Upload
others
View
47
Download
0
Embed Size (px)
Citation preview
Ping End-to-End Reporting
(PingER)presentation for
PERN/PERN2 Special Interest Group for Network Monitoring
http://maggie.seecs.edu.pk http://www-iepm.slac.stanford.edu/pinger
PingER Project• Arguably the world’s most extensive active end-to-end Internet
Performance Project (emphasis Digital Divide)– Partially funded by MoST Pakistan, US State Department
•Last six years - a joint development effort of Stanford University (SLAC) & NUST (SEECS)
•Many SEECS students cut their teeth on it, several research papers & studies
• Results:– Highly successful in quantifying E2E performance (e.g ICFA - SCIC
International Committee for Future Accelerators Standing Committee on Inter-Regional Connectivity)
– Identified & quantified rates of improvement for regions/countries•How far behind, catching up, falling behind
•Many presentations to funding agencies, politicians, NRENs, recommendations
– Case studies identified: fragility of e2e connections, last mile
congestion problems, inefficient routing
http://maggie.seecs.edu.pk http://www-iepm.slac.stanford.edu/pinger
PingER Methodology10 ping requests, each 30 mins
ping response packets
remote host
(typically a
web-server)
monitoring
host
data
repository
once a
day
Measure
Round Trip
Time
& Loss
Internet
http://maggie.seecs.edu.pk http://www-iepm.slac.stanford.edu/pinger
Pinger ArchitectureMonitor hosts send 21 pings each 30 mins to Remote Hosts and cache results
Archive hosts gather data daily, save, analyze & make results available publicly via web
http://maggie.seecs.edu.pk http://www-iepm.slac.stanford.edu/pinger
Pinger Deployment166 countries (98.3% world’s population, >99% world’s connected population)
45 monitoring nodes in 23 countries
890 remote nodes at over 650 sites in over 165 countries (beacons ~ 90)
!
http://maggie.seecs.edu.pk http://www-iepm.slac.stanford.edu/pinger
What is measured?E2E Network Performance Metrics
Round Trip Time (delay/latency) - ms
Packet Loss - %
Throughput (derived) Kbps
Jitter - ms
Unreachability - %
http://maggie.seecs.edu.pk http://www-iepm.slac.stanford.edu/pinger
Why are we here?Deployment of PingER monitoring nodes at the 80+ universities connected via PERN
Facilitate measurement of user-experienceEnd-to-end performance measurement
Urge universities to utilize PERN connectivityHost university servers on PERN links
Provide universities with real-world measurements for analysis and research
Build network monitoring tools
Analyse network measurements
http://maggie.seecs.edu.pk http://www-iepm.slac.stanford.edu/pinger
Why is this important?Problem Diagnosis is challenging
Convince service providers that there is a problem
Need multiple pieces of information from multiple sources
Performance measurements facilitates
Network operations
Allows for trend analysis, planning and provisioning
World View
http://maggie.seecs.edu.pk http://www-iepm.slac.stanford.edu/pinger
Behind Europe6 Yrs: Russia, Latin America 7 Yrs: Mid-East, SE Asia10 Yrs: South Asia11 Yrs: Cent. Asia12 Yrs: Africa
South Asia, Central Asia, and
Africa are in Danger of Falling
Even Farther Behind
!
Throughput as seen from US
Case Study: Pakistan
http://maggie.seecs.edu.pk http://www-iepm.slac.stanford.edu/pinger
Monitoring nodes in Pakistan
LocationNo. of monitoring
nodes
SEECS, NUST, Islamabad 4
NCP, Quaid-e-Azam University, Islamabad 1
COMSATS University, Islamabad 1
Pakistan Education Research Network (PERN), Islamabad 1
Micronet/Nayatel Pakistan, Islamabad 1
http://maggie.seecs.edu.pk http://www-iepm.slac.stanford.edu/pinger
Remote nodes in Pakistan
Remote Node
University/
Organization
Location
Service Provider Available
Bandwidth
End Host
Location
LSE (lahoreschoolofeconomics.edu.pk) Lahore ! ! Lahore
COMSATS (comsats.edu.pk) Islamabad PERN ! Islamabad
BUITMS (buitms.edu.pk) Quetta PERN ! Quetta
SSUET (ssuet.edu.pk) Karachi PERN ! Karachi
UPESH (www.upesh.edu.pk) Peshawar PERN* ! Islamabad
PIEAS (www.pieas.edu.pk) Nilore PERN* ! Islamabad
NUST/SEECS (www.seecs.edu.pk)
formerly NIITIslamabad
Micronet/ Nayatel
(dsl.net.pk) and
PERN
1 - 1.5 Mbps
and
32 Mps
Islamabad
GIKI (www.giki.edu.pk) Topi PERN* ! Topi
UET (uet.edu.pk) Lahore PERN* ! Lahore
HU (hu.edu.pk) Hazara PERN* ! Hazara
PERN (www.pern.edu.pk) Islamabad PERN ! Islamabad
DSL.NET (dsl.net.pk) Islamabad Micronet/ Nayatel
(dsl.net.pk) ! Islamabad
NAYATEL (nayatel.com) Islamabad Micronet/ Nayatel
(dsl.net.pk) Islamabad
SDNPK (wb.sdnpk.org) Islamabad Cyber NET
(cyber.net.pk) ! Islamabad
http://maggie.seecs.edu.pk http://www-iepm.slac.stanford.edu/pinger
Deployment in Pakistan
http://maggie.seecs.edu.pk http://www-iepm.slac.stanford.edu/pinger
Average RTT (ms)
!"
