1

A new infrastructure for high throughput network and

application performance measurement.

Les Cottrell – SLACPrepared for the IPAM meeting, UCLA Mar 19 ‘02

http://www.slac.stanford.edu/grp/scs/net/talk/ipam-mar02.html

Partially funded by DOE/MICS Field Work Proposal on Internet End-to-end Performance Monitoring (IEPM), also supported by IUPAP

2

PingER deployment• Measurements from

– 34 monitors in 14 countries– Over 600 remote hosts

– Over 72 countries

– Over 3300 monitor-remote site pairs

– Measurements go back to Jan-95

– Reports on RTT, loss, reachability, jitter, reorders, duplicates …

• Countries monitored– Contain 78% of world population

– 99% of online users of Internet

• Lightweight (100bps/host pair)– Very useful for inter-regional and poor links, need more intensive for high

performance & Grid sites

3

User interface• Eu Data Grid WP7 have

extended PingER to throughput measurements (iperf, udpmon …) – Manchester, Daresbury,

IN2P3 & UCL

• SLAC has added iperf, bbcp, bbft measurements to PingER tables– www-iepm.slac.stanford.edu/cgi-

wrap/pingtable.pl?dataset=iperf

Metrics:pipecharIperfBbcpbbftp

4

Effect of Sep 11 ’02 on unreachability• For most sites (median) the effect was not noticeable

• There was a big peak in unreachability for 10 sites (5%) on Sep 11 &12

• All 5 Italian sites lost connectivity for about 24 hours, Padova took 8 days to stabilise

• BNL took 6 days to stabilize• 5 sites (2.5%) took 10 days to

recover• 2 sites (MX & Montreal) had not

recovered by end of month, may have had nothing to do with WTC– MX found blocked ping access in Nov

SNVORN

5

Romania performance from US

6

Losses: World by region, Jan ‘02• <1%=good, <2.5%=acceptable, < 5%=poor, >5%=bad

• Russia, S America bad

• Balkans, M East, Africa, S Asia, Caucasus poor

Monitored Region \ Monitor Country

BR (1)

CA (2)

DK (1)

DE (1)

HU (1)

IT (3)

JP (2)

RU (2)

CH (1)

UK (3)

US (16) Avg Region

Avg -(H

Avg NA + WEU + JP Pairs

COM 0.2 0.3 0.3 0.2 COM 0.27 23

Canada 1.8 1.6 0.3 0.5 9.0 0.3 1.4 21.7 0.7 0.7 0.5 3.5 Canada 0.74 126

US 0.4 2.6 0.2 0.3 8.0 0.1 1.4 13.8 0.3 1.3 0.9 2.7 US 0.88 2149

C America 0.9 0.9 C America 0.89 19

Australasia 0.8 1.8 1.3 Australasia 1.30 18

E Asia 1.2 3.5 1.0 1.1 9.0 0.9 2.0 5.2 1.5 1.4 1.5 2.6 E Asia 1.61 215

Europe 0.4 5.6 0.3 0.5 5.4 0.4 1.3 15.5 1.1 1.0 1.0 2.9 Europe 1.38 852

NET 1.7 6.2 1.0 1.3 8.0 1.6 3.6 21.9 0.7 0.8 0.9 4.3 NET 2.00 85

FSU- 4.5 0.5 9.8 0.5 1.6 11.2 4.3 1.2 2.0 4.0 FSU- 2.09 48

Balkans 3.8 3.8 Balkans 3.83 109

Mid East 4.6 1.4 3.0 8.5 2.8 3.2 11.8 2.0 2.5 2.1 4.2 Mid East 2.70 57

Africa 5.8 1.5 12.0 1.2 4.2 11.9 2.0 1.9 2.5 4.8 Africa 2.72 45

Baltics 5.3 0.8 2.3 7.7 2.2 3.5 10.8 4.8 2.1 3.9 4.3 Baltics 3.12 67

S Asia 1.6 7.3 0.1 3.1 9.2 3.0 3.9 17.9 1.5 3.1 3.0 4.9 S Asia 3.12 97

Caucasus 3.2 3.2 Caucasus 3.22 19

S America 24.1 11.3 0.6 0.9 6.7 12.9 7.7 23.0 9.3 1.1 6.6 9.5 S America 6.30 203

Russia 35.9 24.1 22.2 13.4 23.8 21.7 13.6 0.7 8.7 24.1 12.7 18.3 Russia 17.57 91

Avg 7.5 6.9 2.8 2.4 9.8 3.7 3.9 13.8 3.1 3.2 2.8 4.4 Avg 3.16

Pairs 64 144 54 67 70 203 190 114 209 192 1990 Pairs

7

Throughput quality improvementsTCPBW < MSS/(RTT*sqrt(loss)) (1)

