Presentation is loading. Please wait.

Presentation is loading. Please wait.

1 IEPM / PingER project & PPDG Les Cottrell – SLAC Presented at the NGI workshop, Berkeley, 7/21/99 Partially funded by DOE/MICS Field Work Proposal on.

Similar presentations


Presentation on theme: "1 IEPM / PingER project & PPDG Les Cottrell – SLAC Presented at the NGI workshop, Berkeley, 7/21/99 Partially funded by DOE/MICS Field Work Proposal on."— Presentation transcript:

1 1 IEPM / PingER project & PPDG Les Cottrell – SLAC Presented at the NGI workshop, Berkeley, 7/21/99 Partially funded by DOE/MICS Field Work Proposal on Internet End-to-end Performance Monitoring (IEPM)

2 2 Overview of Mechanism 18 Monitor sites, 10 countries 1200 monitor-remote-site pairs 382 unique hosts, 29 countries Measure response, jitter, loss, reachability Data goes back 4.5 years 1 Million probes of Internet/day Treats Internet as black box Uses existing infrastructure (ping) Low cost, well understood Provides active end-to-end measurements Ping WWW HEPNRC SLAC Monitor Remote

3 3 Min ping response HTTPGetHTTPGet Relates to Web performance small files dominated by RTT BW TCP < (MSS/RTT)*(1/sqrt(loss)) GET - 2 * min (ping RTT) IQR=220ms Examples of relevance to applications

4 4 Deployment in HEP Over 50% of HEP collaborator sites are explicitly monitored by PingER –Atlas (37%), BaBar(68%), Belle(23%), CDF(73%), CMS (31%), D0(60%), Zeus (35%) –Created focussed PingER pages for BaBar, CDF, D0... Remainder represented by beacon sites –Selected to represent countries/R&E nets –About 50 beacon sites in 27 countries

5 5 Choose metric loss, RTT, variability, reachability Choice of time ticks hour, daily, monthly ; can also select day, month etc. Choice of group Export data to Excel Drill down to plots Web Interface Other stuff http://www.slac.stanford.edu/ /xorg/iepm/pinger/table.html Sort on columns Help

6 6 September 1 to December 31 1998 0 50 0 200 400 % packet loss ITU G114 300msec RTT limit for voice Effect of STAR-TAP on KEK.jp SLAC

7 7 Performance Trends Bandwidth TCP < 1460/(RTT * sqrt(loss))

8 8 Comparisons of Networks

9 9 Problem areas Germany bad with Canada &.edu yet good with ESnet Russia (W) bad to everywhere yet good with Esnet China poor with most UK varies with time, bandwidth increased by factor of ~80 in 4 years Peering is critical, avoid congestion points, reduce number of NSPs, reliance on your NSP

10 10 Ping details CERN=>SLACSLAC => CERN Surveyor reports for May 11, 99

11 11 Correlation of pings with Surveyor

12 12 Correlation of Surveyor & PingER PingER CERN SLAC May 11, 1999 CERN=>SLACSLAC => CERN Surveyor reports for May 11, 99

13 13 Ping “Jitter” measurement

14 14 Ping “jitter” between SLAC & CERN, DESY, FNAL 12/14-15, 1999 ITU threshold for jitter (75 msec) Big changes during day IQR(Response) ~ IQR(IPD) FNAL order of magnitude better CERN worst during Euro day DESY worst during US day

15 15 PPDG work Monitor testbed network between LBNL & SLAC to see impact of loading and the effect of QoS –what is the domain of ping usefulness, how does it relate to more heavyweight methods (e.g. packet sniffing), what does it show, how to instrument –Simple ttcp loading, using window size to crank up the load

16 16 Value of PingER/Ping Lightweight –uses existing infrastructure - nothing to install @ remote sites –low network (100bytes/sec/pair), disk & cpu requirements –ping itself is mature & well understood Complementary to more in-depth measurements –good agreement with Surveyor, relation to wire times understood –still requires careful validation, calibration & correlation duplicate pings (0.2%), out of order packets (<0.02%), wire time vs. host reported time, rate limiting, implementation shortcomings

17 17 Value of PingER Has excellent grouping & long term reporting –allows comparisons of affinity groups (ISP, location, experiment, collaboration …) –good for baselines, trends, planning, setting expectations where is an application (e.g. Internet voice, FTP of code or event data) likely to work where to locate code development hosts, regional centers Used for selecting routes, choosing ISPs, setting & verifying SLAs, prioritizing which links to upgrade, can application get performance it needs

18 18 More Information WAN Monitoring at SLAC has lots of links –http://www.slac.stanford.edu /comp/net/wan-mon.htmlhttp://www.slac.stanford.edu /comp/net/wan-mon.html Tutorial on WAN Monitoring (including methods, RTT, jitter, loss & QoS thresholds, validation, comparisons with Surveyor etc.) –http://www.slac.stanford.edu /comp/net/wan-mon/tutorial.htmlhttp://www.slac.stanford.edu /comp/net/wan-mon/tutorial.html PingER History tables –http://www.slac.stanford.edu /cgi-wrap/pingtable.plhttp://www.slac.stanford.edu /cgi-wrap/pingtable.pl Internet End-to-end Monitoring (IEPM) home page –http://www-iepm.slac.stanford.edu/http://www-iepm.slac.stanford.edu/

19 19 Deployment of measurement projects


Download ppt "1 IEPM / PingER project & PPDG Les Cottrell – SLAC Presented at the NGI workshop, Berkeley, 7/21/99 Partially funded by DOE/MICS Field Work Proposal on."

Similar presentations


Ads by Google