VM environment• Goal: To infer the communication
behavior• Offline implementation• Evaluating with parallel benchmarks• Online Monitoring in a VM environment• Conclusions
![Page 3: Inferring the Topology and Traffic Load of Parallel Programs in a VM environment](https://reader036.vdocuments.us/reader036/viewer/2022062501/56815cb3550346895dcab166/html5/thumbnails/3.jpg)
Virtuoso: A VM based abstraction for a Grid environment
![Page 4: Inferring the Topology and Traffic Load of Parallel Programs in a VM environment](https://reader036.vdocuments.us/reader036/viewer/2022062501/56815cb3550346895dcab166/html5/thumbnails/4.jpg)
![Page 5: Inferring the Topology and Traffic Load of Parallel Programs in a VM environment](https://reader036.vdocuments.us/reader036/viewer/2022062501/56815cb3550346895dcab166/html5/thumbnails/5.jpg)
Motivation
• A distributed computing environment based on Virtual Machines– Raw machines connected
to user’s network– Our Focus: Middleware support
to hide the Grid complexity
![Page 6: Inferring the Topology and Traffic Load of Parallel Programs in a VM environment](https://reader036.vdocuments.us/reader036/viewer/2022062501/56815cb3550346895dcab166/html5/thumbnails/6.jpg)
![Page 7: Inferring the Topology and Traffic Load of Parallel Programs in a VM environment](https://reader036.vdocuments.us/reader036/viewer/2022062501/56815cb3550346895dcab166/html5/thumbnails/7.jpg)
![Page 8: Inferring the Topology and Traffic Load of Parallel Programs in a VM environment](https://reader036.vdocuments.us/reader036/viewer/2022062501/56815cb3550346895dcab166/html5/thumbnails/8.jpg)
Motivation
• A distributed computing environment based on Virtual Machines– Raw machines connected
to user’s network– Our Focus: Middleware support
to hide the Grid complexity
• Our goal here: Efficient execution of Parallel applications in such an environment
![Page 9: Inferring the Topology and Traffic Load of Parallel Programs in a VM environment](https://reader036.vdocuments.us/reader036/viewer/2022062501/56815cb3550346895dcab166/html5/thumbnails/9.jpg)
ParallelApplication Behavior
Intelligent Placement and virtual networking
of parallel applications
VM Encapsulation Virtual Networks With VNET
![Page 10: Inferring the Topology and Traffic Load of Parallel Programs in a VM environment](https://reader036.vdocuments.us/reader036/viewer/2022062501/56815cb3550346895dcab166/html5/thumbnails/10.jpg)
VNET
• Abstraction: A set of VMs on same Layer 2 network
• Virtual Ethernet LAN
![Page 11: Inferring the Topology and Traffic Load of Parallel Programs in a VM environment](https://reader036.vdocuments.us/reader036/viewer/2022062501/56815cb3550346895dcab166/html5/thumbnails/11.jpg)
Goal of this project
Low Level Traffic Monitoring
?
An online topology inference framework for a VM environment
Application Topology
![Page 12: Inferring the Topology and Traffic Load of Parallel Programs in a VM environment](https://reader036.vdocuments.us/reader036/viewer/2022062501/56815cb3550346895dcab166/html5/thumbnails/12.jpg)
Approach
Design an offline framework
Evaluate with parallel benchmarks
If successful, design an online framework for VMs
![Page 13: Inferring the Topology and Traffic Load of Parallel Programs in a VM environment](https://reader036.vdocuments.us/reader036/viewer/2022062501/56815cb3550346895dcab166/html5/thumbnails/13.jpg)
An offline topology inference framework
Goal: A test-bed for traffic monitoring
and evaluating topology inference methods
![Page 14: Inferring the Topology and Traffic Load of Parallel Programs in a VM environment](https://reader036.vdocuments.us/reader036/viewer/2022062501/56815cb3550346895dcab166/html5/thumbnails/14.jpg)
The offline method
Synced Parallel Traffic Monitoring
Traffic Filtering and Matrix Generation
Matrix Analysis and Topology Characterization
![Page 15: Inferring the Topology and Traffic Load of Parallel Programs in a VM environment](https://reader036.vdocuments.us/reader036/viewer/2022062501/56815cb3550346895dcab166/html5/thumbnails/15.jpg)