(1) Macroscopic Behavior of the TCP Congestion Avoidance Algorithm, Matthis, Semke, Mahdavi, Ott, Computer Communication Review 27(3), July 1997

China

80% annual improvement ~ factor 10/4yr

SE Europe not keeping up

8

Quality improvement

seen from SLAC & NASA

• NASA results courtesy of Andy Germain, NASA, GSFC

9

We need to better understand• PingER losses insufficient for high perf links

– Need more measurements & ping losses != TCP losses

• Closer to applications, e.g. FTP• Understand how to make throughput measurements:

– Duration & frequency (balance impact against granularity needed), – Windows and or vs parallel streams, – OS dependencies, cpu utilization, interface speeds, security (e.g. ssh)– Impact on others, variability on different time-scales– Can we use QBSS, can/should application self limit?– How well does simulation work, how to improve?– How to relate to simpler measurements– How does file transfer work compared to iperf?– Is compression useful and when?– How to steer applications

10

Even with big windows (1MB)

still need multiple streams

• Above knee performance still improves slowly, maybe due to TCP fair share as add more streams

• ANL, Caltech & RAL reach a knee (between 2 and 24 streams) above this gain in throughput slow

11

IEPM-BW: Main issues being addressed• Provide a simple, robust infrastructure for:

– Continuous/persistent and one-off measurement of high network AND application performance

– management infrastructure – flexible remote host configuration • Optimize impact of measurements

– Duration, frequency of active measurements, and use passive• Integrate standard set of measurements including: ping, traceroute, pipechar,

iperf, bbcp …• Allow/encourage adding measure/app tools

• Develop tools to gather, reduce, analyze, and publicly report on the measurements:– Web accessible data, tables, time series, scatterplots, histograms, forecasts …

• Compare, evaluate, validate various measurement tools and strategies (minimize impact on others, effects of app self rate limiting, QoS, compression…), find better/simpler tools, choose best set

• Provide simple forecasting tools to aid applications and to adapt the active measurement frequency

• Provide tool suite for high throughput monitoring and prediction

12

IEPM-BW Deliverables• Understand and identify resources needed to achieve high

throughput performance for Grid and other data intensive applications

• Provide access to archival and near real-time data and results for eyeballs and applications:– planning and expectation setting, see effects of upgrades– assist in trouble-shooting problems by identifying what is

impacted, time and magnitude of changes and anomalies– as input for application steering (e.g. data grid bulk data transfer),

changing configuration parameters– for prediction and further analysis

• Identify critical changes in performance, record and notify administrators and/or users

• Provide a platform for evaluating new network tools (e.g. pathrate, pathload, GridFTP, INCITE, UDPmon …)

• Provide measurement/analysis/reporting suite for Grid & hi-perf sites

IEPM-BW Deployment

14

Results so far 1/2• Reasonable estimates of

throughput achievable with 10 sec iperf meas.

• Multiple streams and big windows are critical– Improve over default

by 5 to 60.– There is an optimum

windows*streams• Continuous data at 90

min intervals from SLAC to 33 hosts in 8 countries since Dec ‘01

15

Results so far 2/2

0 400

Dis

k M

bps

80

Iperf Mbps

• 1MHz ~ 1Mbps

• Bbcp mem to mem tracks iperf

• BBFTP & bbcp disk to disk tracks iperf until disk performance limits

• High throughput affects RTT for others

– E.g. to Europe adds ~ 100ms

– QBSS helps reduce impact

• Archival raw throughput data & graphs already available via http

16

E.g. Iperf vs File copy (mem-to-mem)