#!!"
$!!"
%!!"
&!!"
'!!"
(!!"
)!!"*+,-!)"
./0-!)"
120-!)"
./3-!)"
456-!)"
457-!)"
158-!)"
9+2-!)"
:;<-!)"
=>?-!)"
@+;-!)"
4/6-!A"
*+,-!A"
./0-!A"
120-!A"
./3-!A"
456-!A"
457-!A"
158-!A"
9+2-!A"
:;<-!A"
=>?-!A"
@+;-!A"
4/6-!B"
!"#$%%&'()*&
!"#$%%&+)&),,-&./0(&122314&561%&7,8&9:&;&<+-&9=&
9@=CD"
CEF="
@9GH=EI"
G9E"
J:.91I9"
=1K1IEG"
LMN.I9"
99MEI"
MCE9O"
CNE19"
PNDN"
MEI"
OM"
http://maggie.seecs.edu.pk http://www-iepm.slac.stanford.edu/pinger
Average RTT (ms)
!"
#!"
$!!"
$#!"
%!!"
%#!"
&!!"
&#!"
'!!"
'#!"
#!!"
()*+,"
+-.*"
)(/0*-1"
/(-"
234(51("
*5651-/"
78941("
((8-1"
8+-(:"
+9-5("
*8(1"
;9,9"
8-1"
:8"
!"#$%&''%()*+%
,-./0-123%4/02*%
!"215#2%&''%5*%)25*6123%78%9/.#:&%)-./0-1/.#%*/02*%
;/0</.%95=/*05.%>5.%?@@A%B%>5.%?@@C%
(--2(<"*8(1"
*2+<"=58"
http://maggie.seecs.edu.pk http://www-iepm.slac.stanford.edu/pinger
Packet Loss (%)
!"
#"
$"
%"
&"
'!"
'#"
'$"
()*+!,"
-./+!,"
01/+!,"
-.2+!,"
345+!,"
346+!,"
047+!,"
8)1+!,"
9:;+!,"
<=>+!,"
?):+!,"
3.5+!&"
()*+!&"
-./+!&"
01/+!&"
-.2+!&"
345+!&"
346+!&"
047+!&"
8)1+!&"
9:;+!&"
<=>+!&"
?):+!&"
3.5+!@"
!"#$%&'()**'+,-'
!"#$%&'()**'"*'*%%.'/0)1'233425'6728'9%:';<'='>".';?'
8?<AB"
ACD<"
?8EF<CG"
E8C"
H9-80G8"
<0I0GCE"
JKL-G8"
88KCG"
KAC8M"
ALC08"
NLBL"
KCG"
MK"
http://maggie.seecs.edu.pk http://www-iepm.slac.stanford.edu/pinger
Packet Loss (%)
!"
#"
$"
%"
&"
'"
("
)"
*+,-!)"
./0-!)"
120-!)"
./3-!)"
456-!)"
457-!)"
158-!)"
9+2-!)"
:;<-!)"
=>?-!)"
@+;-!)"
