wiki:OnTime-Dec09

OnTimeMeasure: Centralized and Distributed Measurement Orchestration Software

Quarterly Status Report Document for Global Environment for Network Innovations (GENI) Spiral 2 Project # 1764

Prasad Calyam (PI) and Paul Schopis (Co-PI)

Ohio Supercomputer Center/OARnet, The Ohio State University

December 2009


I. Major accomplishments

A. Milestones achieved

In our first quarter in the GENI Spiral 2, we have made significant progress towards our milestones [1]. A summary of our accomplishments is as follows:

  • We have reviewed several GPO and GENI Spiral 1 project documents to familiarize ourselves with previous/on-going efforts. Specifically, we focused on the ProtoGENI [2] and GMOC [3] activities to understand the infrastructure and tools available to experimenters. In addition, we reviewed the activities of GENI Spiral 1 measurement projects such as the Measurement System [4], Instrumentation Tools [5] and Data Plane Measurements [6].
  • Project PI (Prasad Calyam) attended the GEC6 in Salt Lake City and met with the various project teams and informed each other about project objectives. PI also discussed software requirements with potential measurement service users.
  • We successfully completed our first milestone (OnTime: S2.a: OnTimeMeasure for ProtoGENI design v1.0). Based on our GENI Spiral 1 projects literature review and our GEC6 interactions, we have developed a high-level design document of our OnTimeMeasure: centralized and distributed measurement orchestration software v1.0 for experimenters in ProtoGENI as well as for GMOC personnel.
  • We successfully completed our second milestone (OnTime: S2.b: PoC for response and escalation group). We have identified Weiping Mandrawa of OARnet/The Ohio State University as the point-of-contact (PoC), who will respond to queries from GENI Operations and take action if any severe or unexpected scenarios occur in OnTimeMeasure software deployments.
  • Notable progress has been made towards our third milestone (OnTime: S2.c: Platform setup for OnTimeMeasure-GENI software development, testing and deployment) due to be completed by Jan 31st 2010. We are working towards setting up a measurement testbed involving 4-sites on the OARnet backbone, and have identified relevant ProtoGENI and GMOC resources (e.g. node configurations, node locations, databases). Also, we are working on coming up with a well defined process for internal project tasks tracking, source version control management and bug-tracking.
  • We have already started planning activities to work on our fourth milestone (OnTime: S2.d: Demo of basic OnTimeMeasure-GENI functionality on ProtoGENI at GEC7 in March 2010). Specifically, we have identified the testbed setup and software modules to be developed to enable an experimenter to: (i) submit measurement requirements via the ProtoGENI control framework, (ii) control the measurement service in experimenter’s slice, and (iii) gather measurement results for analysis.

B. Deliverables made

  • Published the high-level design document of our OnTimeMeasure: centralized and distributed measurement software v1.0 on the GENI project website.
  • Identified Weiping Mandrawa of OARnet/The Ohio State University as the point-of-contact (PoC) for response and escalation group
  • Published quarterly status report document on the GENI project website.

II. Description of work performed during last quarter

A. Activities and findings

Details of our activities and findings in this quarter are as follows:

i. GEC6 trip report:

Project PI (Prasad Calyam) attended the GEC6 in Salt Lake City. He presented brief overviews of project objectives at the Cluster C Breakout Session and Instrumentation and Measurement Working Group meeting. The poster session was an excellent venue to meet with various project teams in person and learn about their project objectives/accomplishments. In addition to attending the plenary sessions, Prasad attended the ProtoGENI cluster meeting that featured tutorials on how to setup slices manually and via the ProtoGENI API. Prasad was also able to interact with potential measurement service users (e.g., Rob Ricci - ProtoGENI, Chris Small - GMOC, Justin Cappos - Seattle, Deniz Gurkhan - Data Plane Measurements) and other measurement related projects (e.g., Martin Swany - LAMP, Sonia Fahmy - Scalable Monitoring).

ii. High-level design document discussions:

Based on our GENI Spiral 1 projects literature review and our GEC6 interactions, we developed a high-level design document of our OnTimeMeasure: centralized and distributed measurement orchestration software v1.0 for experimenters in ProtoGENI as well as for GMOC operations. In the process of preparing the document, we were able to share ideas and solicit feedback from Rob Ricci - ProtoGENI project, Chris Small and Camille Davis - GMOC project, and Mukundan Sridharan - KanseiGENi project. The major challenges in the design process were the integration aspects with ProtoGENI and measurement data accessibility to users. We were able to find out that we needed to use RSpec extensions [7] [8] to be able to have ProtoGENI users submit measurement requests. Also, we reviewed some of the OGF-NMWG schemas for measurement data exchange and looked at sample implementations of corresponding web-service clients [9] [10]. Given the fact that the GENI-specific schemas for RSpec extensions and LAMP project’s prospectus (i.e., common schema for measurement data and metadata) [11] are works-in-progress, we plan to work closely with these teams in order to take advantage of their efforts.

Another investigation we pursued related to how we would integrate with the GMOC monitoring efforts. In our discussions with the GMOC team, we were able to gather that there is no central measurements repository of all experiment slices, and that the NetKarma metadata database [12] was going to the location for us to report any measurements metadata. The GMOC team expressed that they could take advantage of the measurement service for both monitoring and analysis of network performance at various strategic points in the GENI facility. In our discussions with Rob Ricci, we determined the deployment options of our software. Essentially, we found out that we could use their SSL implementation to authenticate experimenters, and address their measurement privacy aspects. Also, based on the backbone setup of the ProtoGENI [13] and the Instrumentation Tools project architecture [5], we learnt that we could use the ProtoGENI control network for coordinating measurement schedules between our Root and Node Beacons, and the ProtoGENI experimental network for collecting measurements within a slice. In our design discussions, we investigated the feasibility of having a separate measurement slice dedicated to run an OnTimeMeasure-GENI daemon or a global Root Beacon that controls all the Node Beacons to service measurement requests in all experiments. We decided against these options due to scalability issues of managing such a system, and besides experiment users may not need access to all their experiment measurement results. If they need the measurement results of an experiment for any reason, we decided to serve these users with the option to download the measurement results archive (e.g., MySQL dump, graph plots, time series data files) using the RequestMeasurementResultsArchive web-service functionality in OnTimeMeasure-GENI.

iii. Platform setup activities for OnTimeMeasure-GENI:

We are working on two major sets of activities for the platform setup. First, we have created experiment slices in ProtoGENI (single-hop network) and have run example measurements of Ping, Traceroute and Iperf to a measurement server on our OARnet backbone. We are working on creating geographically distributed multi-hop experiment slices in ProtoGENI using both the command-line scripts as well as the flash interface.

In the second set of activities, we are working on setting up a measurement testbed involving 4-sites that are geographically distributed on the OARnet backbone. This testbed is our project’s development platform and machines in this testbed will be used for the researcher portal and other functionalities demonstration of our prototypes. The developed software will be deployed on experiment slices created in the ProtoGENI infrastructure.

In the course of our experiment slices setup, we learnt that the Emulab PCs provisioned in a slice in the single-hop case run Red Hat Linux Release 9 (Shrike) with Kernel 2.4.20-31 and gcc-build version 3.2.2 image, whereas the Emulab PCs provisioned in a slice in the geographically distributed multi-hop case run a Fedora 9 image. This finding is important because we need to install similar images in our measurement testbed so that our development platform matches the ProtoGENI deployment platform. Our current measurement testbed is based upon a Debian image and we are in the process of changing it to the Fedora 9 image. The access to our measurement testbed is restricted, and our development team has been provided with secure key cards for login and sudo root permissions by OARnet network operations. We have uploaded SSH public keys from our measurement testbed to the My Emulab web-portal to be able to create experiment slices. In addition to the measurement testbed setup, we have setup SVN/Trac server configurations at OSC with appropriate user accounts for internal project tasks tracking, source version control management and bug-tracking.