0 400

350

Iperf TCP Mbits/s

Fil

e co

py m

em to

mem

File copy mem to mem ~ 72% iperf

17

E.g. Iperf vs file copy disk to disk

0 400

100

Iperf TCP Mbits/s

File copy dis k-to- dis k

Fast Ethernet

OC3

Disklimited

Over 60Mbits/s iperf >> file copy

18

Windows and Streams 1/2• Now well accepted that multiple streams and/or big

windows are important to achieve optimal throughput• Can be unfriendly to others• Optimum windows & streams changes with changes in path• So working on new methods to predict both windows &

streams automatically & optimize– Netest from LBNL uses UDP& TCP

• Predicts window sizes and streams using packet streams, UDP & TCP

• Needs to be validated

– Brute force: run iperf multi windows & streams & find optimum, need to automate measurement & analysis

– Only need to run occasionally or when suspect change • e.g. Step change in number of hops or RTT or other measurement

19

Optimizing streams & flows vs losses

Throughput = i=1,n(MSSi/(RTTi/sqrt(lossi)) (1)

~ (MSS/RTT)(1/sqrt(lossi) ~ [n*MSS]/(RTT*sqrt(loss))So problem reduces to optimizing throughput w.r.t. loss (2)

(2) “The End-to-End Performance Effects of parallel TCP sockets on a Lossy Wide-Area network”, Hacker & Athey, Proc 16th Int Parallel & Distributed ProcessingSymposium, Ft Lauderdale, FL April 2002

20

Effect of saturation

• If keep below knee then RTT stays low

21

Estimated vs

measured throughput

• Est BW~C*Streams*MSS/(RTT*sqrt(loss)) (Mathis et. al. & Hacker)• Measure retransmissions using Web100, loss~Retransmissions/total

packet• Note high degree of correlation (R^2 >0.9)• If window large enough ((>=128KB) then ~ common relation,

window threshold varies from link to link

22

Forecasting• Given access to the data one can do real-time forecasting for

– TCP bandwidth, file transfer/copy throughput• E.g. NWS, Predicting the Performance of Wide Area Data Transfers by

Vazhkudai, Schopf & Foster

• Developing simple prototype using average of previous measurements– Validate predictions versus observations– Get better estimates to adapt frequency of active measurements &

reduce impact• Also use ping RTTs and route information

– Look at need for diurnal corrections– Use for steering applications

• Working with NWS for more sophisticated forecasting• Can also use on demand bandwidth estimators (e.g.

pipechar, but need to know range of applicability)

23

Forecast results

33 nodes Iperf TCP

Bbcp mem

Bbcp disk

bbftp pipechar

Average % error

13%

+- 11%

23%

+- 18%

15%

+-13%

14%

+-12%

13%

+-8%

Predict=Moving average of last 5 measurements +- Iperf TCP throughput SLAC to Wisconsin, Jan ‘02

100

60M

bits

/s

x

PredictedObserved

% average error = average(abs(observe-predict)/observe)

24

Passive (Netflow) 1/2• Use Netflow measurements from border router

– Netflow records time, duration, bytes, packets etc./flow

– Calculate throughput from Bytes/duration for big flows (>10MB)

– Validate vs. iperf, bbcp etc.

– No extra load on network, provides other SLAC & remote hosts & applications, ~ 10-20K flows/day, 100-300 unique pairs/day

25

Passive 2/2M

bits

/s

Date0

450

Iperf SLAC to Caltech (Feb-Mar ’02)

0

80M

bits

/s

Date

Bbftp SLAC to Caltech (Feb-Mar ’02)

+ Active+ Passive

+ Active+ Passive

Iperf matches well

BBftp reports underwhat it achieves

26

CompressionBbcp file throughput from SLAC to CERN

by compression factor, Dec 5 '01

0

10

20

30

40

50

0 2 4 6 8 10Compression factor

Bb

cp

thro

ug

hp

ut

Mb

its/

s

0

2

4

6

8

10

Co

mp

ress

ion

rat

io

•60Mbyte Objectivity file, using zlib, 8 streams, 64KB window•Can improve throughput on this link with these hosts (Sun Ultra Sparcs with 360MHz cpus) by more than a factor of 2.•Working on formula f(file compressibility, host MHz, link Mbits/s) whether compression is worth it

