Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
1
TransPAC3-‐ Asia US High Performance International Networking NSF Award #0962968 Year 5 Annual Report
1 March 2014 through 28 February 2015
Jennifer Schopf – Principal Investigator
(Prepared by Alice Jackson, Jennifer Schopf)
Summary During the Year 5, the TransPacific Asia-‐US High Performance International Networking project (TransPAC3) continued its collaborative and engineering activities to support international science. This report outlines collaborations, software and systems work, operational activities, and usage statistics for the project. Highlights include a 99.98% uptime record for the circuits and their infrastructure, hiring a replacement TransPAC engineer, participation in 15 meetings and workshops, including presentations on several aspects of the TransPAC project, and extending the monitoring infrastructure for the circuits.
1. TransPAC Overview The TransPAC project supports three circuits and a set of network services between the US West coast and Asia. In the current set up, these circuits are:
• The TransPAC Circuit: a10Gbps link between Los Angeles, California, and Tokyo, Japan. This is the primary, NSF-‐funded circuit for the project and used for the bulk of the production project network bandwidth.
• The JGN-‐X Circuit: a 10Gbps layer-‐2 circuit, largely used for experiments and Software Defined Networking (SDN) trials. The Japan Gigabit Network Extension (JGN-‐X) project is a testbed funded by the Japanese National Institute of Information and Communications Technology (NICT) (http://www.nict.go.jp/en). This link is not supported by NSF funds. A backup routed peering connection between TransPAC and APAN also runs across this link.
• The CERNET circuit: a 10Gbps link between Los Angeles, California, and Beijing, China, which is only partially funded by NSF. This is a layer3-‐only connection at this time.
2
These circuits are used in production to support a wide variety of science applications and demonstrations of advanced networking technologies. In addition, the TransPAC award supports tool development, SDN experimental work, measurement deployments, and security activities. Please note that some of the activities (outreach to Africa, PerfSONAR training, etc.) are also included in the ACE annual report project, as appropriate, as joint funding supports these efforts.
2. Financials The TransPAC project over the course of Year 5 spent approximately $1,134,297, which was on target for the project to be spent down on schedule for the original May 31, 2015 end date. See Appendix A for full year accounting. However, in September the IU Finance office finally got a fully updated accounting, including a ground truth of the state of the finances for the project as a whole. At that time it was discovered that the project was estimated to be underspent by $1,440,000 at the project end date if spending did not change significantly. In May it was determined that the underspend amount was closer to $1,050,000. Further investigation found a severe underspend in the first 2 years of the project, which the current PI was unaware of. In order to address this issue, Schopf met with NSF PO Thompson September 30th to discuss future plans and alternatives. The project requested and was granted a one-‐year No Cost Extension, and several possible plans were discussed. In November, Schopf was informed that the proposal for a follow-‐on project, TransPAC4, was being recommended for funding, however at a reduced budget, with the TransPAC3 underspend to cover the first year of expenses for the new award. The TransPAC4 project is being announced March 1, 2015. Over the next 3 months, additional transition planning will take place. A high level spend-‐down plan for the last 15 months (since this report is being submitted Feb 28, but the project end date is May 31) includes:
• Continue all staffing through May 2016 • Continue production LA-‐Tokyo circuit through December 2016 • Continue funding for GlobalNOC support through December 2016 • Continue funding for production IU systems support through December
2016 • Discontinue support for LA-‐Beijing circuit in May 2015, per prior agreed to
MOU with Internet2 • Upgrade LA PoP to collect active and passive monitoring data • Explore additional TransPacific circuits to be supported through TransPAC4,
to begin fall 2015 – this may include additional expenditures if a Seattle PoP is involved
3
• Expand application support team and consider developing workshops to increase use of circuits
With these plans in place, the project should spend out its balance by the end of the sixth year, and insure a smooth transition to TransPAC4.
3. Staffing There were several staffing changes during Year 5 of the project. At the beginning of the year, project staff consisted of:
• Jennifer Schopf, Director • Brent Sweeny, primary TransPAC senior network engineer • Scott Chevalier, primary contact for GlobalNOC support desk • Alice Jackson, administration • George McLaughlin, Asia-‐Pacific network consultant • Dale Smith, consultant
Brent Sweeny retired from Indiana University in May. The job was posted, and Joe Lappa, formerly of Pittsburg SuperComputer Center was hired in July to be the primary engineer for TransPAC. However, in January, Lappa shifted over to the IU iLight project. The job was reposted, and Hans Addleman will join the team in this role March 9. Addleman has a 10-‐year history with the IU GlobalNOC, and a depth of experience working with other International Network team members. George McLaughlin’s contract was going to be competed July, 2014, but he has requested an earlier completion date due to health issues. His last charge for services was in May. Over the next 6 months, we expect to bring on an additional applications support staff member to increase the use of the circuits. We will also explore hiring an additional Asia-‐area consultant to replace McLauglin and aid in science support work.
