80 likes | 175 Views
USATLAS perfSONAR Status. Shawn McKee for the USATLAS Throughput Working Group. perfSONAR. perfSONAR is being developed as a common network measurement and diagnostic infrastructure by a consortium including ESnet, Internet2, CANARIE, GEANT and other world-wide R&E networks.
E N D
USATLAS perfSONAR Status Shawn McKee for the USATLAS Throughput Working Group
perfSONAR • perfSONAR is being developed as a common network measurement and diagnostic infrastructure by a consortium including ESnet, Internet2, CANARIE, GEANT and other world-wide R&E networks. • Many network PoPs (Points-of-Presence) are being instrumented with perfSONAR to help enable quick problem isolation • USATLAS would like to benefit from this infrastructure by deploying (on the identical hardware) perfSONAR instances at our primary resource sites • Such a deployment yields a consistent measurement of the network between resource locations and provides test points on demand for problem diagnosis.
USATLAS perfSONAR Plan • USATLAS agreed to have a milestone of Fall 2008 for deployment of a set of USATLAS perfSONAR boxes • Each “site” (distinct resource location) should purchase two of the agreed upon standard systems from KOI computing and co-locate them with their resources. • The two boxes have different roles: one is to monitor one-way delays between itself and its peers and the other is used to make bandwidth measurements. • Additional services provide test points at the site.
USATLAS perfSONAR spreadsheet Missing info on yellow status lines
Status Today • We have boxes deployed completely at BNL, AGLT2, SWT2, partial deployment at MWT2 and are waiting for NET2 and WT2 to finish their installations. • We don’t yet have a consistent deployment: • Need Communities of Interest set to ‘LHC USATLAS’ • Need to configure scheduled tests at each site • Only BNL runs regular tests so far • Tests should have regular 15 second bandwidth test to all peer sites • Need to setup latency measurements to peers as well • AGLT2 will set this up and document it • We have requested some changes to default installation: • Bandwidth summary page at default URL (matrix of results) • USATLAS peer list with status (OK/NOTOK) • See next page for examples
perfSONAR and Throughput • Within the Throughput group, we are trying to define what types of monitoring and measurements are useful for finding throughput problems or bottlenecks • perfSONAR is intended to provide an independent, ongoing measurement of the network between sites • WAN/LAN issues can be found quickly. • Diagnostics can be run on-demand • Throughput depends upon many factors: network, I/O of end-systems, hardware configurations, applications, system tunings. • Next up is a regularly scheduled “test dataset” copy to provide an “end-to-end” measurement of the throughput (inclusive of the factors above) • Goal is to identify the minimal set of tests and monitored data to quickly isolate throughput problems
Timeline • Existing perfSONAR sites should reconfigure their “Communities of Interest” this week (LHC USATLAS) • AGLT2_UM will document steps needed to setup regular USATLAS peer tests for perfSONAR by next throughput meeting. • Hiro (& Jay for graphics?) will create a prototype standardized dataset copy test to measure end-to-end throughput by the next meeting. • Missing perfSONAR sites need to update the spreadsheet to provide their timelines ASAP