27

Impact on Others• Make ping measurements with & without iperf

loading– Loss loaded(unloaded)– RTT

• Looking at how to avoid impact: e.g. QBSS/LBE, application pacing, control loop on stdev(RTT) reducing streams, want to avoid scheduling

28Pings to host on show floor

Priority: 9+-2 ms

BE: 18.5+-3ms

QBSS: 54+-100 ms

SC2001 QBSS demo• Send data from 3 SLAC/FNAL booth computers (emulate a tier 0 or

1 HENP site) to over 20 other sites with good connections in about 6 countries– Iperf TCP throughputs ranged from 3Mbps to ~ 300Mbps

• Saturate 2Gbps connection to floor network– Maximum aggregate throughput averaged over 5 min. ~ 1.6Gbps

• Apply QBSS to highest performance site, and rest BE

100

0Time

Mbi

ts/s

Iperf TCP Throughput Per GE interface

QBSSNo QBSS

5mins

29

Possible HEP usage of QBSS• Apply priority to lower volume interactive

voice/video-conferencing and real time control• Apply QBSS to high volume data replication• Leave the rest as Best Effort• Since 40-65% of bytes to/from SLAC come from a

single application, we have modified to enable setting of TOS bits

• Need to identify bottlenecks and implement QBSS there

• Bottlenecks tend to be at edges so hope to try with a few HEP sites

30

Possible scenario• BaBar user wants to transfer large volume (e.g. TByte) of data from

SLAC to IN2P3: – Select initial windows and streams from a table of pre-measured optimal values,

or use an on demand tool (extended iperf), or reasonable default if none available

• Application uses data volume to be transferred and simple forecast to estimate how much time is needed

– Forecasts from active archive, Netflow, on demand use one-end bandwidth estimation tools (e.g. pipechar, NWS TCP throughput estimator)

– If estimate duration is longer than some threshold, then more careful duration estimate is made using diurnal forecasting

– Application reports to user who decides whether to proceed

• Application turns on QBSS and starts transferring – For long measurements, provide progress feedback, using progress so far,

Netflow measurements of this flow for last few half hours, diurnal corrections etc.

– If falling behind required duration, turn off QBSS, go to best effort – If throughput drops off below some threshold, check for other sites

31

Experiences• Getting ssh accounts and resources on remote hosts

– Tremendous variation in account procedures from site to site, takes up to 7 weeks, requires knowing somebody who cares, sites are becoming increasingly circumspect

– Steep learning curve on ssh, different versions • Getting disk space for file copies (100s Mbytes)

– Diversity of OSs, userids, directory structures, where to find perl, iperf ..., contacts • Required database to track

– Also anonymizes hostnames, tracks code versions, whether to execute command (e.g. no ping if site blocks ping) & with what options,

• Developed tools to download software and to check remote configurations – Remote server (e.g. iperf) crashes:

• Start & kill server remotely for each measurement – Commands lock up or never end: – Time out all commands

• Some commands (e.g. pipechar) take a long time, so run infrequently • AFS tokens to allow access to .ssh identity timed out, used trscron

– Protocol port blocking – Ssh following Xmas attacks; bbftp, iperf ports, big variation between sites – Wrote analyses to recognize and worked with site contacts

• Ongoing issue, especially with increasing need for security, and since we want to measure inside firewalls close to real applications

• Simple tool built for tracking problems

32

Next steps• Develop/extend management, analysis, reporting, navigating tools – improve robustness,

manageability, workaround ssh anomalies • Get improved forecasters and quantify how they work, provide tools to access • Optimize intervals (using forecasts, and lighter weight measurements) and durations • Evaluate use of compression • Evaluate self rate limiting application (bbcp)• Tie in passive Netflow measurements • Add gridFTP (with Allcock@ANL), UDPmon (RHJ@manchester) & new BW measurers –

netest (Jin@LBNL), INCITE (Reidi@Rice), pathrate, pathload (Dovropolis@Udel) – Make early data available via http to interested & “friendly” researchers – CAIDA for correlation and validation of Pipechar & iperf etc. (sent documentaion) – NWS for forecasting with UCSB (sent documentation)

• Understand correlations & validate various tools, choose optimum set• Make data available by std methods (e.g. MDS, GMA, …) – with Dantong@BNL, Jenny

Schopf@ANL & Tierney@LBNL • Make tools portable, set up other monitoring sites, e.g. PPDG sites

– Working with INFN/Trieste to port– Work with NIMI/GIMI to evaluate deploying dedicated engines – More uniformity, easier management, greater access granularity & authorization

• Still need non dedicated: • Want measurements from real application hosts, closer to real end user • Some apps may not be ported to GIMI OS

– Not currently funded for GIMI engines – Use same analysis, reporting etc.