The offline method
Synced Parallel Traffic Monitoring
Traffic Filtering and Matrix Generation
Matrix Analysis and Topology Characterization
![Page 16: Inferring the Topology and Traffic Load of Parallel Programs in a VM environment](https://reader036.vdocuments.us/reader036/viewer/2022062501/56815cb3550346895dcab166/html5/thumbnails/16.jpg)
The offline method
Synced Parallel Traffic Monitoring
Traffic Filtering and Matrix Generation
Matrix Analysis and Topology Characterization
![Page 17: Inferring the Topology and Traffic Load of Parallel Programs in a VM environment](https://reader036.vdocuments.us/reader036/viewer/2022062501/56815cb3550346895dcab166/html5/thumbnails/17.jpg)
The offline method
Synced Parallel Traffic Monitoring
Traffic Filtering and Matrix Generation
Matrix Analysis and Topology Characterization
![Page 18: Inferring the Topology and Traffic Load of Parallel Programs in a VM environment](https://reader036.vdocuments.us/reader036/viewer/2022062501/56815cb3550346895dcab166/html5/thumbnails/18.jpg)
The offline method
Synced Parallel Traffic Monitoring
Traffic Filtering and Matrix Generation
Matrix Analysis and Topology Characterization
PVMPOV Inference
![Page 19: Inferring the Topology and Traffic Load of Parallel Programs in a VM environment](https://reader036.vdocuments.us/reader036/viewer/2022062501/56815cb3550346895dcab166/html5/thumbnails/19.jpg)
Synced Parallel Traffic Monitoring
Traffic Filtering and Matrix Generation
Matrix Analysis and Topology Characterization
Infer.pl
![Page 20: Inferring the Topology and Traffic Load of Parallel Programs in a VM environment](https://reader036.vdocuments.us/reader036/viewer/2022062501/56815cb3550346895dcab166/html5/thumbnails/20.jpg)
Parallel Benchmarks Evaluation
Goal:To test the practicality of low level
traffic based inference
![Page 21: Inferring the Topology and Traffic Load of Parallel Programs in a VM environment](https://reader036.vdocuments.us/reader036/viewer/2022062501/56815cb3550346895dcab166/html5/thumbnails/21.jpg)
Parallel Benchmarks used
• Synthetic benchmarks: Patterns– N-dimensional mesh-neighbor– N-dimensional toroid-neighbor– N-dimensional hypercubes– Tree reduction – All-to-All
• Scheduling mechanism to generate deadlock free and efficient schemes
1 2 3
![Page 22: Inferring the Topology and Traffic Load of Parallel Programs in a VM environment](https://reader036.vdocuments.us/reader036/viewer/2022062501/56815cb3550346895dcab166/html5/thumbnails/22.jpg)
Application benchmarks
• NAS PVM benchmarks– Popular benchmarks for parallel computing– 5 benchmarks
• PVM-POV : Distributed Ray Tracing• Many others possible…
• The inference not PVM specific– Applicable to all communication .– e.g. MPI, even non-parallel apps
![Page 23: Inferring the Topology and Traffic Load of Parallel Programs in a VM environment](https://reader036.vdocuments.us/reader036/viewer/2022062501/56815cb3550346895dcab166/html5/thumbnails/23.jpg)
Patterns application
2-D Mesh 3-D Toroid 3-D Hypercube
Reduction Tree All-to-All
![Page 24: Inferring the Topology and Traffic Load of Parallel Programs in a VM environment](https://reader036.vdocuments.us/reader036/viewer/2022062501/56815cb3550346895dcab166/html5/thumbnails/24.jpg)
PVM NAS benchmarks
Parallel Integer Sort
![Page 25: Inferring the Topology and Traffic Load of Parallel Programs in a VM environment](https://reader036.vdocuments.us/reader036/viewer/2022062501/56815cb3550346895dcab166/html5/thumbnails/25.jpg)
Traffic Matrix for PVM IS benchmark
![Page 26: Inferring the Topology and Traffic Load of Parallel Programs in a VM environment](https://reader036.vdocuments.us/reader036/viewer/2022062501/56815cb3550346895dcab166/html5/thumbnails/26.jpg)
Traffic Matrix for PVM IS benchmark
Placement of host1 is crucial on the network
![Page 27: Inferring the Topology and Traffic Load of Parallel Programs in a VM environment](https://reader036.vdocuments.us/reader036/viewer/2022062501/56815cb3550346895dcab166/html5/thumbnails/27.jpg)
An Online Topology Inference Framework: VTTIF