4. Collaborations, Travel, and Training TransPAC staff participated in various meetings to support their role in collaborations in Asia. These included:
• Internet2 Global Summit, Apri 6-‐10 2014, Denver, CO – Schopf, Sweeny, Smith
o Sweeny received lifetime achievement award o Schopf ran session on use of 100G networks
• U Hawaii Manoa, April 24-‐25, Schopf met with Jason Leigh, David Lassner, Gwen Jacobs, Alan Whinery
• CC:NIE PI workshop, April 30-‐May 1, 2014, Arlington, VA – Schopf o Schopf presented on Campus-‐International interactions
4
• Woods Hole Oceanographic Institution, May 23-‐24, Schopf met with former collaborators and workshop organizers
• TERENA Networking Conference, May 17-‐22, Dublin, Ireland – Schopf, Sweeny
o Sweeny presented on BGP PathHinting • 2014 Environmental Data Summit
(http://environmentaldatasummit2014.deltacouncil.ca.gov/ ), June 4-‐6, Sacramento, CA
o Schopf gave opening Keynote o Additional meetings with ESnet
• Focused Technical Workshop on Global climate science, July 13-‐16, Boulder, CO (http://meetings.internet2.edu/2014-‐cc-‐climate/)
o Schopf organized • APAN 38, Aug 9-‐15, Nantou, Taiwan – Schopf, Lappa
o Schopf chaired 2 sessions o Schopf gave an overview of TransPAC plans
• Lower Mekong Water Resource Workshop (http://internationalnetworking.iu.edu/archives/LMI/index.html), August 18-‐22, Hanoi, Vietnam – Schopf, Lee
o Schopf co-‐organized • Chinese American Networking Symposium, September 15-‐17, New York City
– Lappa o Lappa presented on use of CERNET circuit
• PRAGMA 27, Oct 15-‐17, Bloomington, IN o Schopf was a co-‐organizer for the event o Schopf chaired the Network Engineering Workshop o Lappa and Sweeny presented
• Internet2 Technology Exchange, October 26-‐30, Indianapolis, IN – Schopf, Lappa, Lee
o Lee also attended the Sunday program “Broadening the Reach: A Cyberinfrastructure Program for Non-‐Researh-‐Intensive and EPSCoR Institutions”
• SuperComputing, November 16-‐21, New Orleans – Schopf, Lee, Lappa o Schopf co-‐organized “Innovating the Network for Data-‐Intensive
Science” workshop. • ESIP Winter meeting, January 6-‐8, Washington, DC – Schopf • PerfSonar Workshop, January 21-‐23, Columbus, OH – Schopf, Lee
Throughout the year, the team increased its collaborations with international groups significantly. At the upcoming APAN meeting in March we will begin hosting TransPAC-‐specific sessions. Over the next year, we plan to stay involved in many of these conferences and workshops, and increase our work with end-‐users.
5
5. Software and Systems Work
A. Tool Development Tool development continued this quarter with expected patches and upgrades to existing tools.
B. Dynamic Circuits The TransPAC project provides Dynamic Layer 2 Network services through the OSCARS software suite. This service peers with the Internet2 ION project and the JGN-‐X Dynamic network facilities. Researchers and scientists can interactively create a layer2 dynamic circuit between Asia and the US to transfer data.
C. Software Defined Networking (SDN) Activities One of the primary research goals of the TransPAC project is to enable Software Defined Networking (SDN) using Open Flow to provide a mechanism to dynamically configure and control circuit behavior between the US and Asia. Support for JGN-‐X’s RISE and Open Flow experiments continued. Unfortunately, due to staff changes, the SDN work over TransPAC did not move forward as much as hoped in Project Year 5. This will be a focus of effort over the last year of the project, including several talks and demos by the group.
D. Measurement Activities The TransPAC project supports a perfSONAR deployment in Los Angeles that provides periodic testing between several US and Asian sites. A testing matrix is available at http://tp3-‐3.transpac3.iu.edu/maddash-‐webui/ , however it is currently not collecting any data. We have started up conversations with the JGN monitoring group 3 times over the last 12 months to try to get this addressed to no avail. We will continue to press on this issue. We negotiated with the CERNet group to begin to collect Flow data in the form of “Top Talkers” list. They gave us a sample in August and another in December, see Appendix B for the slides. With our new staffing we will be following up to collect this data on a monthly basis for the remaining lifespan of the circuit. In Year 6 we will also be collecting our own flow data on the NSF-‐funded TransPAC circuit, as part of the upgrade to the LA POP.