33

More Information• IEPM/PingER home site:

– www-iepm.slac.stanford.edu/• Bulk throughput site:

– www-iepm.slac.stanford.edu/monitoring/bulk/• SC2001 & high throughput measurements

– www-iepm.slac.stanford.edu/monitoring/bulk/sc2001/• Transfer tools:

– http://dast.nlanr.net/Projects/Iperf/release.html– http://doc.in2p3.fr/bbftp/– www.slac.stanford.edu/~abh/bbcp/– http://hepwww.rl.ac.uk/Adye/talks/010402-ftp/html/sld015.htm

• TCP Tuning:– www.ncne.nlanr.net/training/presentations/tcp-tutorial.ppt– www-didc.lbl.gov/tcp-wan.html

• QBSS measurements– www-iepm.slac.stanford.edu/monitoring/qbss/measure.html

34

Further slides

35

0 400

300

Iperf TCP Mbits/sWorking with Developer. Not using TCP, but should track Pipechar disagrees badly above 100Mbits/s (6 hosts, 20%),50% of hosts have reasonable agreement

Pip

echa

r m

in th

roug

hpt M

bits

/sE.g. iperf vs pipechar

36

Optimizing CPU vs

window vs streams vs throughput

•MHz ~ 0.97 * Mbps•Bigger windows = less cpu for fixed throughput

Iperf % cpu vs throughput vs window SLAC to RIKEN Nov 29 '01

0

5

10

15

20

0 5 10 15 20Throughput Mbits/s

CP

U M

Hz

8K 16KB32KB 64KB128KB 256KB512KB 1024KB

Increasing window

Increasing streams

Hooks at end = saturation

37

Typical QBSS test bed• Set up QBSS testbed

• Configure router interfaces– 3 traffic types:

• QBSS, BE, Priority

– Define policy, e.g.• QBSS > 1%, priority < 30%

– Apply policy to router interface queues

10Mbps

100Mbps

100Mbps

100Mbps

1Gbps

Cisco 7200s

38

Example of effects

•Also tried: 1 stream for all, and priority at 30%, 100 Mbps & 2 Gbps bottlenecks

•2Gbps challenge to saturate: did at SC2001, 3 Linux cpus with 5*1 Gbps NIC cards and 2 Gbps trunk from subnet to floor network, sending to 17 hosts in 5 countries

Kicks in fast (<~ 1 s)

39

Impact on response time (RTT)• Run ping with Iperf loading with various QoS

settings, iperf ~ 93Mbps– No iperf ping avg RTT ~ 300usec (regardless of QoS)– Iperf = QBSS, ping=BE or Priority: RTT~550usec

• 70% greater than unloaded

– Iperf=Ping QoS (exc. Priority) then RTT~5msec• > factor of 10 larger RTT than unloaded

40

File Transfer• Used bbcp (written by Andy Hanushevsky)

– similar methodology to iperf, except ran for file length rather than time, provides incremental throughput reports, supports /dev/zero, adding duration

– looked at /afs/, /tmp/, /dev/null– checked different file sizes

• Behavior with windows & streams similar to iperf

• Thrubbcp ~0.8*Thruiperf

•For modest throughputs (< 50Mbits/s) rates are independent of whether destination is /afs/, /tmp/ or /dev/null. •Cpu utilization ~ 1MHz/Mbit/s is ~ 20% > than for iperf

41

Application rate-limiting• Bbcp has transfer rate limiting

– Could use network information (e.g. from Web100 or independent pinging) to bbcp to reduce/increase its transfer rate, or change number of parallel streams

No rate limiting, 64KB window, 32 streams

15MB/s rate limiting, 64KB window, 32 streams