Goal:To automatically detect, monitor and
report the global traffic matrix for a set of VMs running on a overlay network
![Page 28: Inferring the Topology and Traffic Load of Parallel Programs in a VM environment](https://reader036.vdocuments.us/reader036/viewer/2022062501/56815cb3550346895dcab166/html5/thumbnails/28.jpg)
Overall Design
• VNET– Abstraction: A set of VMs on same Layer 2
network– Virtual Ethernet LAN
![Page 29: Inferring the Topology and Traffic Load of Parallel Programs in a VM environment](https://reader036.vdocuments.us/reader036/viewer/2022062501/56815cb3550346895dcab166/html5/thumbnails/29.jpg)
A VNET virtual layer
VNET Layer
Physical Layer
A Virtual LAN over wide area
![Page 30: Inferring the Topology and Traffic Load of Parallel Programs in a VM environment](https://reader036.vdocuments.us/reader036/viewer/2022062501/56815cb3550346895dcab166/html5/thumbnails/30.jpg)
Overall Design
• VNET– Abstraction: A set of VMs on same Layer 2
network• Extend VNET to include the required features
– Monitoring at Ethernet packet level• The Challenge here
– Lacks manual control– Detecting interesting parallel program
communication ?
![Page 31: Inferring the Topology and Traffic Load of Parallel Programs in a VM environment](https://reader036.vdocuments.us/reader036/viewer/2022062501/56815cb3550346895dcab166/html5/thumbnails/31.jpg)
Detecting interesting phenomenon
Reactive Mechanisms Proactive Mechanisms
•Certain address properties
•Based on Traffic rate
•Etc.
Provide support for queries by external agent
Rate based monitoringNon-uniform discrete event sampling
What is the Traffic Matrix for the last n seconds ?
Like a Burglar Alarm Video Surveillance
![Page 32: Inferring the Topology and Traffic Load of Parallel Programs in a VM environment](https://reader036.vdocuments.us/reader036/viewer/2022062501/56815cb3550346895dcab166/html5/thumbnails/32.jpg)
Traffic Analyzer
Rate based Change detection
Traffic MatrixQuery Agent
VM Network Scheduling Agent
VNET daemon
VM
VNET overlay network
To other VNET daemons
Physical Host
![Page 33: Inferring the Topology and Traffic Load of Parallel Programs in a VM environment](https://reader036.vdocuments.us/reader036/viewer/2022062501/56815cb3550346895dcab166/html5/thumbnails/33.jpg)
Traffic Matrix Aggregation
• Each VNET daemon keeps track of local traffic matrix– Need to aggregate this information for a global view– When the rate falls, the local daemons push the traffic
matrix (When do you push the traffic matrix ?)– Operation is associative: reduction trees for scalability
The proxy daemon
![Page 34: Inferring the Topology and Traffic Load of Parallel Programs in a VM environment](https://reader036.vdocuments.us/reader036/viewer/2022062501/56815cb3550346895dcab166/html5/thumbnails/34.jpg)
Evaluation
• Used 4 Virtual Machines over VNET • NAS IS benchmark
![Page 35: Inferring the Topology and Traffic Load of Parallel Programs in a VM environment](https://reader036.vdocuments.us/reader036/viewer/2022062501/56815cb3550346895dcab166/html5/thumbnails/35.jpg)
Conclusions
Possible to infer the topology with
low level traffic monitoring
A Traffic Inference Framework for Virtual Machines
Ready to move on to future steps:Adaptation for Performance
![Page 36: Inferring the Topology and Traffic Load of Parallel Programs in a VM environment](https://reader036.vdocuments.us/reader036/viewer/2022062501/56815cb3550346895dcab166/html5/thumbnails/36.jpg)
Current Work
• Capabilities for dynamic adaptation into VNET
• Spatial Inference Network Adaptation for Improved Performance
• Prelim Results: Improved performance upto 40% in execution time
• Looking into benefits of Dynamic Adaptation
![Page 37: Inferring the Topology and Traffic Load of Parallel Programs in a VM environment](https://reader036.vdocuments.us/reader036/viewer/2022062501/56815cb3550346895dcab166/html5/thumbnails/37.jpg)
For more information
• http://virtuoso.cs.northwestern.edu• VNET is available for download
• PLAB web site:
plab.cs.northwestern.edu