UC Irvine OptIPuter Link

1
UC Irvine OptIPuter Link

• Dana Roode
• UCI Network and Academic Computing Services
• Dana.Roode_at_uci.edu

2
Network Academic Computing Services - NACS

• UCInet
• 23,244 connections (1,773 switches in 601
  equipment closets serving 219 buildings)
• Research computing and networking support
• NACS support of Optiputer
• Under partnering agreement with Calit2
• Desire to understand ins and outs of dedicated
  research networks
• My role

3
Initial UCI Link Efforts

• Exploration of GRE tunneling (Early 2004)
• Initial UCI Optiputer PI funding did not cover
  network infrastructure
• Deployed equipment did not support tunneling,
  idea dropped
• UCI connection via CalREN HPR (June 2004)
• Performance extremely poor (7-8 mbps)
• Led to investigation of bottlenecks on and off
  campus
• UCI testing (iPerf) showed backbone 5500s to be a
  significant bottleneck
• 5500s later replaced with 6513s in backbone
  upgrade

4
CalREN-XD Optiputer Circuit

• CENIC crafted approach to make 1gbps of 10gbps
  SD-LA-Tustin link available for Optiputer
• CENIC/UCSD/UCI collaborated to engineer Optiputer
  connectivity paths on each campus
• Wave 1 Link became available to UCI Engineering
  Gateway in September 2004
• Wave 2 Link available in new UCI Calit2
  building in summer 2005 (with switch funded by
  Charlie Zender ESMF Optiputer grant)

5
Created 09-27-2005 by Garrett Hildebrand Updated
11-03-2005 by Jessica Yu
at UC Irvine
ONS 15540 WDM at UCI campus MPOE (CPL)
10 GE DWDM Network Line
Wave-2 layer-2 GE. UCSD address space
137.110.247.210-222/28
Floor 4 Catalyst 6500
Engineering Gateway Building,
SPDS
Viz Lab
Floor 3 Catalyst 6500
Catalyst 3750 in 3rd floor IDF
Wave-1 UCSD address space 137.110.247.242-246
NACS-reserved for testing
Floor 2 Catalyst 6500
Catalyst 3750 in NACS Machine Room (Optiputer)
ESMF
10 GE
Wave 1 1GE
Wave 2 1GE
Catalyst 3750 in CSI
MDF Catalyst 6500 w/ firewall, 1st floor closet
6
Early Experiences

• System connections made, NACS is done, right?
• Wrong.
• Lingering negative perceptions from earlier
  experiences
• Expectations of performance not met
• Metrics and testing
• Remaining performance issue
• UCSD -gt UCI 885 mbits/sec
• UCI -gt UCSD 60 mbits/sec

7
Lessons Learned

• Communication is paramount
• Goals, needs and open issues
• Communication must span researchers, support
  staff, and campuses
• Pro-active communication by staff is of great
  value
• Expectations and perceptions must be factored in
• Bandwidth performance testing is essential in
  implementation of high-speed networks
• Tune the network, but tune the computer too

8
Whats Next?

• Correct asymmetrical UCIgtUCSD performance issue
• Ensure all Optiputer researchers have what they
  need to proceed with their research at 1gbps
• Review topology and equipment at UCSD and UCI
• Dedicated Optiputer circuit / hardware in UCI
  Calit2 building?
• Optical switch?
• Upgrade UCIlt-gtUCSD path to 10G?
• How does this fit in with proposed larger-scale
  dedicated optical path networks and CalREN-XD?
• What locations at UCI will be served?