4/6-!A"
*+,-!A"
./0-!A"
120-!A"
./3-!A"
456-!A"
457-!A"
158-!A"
9+2-!A"
:;<-!A"
=>?-!A"
@+;-!A"
4/6-!B"
!"#$%&'()**'+,-'
!"#$%&'()**'.%%/'01)2'.334.5'6%7'89':';"/'8<'
$'<C"2+0;+6D7+"
.+EF/6"
)'<C"2+0;+6D7+"
http://maggie.seecs.edu.pk http://www-iepm.slac.stanford.edu/pinger
Packet Loss (%)
!"!!#
$"!!#
%"!!#
&"!!#
'"!!#
("!!#
)"!!#
*"!!#
+"!!#
,"!!#
$!"!!#
-./01#
023/#
.-4"/25#
4-2#
678-95-#
/9:9524#
;<=85-#
--<25#
<02->#
0=29-#
/<-5#
?=1=#
<25#
><#
!"#$"%&'(")*+,,)-./)
0+%1&+#"2)31&",)
!'$4"&)*+,,)',)5"',6#"2)78)!1%(9:))
5+%1&+#1%(),1&",);1&<1%)!'41,&'%)
='%)>??@)A)='%)>??B)
-226-@#/<-5#
/60@#A9<#
http://maggie.seecs.edu.pk http://www-iepm.slac.stanford.edu/pinger
Throughput (Kbps)
!"!!#
$!!!"!!#
%!!!"!!#
&!!!"!!#
'!!!"!!#
(!!!"!!#
)!!!"!!#
*!!!"!!#
+!!!"!!#
,-./0#
/12.#
-,3".14#
3,1#
567,84,#
9:;74,#
,,:14#
:/1,<#
/;18,#
.:,4#
=;0;#
:14#
!"#$!%&'()%*(+$,-$./+0102$
3'4/+'&56$7/+50$
!"#$!%&'()%*(+$80$9580(&56$.:$#/4);<$9'4/+'&/4)$
0/+50$=/+%/4$#8-/0+84$
$>84$?@@A$B$>84$?@@C$ ,115,>#.:,4#
.5/>#?8:#
Mathis et. alThroughput or Rate < (MSS ~1460 Bytes/RTT)*(1 / sqrt(loss))
The macroscopic behavior of the TCP congestion avoidance algorithm by Mathis, Semke, Mahdavi & Ott in Computer Communication Review, 27(3), July 1997
Traceroute resultsLast mile effects
http://maggie.seecs.edu.pk http://www-iepm.slac.stanford.edu/pinger
Unreachability & Loss
!
http://maggie.seecs.edu.pk http://www-iepm.slac.stanford.edu/pinger
Average RTT (ms)
http://maggie.seecs.edu.pk http://www-iepm.slac.stanford.edu/pinger
Average RTT (ms)
!"
#!!"
$!!"
%!!"
&!!"
'!!"
(!!"
)!!"
*+,-!
%"./0-!
&"123-!&"
4+5-!
&"*+,-!
&"./0-!
'"123-!'"
4+5-!
'"*+,-!
'"./0-!
("123-!("
4+5-!
("*+,-!
("./0-!
)"123-!)"
4+5-!
)"*+,-!
)"./0-!
6"123-!6"
4+5-!
6"*+,-!
6"
!"#$%%&'()*&
!"#$%%&+,-&./01)2/3&/)&)443&+-,(&56!7&
$'78"5+0,+39:+"
.+;</3"
)'78"5+0,+39:+"
http://maggie.seecs.edu.pk http://www-iepm.slac.stanford.edu/pinger
InferencesNational backbone is well provisioned
Last mile effects
congestion & losses
power outages
http://maggie.seecs.edu.pk http://www-iepm.slac.stanford.edu/pinger
What to do?Deploy PingER
Monitoring & reverse traceroute scripts to be installed
Node owned by the respective university
PingER requirementsNode should run Linux
Node name must be registered with a DNS
Web (http) and ICMP (ping traceroute) traffic should be allowed
Node preferably close to the border (DMZ), can be within the local LAN
Minimal disk storage and network access
http://maggie.seecs.edu.pk http://www-iepm.slac.stanford.edu/pinger
ConclusionPerformance monitoring is vital for
network operations,
trend analysis,
planning and provisioning
Provides opportunities for computer network research
We encourage all universities to participate
and utilize the PERN connectivity
Thank youQuestions?
http://maggie.seecs.edu.pk http://www-iepm.slac.stanford.edu/pinger
Selected ProjectsAnomalous Event Detector
TULIP - Trilateration Utility for Locating IP Addresses
PingER validation toolkit
PingER regional case studiesAfrica, Pakistan, South Asia, East Asia, Latin America etc
ICFA reports 2007, 2008 and 2009
ViPER - PingER Visualization
PingER motion charts
PingER executive plots
Google maps for PingER
and many others ...
http://maggie.seecs.edu.pk http://www-iepm.slac.stanford.edu/pinger
TULIP
!
Trilateration Utility for locating IP addresses
http://maggie.seecs.edu.pk http://www-iepm.slac.stanford.edu/pinger
PingER Motion Charts