6
6. Operational Activities The TransPAC circuits were very stable over the year, with only 6 downtimes, of which only one was unscheduled. The circuits have not changed over the course of the project year. We are in the process of upgrading the Los Angeles PoP to enable additional monitoring and measurement capabilities, as well as better supporting upcoming SDN activities. The Ixia tester, purchased as part of the ACE award, will also be placed in LA to allow better use of it for testing. This will require an upgrade to our router and port to support a 100G connection. Figure 1 shows the current set up at the POP, and Figure 2 diagrams the planned upgrade.
Figure 1: LA POP at present time.
Figure 2: LA POP when upgrade is completed.
7
A. Network Engineering The TransPAC circuit between Los Angeles, California, and Tokyo, Japan, continues to function as designed with no unscheduled outages. It supports IPv4, IPv6, and dynamic-‐circuit functions, and is prototyping OpenFlow 1.0 capabilities. The JGN-‐X circuit between Los Angeles, California, and Tokyo, Japan, operates in layer 2 between switches in Tokyo and Los Angeles, and is primarily used for experimental network research, particularly DCN, OpenFlow, GLIF, Optical testbeds, and OpenGOLEs. The Beijing CERNET link will be no longer part of the TransPAC project after May 31, 2015, per prior agreement with Internet2. Planning for this shift has taken place.
B. Traffic Graphs Figure 3 and Figure 4 show the traffic on the TransPAC network during Project Year 5, and Figures 5 and 6 shows data for the CERNET connection for the same period. Monitoring data is not available for the JGN-‐X circuit.
Figure 3: Aggregate traffic using average daily values on the 10Gbps TransPAC (NSF-funded) circuit between Los Angeles and Tokyo
8
Figure 4: Aggregate traffic using maximum daily values on the 10Gbps TransPAC (NSF-funded) circuit between Los Angeles and Tokyo
Figure 5: Aggregate traffic using average daily values on the 10G CERNET circuit between Beijing and Los Angeles
9
Figure 6: Aggregate traffic using maximum daily values on the 10G CERNET circuit between Beijing and Los Angeles. Gaps in data are periods where the polling server is unable to contact the switch to gather data.
C. Trouble Tickets During this year, there were 6 reported downtimes, five scheduled and one unscheduled. Table 1 shows detail on the unscheduled downtime. Table 2 shows the detail on the scheduled downtime. Additional trouble ticket information is available at https://tick.globalnoc.iu.edu/fp_tools/public_ticket_viewer/index.cgi by entering the ticket number. The unscheduled outage on October 31, 2014, (ticket 1499:62) on the TransPAC Core Node experienced was due to maintenance performed by Pacwave during which the buffering capability of the line cards was overwhelmed. This outage identified an issue with the core equipment that eventually resulted in an upgrade to minimize any potential re-‐occurrence.
Ticket
Number Customer
Impact Network Impact Title Outage Type Source Of
Impact Start Time
(UTC) End Time
(UTC)
1499
1-Critical 1-Critical Brief Outage Resolved - TransPAC Core Node LOSA
Unannounced Maintenance
Vendor 10/31/2014 7:26 AM
10/31/2014 7:42 AM
Table 1: Unscheduled maintenance tickets for the TransPAC this quarter
10
Ticket Number
Customer Impact
Network Impact Title Maintenance
Type Source
Of Impact
Start Time (UTC)
End Time (UTC)
1438
4-Normal 3-Elevated
Maintenance 1 of 2 Completed - TransPAC Backbone TOKY-TransPAC LOSA
Circuit Vendor 03/06/2014 11:39 AM
03/06/2014 12:50 PM
1507
4-Normal 2-High Emergency Maintenance Completed - TransPAC Peer PacWave (All Connections via PacWave)
Hardware
Internal
11/13/2014 3:36 PM
11/13/2014 5:48 PM
1520
4-Normal 2-High Maintenance Completed - TransPAC Core Node LOSA Software and Hardware Upgrade
Hardware Internal 12/12/2014 3:03 PM
12/12/2014 3:13 PM
1532
3-Elevated 3-Elevated
Maintenance Complete - TransPAC Core Node Hardware Upgrade
Hardware Internal
01/21/2015 5:15 PM
01/21/2015 6:25 PM
1537
3-Elevated 2-High Maintenance 1 of 2 Completed - TransPAC Backbone TOKY-TransPAC LOSA
Circuit Vendor 01/27/2015 11:12 AM
01/27/2015 11:33 AM
Table 2: Scheduled maintenance tickets for the TransPAC circuits this year
D. Downtime Table 3 shows the downtime information for TransPAC core nodes and Table 4 Shows the downtime for TransPAC Circuits. The single unscheduled outage on October 31, 2014, (ticket 1499:62) on the TransPAC Core Node was due to maintenance performed by Pacwave. During this maintenance the buffering capability of the line cards was overwhelmed. This outage identified an issue with the core equipment that eventually resulted in an upgrade to minimize any potential re-‐occurrence.