B. Project participants

The following participants have helped in the execution of the project activities this quarter:

  • Prasad Calyam (PI) - Ohio Supercomputer Center/OARnet
  • Paul Schopis (Co-PI) - OARnet
  • Weiping Mandrawa (Network Software Engineer) - OARnet
  • Chris Hartley (Network Testbed Engineer) - OARnet

C. Publications (individual and organizational)

D. Outreach activities

http://groups.geni.net/geni/attachment/wiki/Gec6ClusterCMondaySlides/Calyam_OnTimeMeasure-GENI_GEC6.ppt

E. Collaborations

We have started collaborations with ProtoGENI and GMOC project teams, and plan to attend Cluster C conference calls when they start next quarter. We have been using the ProtoGENI mailing list to ask questions and report problems, and the ProtoGENI team responses have been prompt and helpful. We are planning on starting collaboration with the Instrumentation Tools project team. As they have already made substantial progress in integrating their software into ProtoGENI in GENI Spiral 1, we are hoping to learn from their experiences. Given the fact that the GENI-specific schemas for RSpec extensions and LAMP project’s prospectus (i.e., common schema for measurement data and metadata) [11] are works-in-progress, we plan to work closely with these teams in order to take advantage of their efforts in our project.

F. Other Contributions

We notified the ProtoGENI team through examples that the default sliver expiration time was too short. This has been addressed by the ProtoGENI team in their updated version i.e., the timer in the RenewSliver() call in createsliver.py has been increased from 600 to 6000.

We also pointed out to the ProtoGENI team that the OS image running in the single-hop experiment slice is quite old, and hence newer programs developed by the GENI community may not run as expected. The ProtoGENI team is looking into making the new Fedora 9 image used in the geographically distributed experiment slices as the default for the single-hop experiment slices.

We are working on documenting RSpec extensions to specify centralized and distributed measurement orchestration requests. We plan to have them publicly available on our project website so that users of our measurement service and other measurement projects could use them as a reference.


Acknowledgements

We would like to acknowledge the following individuals for participating in our project discussions and providing useful suggestions:

  • Thomas Bitterman - Ohio Supercomputer Center
  • Robert Ricci - University of Utah
  • Chris Small - Indiana University
  • Camille Davis - Indiana University
  • Mukundan Sridharan - The Ohio State University
  • Amruta Joshi - The Ohio State University

References

[1] OnTimeMeasure - http://groups.geni.net/geni/wiki/OnTimeMeasure

[2] ProtoGENI - http://www.protogeni.net

[3] GMOC - http://gmoc.grnoc.iu.edu

[4] Measurement System - http://groups.geni.net/geni/wiki/MeasurementSystem

[5] Instrumentation Tools - http://groups.geni.net/geni/wiki/InstrumentationTools

[6] Data Plane Measurements - http://groups.geni.net/geni/wiki/Data%20Plane%20Measurements

[7] ProtoGENI RSpec - http://www.protogeni.net/trac/protogeni/wiki/RSpec

[8] ProtoGENI RSpec Extensions - http://www.protogeni.net/trac/protogeni/wiki/RSpecExtensions2

[9] OGF-NMWG - http://nmwg.internet2.edu

[10] OGF-NMWG Schema Sample Implementations - http://www.cedps.net/index.php/PerfSONAR_overview https://wiki.man.poznan.pl/perfsonar-mdm/index.php/Creating_a_web-service_using_perfSONAR_base

[11] LAMP - http://groups.geni.net/geni/wiki/LAMP

[12] NetKarma - http://groups.geni.net/geni/wiki/netKarma

[13] ProtoGENI Backbone - http://www.protogeni.net/trac/protogeni/wiki/Backbone

This material is based upon work supported by the National Science Foundation under Grant No. CNS-0940805. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of BBN Technologies, Corp., the GENI Project Office, or the National Science Foundation.

Last modified 10 years ago Last modified on 01/27/10 16:32:12