Paul Ruth Anirban Mandal Brian Blanton Jeffery Tilson The 20th GENI Engineering Conference June 2124 2014 University of California Davis Davis CA ExoGENI Custom Images ID: 635459
Download Presentation The PPT/PDF document "GENI Science Shakedown Experiments" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
GENI Science Shakedown Experiments
Paul Ruth,
Anirban
Mandal, Brian Blanton, Jeffery Tilson
The
20th GENI Engineering Conference June 21-24, 2014 University of California Davis, Davis, CA
ExoGENI
Custom
Images
ExoGENI
InstaGENI
Required ConfigurationNEuca tools. Available rpm, deb, and source. Emulab client side tools. Source only.Image FormatAMI, AKI, and ARI (standard formats).Frisbee disk image (Emulab format).RegistrationHosted on any http server. Requires XML metadata.Pre-registered with Emulab.Image DebuggingFew errors seen by user. No access to console log.Many errors and console log available to user.
Porting Images
Between
Testbeds
Motifnetwork
Scaled
Images On Each
Testbed
Porting
images EG to
IG
Successfully ported images from EG to IGToo challenging to be recommended to most usersPorting IG to EG Should be possible with ExoGENI snapshot script
New script to snapshot ExoGENI VMsImages defined by XML metadata fileImage (AMI)Kernel (AKI)Ramdisk (ARI)Hosted on HTTP serverSnapshot script creates: image, kernel, ramdisk, metadata from running VMHigh level stepsCreate/modify a VMRun the scriptCopy the new image files to an http serverInsert metadata URL and hash into a request
Shakedown
Applications
ADCIRC
(Storm surge model)
Tightly coupled MPI applicationCurrent running on ExoGENI and InstaGENIMotifNetwork (Computational Genomics)Running on ExoGENIScaling to 100+ coresStorage Limitations on InstaGENIRemaining ChallengesObtaining larger amount of storage on InstaGENI (~50 GB required for Motifnetwork)Starting significant numbers of VMs on InstaGENI (limit ~16)Future GECsPerformance evaluations
ADCIRC Initial Performance Results
Scaling
of ADCIRC MPI application, 4-16 VMs
With 100 Mb/s, performance is better, but no scaling; placement issues?
P
oor Scaling
Scaling of ADCIRC MPI application, 4-16 VMs, for InstaGENI vs. ExoGENI for medium bandwidth case (500 Mb/s)Performance on ExoGENI is 35–48% betterWith larger scale performance difference is greater