TransPAC Core Nodes Down Time Reporting Period Availability 52 Week Availability
TransPAC MX480 - LA 0 hr 16 min 100.00% 100.00%
Brocade MLXe4 3 hr 32 min 99.96% 99.96%
3410 Ethernet Switch 0 hr 0 min 100.00% 100.00%
OOB Router 0 hr 0 min 100.00% 100.00% Aggregate TransPAC Core Nodes 3 hr 48 min 99.99% 99.99%
Table 3: Reported downtimes for core nodes of TransPAC this year
11
TransPAC Backbone Circuits Down Time Reporting Period Availability 52 Week Availability
TransPAC LOSA-JGN2 LOSA 10GigE 0 hr 16 min 100.00% 100.00%
TransPAC LOSA-Pacific Wave LOSA 10GigE 1 hr 36 min 99.98% 99.98%
TransPAC TOKY-TransPAC2 LOSA 2 hr 58 min 99.97% 99.97% Aggregate All TransPAC Backbone Circuits 4 hr 50 min 99.98161% 99.98166%
Table 4: Reported downtimes for circuits of TransPAC this year
7. Security Events and Activities Basic security measures were maintained over the course of the year, and there were no security incidences to report. We began to examining the possibility of contracting out for some additional international security work, for example, Jacob Farmer, IU, attended the CANS meeting on behalf of the TransPAC project.
8. Year 6 and Transition As discussed in Section 2, we have requested and been granted a one-‐year No-‐Cost-‐Extension for the TransPAC3 project, with a new expected end date of May 2016. We have also been informed that TransPAC4 has been recommended for funding, with an expected start date of March 2015, however with greatly reduced Year 1 funding. Year 6 or TransPAC3 will be spent in transition to the new TransPAC4 project. We split this plan into several components, and address them separately below. Production Traffic: The current 10G NSF-‐funded TransPAC circuit will be supported through the end of calendar year 2015. Between March and August, we will begin negotiations for additional trans-‐Pacific circuits for a 2-‐5 year contract duration. We will ensure a smooth transition of any production traffic if the routes change. Per agreed upon MOU, the production CERNET link will transition to Internet2 on June 1, 2015. However, it could still carry production traffic for the TransPAC project is that is deemed appropriate. Experimental SDN Traffic: . JGN has informed us that in their next funding round, they will concentrate on intra-‐Asia connectivity, so it is possible their 10G experimental link will cease. Our new production links will need to also be able to carry experimental traffic as well. In addition, SDN experiments will be a focus of Year 6 of the project. Monitoring Work: The TransPAC monitoring work increased over Year 5, and is likely to be extended going forward. We have been arguing for a strong PerfSonar presence in APAN, and are committed to ensuring this as the project transitions.
9. Reporting against Objectives Year 5
12
1. Overall a. Planned: New director to review activities and adjust as needed
(Ongoing) b. Additional: Address budget surplus (See Sec 2)
2. Staffing a. Planned: Recruitment of replacement half-‐time engineer for the
project (See Sec 3: Lappa hired July 2015, Addleman hired March 2015)
3. Collaboration, Travel, and Training a. Planned: Attend and speak at major Asian meetings (See Sec 4) b. Planned: Co-‐chair PRAGMA meeting in October with Networking focus
(See Sec 4) c. Planned: Additional TransPAC meetings at APAN in August (See Sec 4) d. Planned: Identify scientific applications using TransPAC more
specifically (Delayed to Year 6) 4. Systems and Software Work
a. Planned: Re-‐focus on monitoring of all circuits (See Sec 5.D) b. Planned: Identify and if possible implement flow-‐analysis tools (See
Sec 5.D) c. Planned: Extend current SDN experiments with JGN (See Sec 5.C)
5. Operational Activities a. Planned: Continue full support of TransPAC circuit (See Sec 6) b. Planned: Continue to shift production system support to production
GlobalNOC teams (See Sec 6)
10. Plans for Year 5 1. Overall
a. New director to review activities and adjust as needed b. Ensure smooth transition to TransPAC4 project
2. Staffing a. Recruitment of additional Outreach Specialist
3. Collaboration, Travel, and Training a. Attend and speak at major Asian meetings b. Identify scientific applications using TransPAC more specifically c. Consider organization of application workshops to support end users
4. Systems and Software Work a. Update LA PoP to support additional monitoring, use of IXIA tester b. Update active monitoring of all circuits c. Update APAN active monitoring systems d. Identify and if possible implement flow-‐analysis tools e. Extend current SDN experiments with JGN
5. Operational Activities a. Continue full support of TransPAC circuit
13
Appendix B: Sample Flow Data from CERnet