wiki:MAX.S2.v

Version 4 (modified by Tom Lehman, 9 years ago) (diff)

--

MAX.S2.v: DRAGON Aggregate Manager Enhancement Performance Testing and Evaluation

Conduct performance testing and evaluation of DRAGON Aggregate Manager. Solicit user feedback on use of DRAGON Aggregate Management. Summarize strengths and weakness of DRAGON Aggregate Manager and develop recommendations for future upgrades.

Status:

During this reporting period we completed the implementation and performance testing of topology based MAX Aggregate Manager provisioning. This testing includes use of a PlanetLab based MAX RSpec to specify a GENI experiment topology. These topologies can include PlanetLab slices and dynamically provisioned network paths. A summary of testing procedures and performance results are available on the project web site:

This link includes detailed instructions for how to exercise these features, QuickTime movies showing example provisioning commands, and typical performance results. An example performance test shows that it typically takes about 3 1/2 minutes to provision an experiment topology consisting of three PlanetLab Nodes, three dynamically provisioned network paths, and a dynamically provisioned !ProtoGENI network path.

We have solicited users feedback on the MAX AM. This includes working with a researcher at Yonsei University of Seoul Korea who have installed the MAX AM on their testbed for use and evaluation. This is early work at this stage, and our main focus has been on assisting with initial install and testing. We have also discussed potential areas for extended testing and development collaborations between us. We hope to obtain additional feedback and possible development contributions in the future. We also are working on finding additional users for the MAX AM and GENI facilities. Additional information on general user support status is provided in MAX.S2.x summary below.

Based our implementation and testing of the MAX AM so far, we have the following observations regarding strengths, weaknesses, and focus areas for future improvements:

-The strengths include the modular and extensible design of the MAX AM. We feel this is proper architecture for regional networks like MAX wishing to combine multiple resources into integrated GENI Experiment Topologies.

-Another strength is this modular approach has forced us to deal with the heterogeneous nature of multiple "resource managers" somewhat earlier then others in the community. However, this will be a general challenge for multi-aggregate GENI Experiment Topology building, and we feel the techniques we have applied within the MAX AM can be applied more broadly to the GENI control framework.

-The primary weakness areas are in monitoring and status reporting after instantiation of topology. We expect that these issues will be addressed as part of the next set of development activities which will focus on making the current MAX AM capability set operational. This is expected to include status reporting to users and the GMOC.