Internet2 E2E piPEs EndtoEnd Performance Initiative Performance Improvement System

1
Internet2 E2E piPEsEnd-to-End Performance
Initiative Performance Improvement System

• Eric L. Boyd
• Internet2

2
Overview

• E2E piPEs Overview
• What is piPEs?
• Goals
• E2E piPEs Measurement Infrastructure
• Abilene Measurement Domain
• Data Analysis Status
• Preliminary Data Discovery Approach
• AMI Web Service

3
Internet2 E2E piPEs

• Project End-to-End Performance Initiative
  Performance Environment System (E2E piPEs)
• Approach Collaborative project combining the
  best work of many organizations, including
  DANTE/GEANT, Daresbury, EGEE, GGF NMWG,
  NLANR/DAST, UCL, Georgia Tech, etc.

4
Internet2 E2E piPEs Goals

• Enable end-users network operators to
• determine E2E performance capabilities
• locate E2E problems
• contact the right person to get an E2E problem
  resolved.
• Enable remote initiation of partial path
  performance tests
• Make partial path performance data publicly
  available
• Interoperable with other performance measurement
  frameworks

5
Measurement Infrastructure Components
6
Sample piPEs Deployment
7
piPEs Deployment
8
Measurement Software Components
9
Abilene Measurement Domain

• Part of the Abilene Observatory
• http//abilene.internet2.edu/observatory
• Regularly scheduled OWAMP (1-way latency) and
  BWCTL (Iperf wrapper) Tests
• Web pages displaying
• Latest results http//abilene.internet2.edu/ami/bw
  ctl_status.cgi/TCP/now Weathermap
  http//abilene.internet2.edu/ami/bwctl_status_map.
  cgi/TCP/now
• Worst 10 Performing Links http//abilene.internet2
  .edu/ami/bwctl_worst_case.cgi/TCP/now
• Data available via web service
• http//abilene.internet2.edu/ami/webservices.html

10
Overview

• E2E piPEs Overview
• What is piPEs?
• Goals
• E2E piPEs Measurement Infrastructure
• Abilene Measurement Domain
• Data Analysis Status
• Preliminary Data Discovery Approach
• AMI Web Service

11
Data Collection / Correlation

• Collection Today
• Iperf (Throughput)
• OWAMP (1-Way Latency, Loss)
• SNMP Data
• Anonymized Netflow Data
• Per Sender, Per Receiver, Per Node Pair
• IPv4 and IPv6
• Collection in the Future
• NTP (Data)
• Traceroute
• BGP Data
• First Mile Analysis

• Correlation Today
• Worst 10 Throughputs
• Worst 10 Latencies
• Correlation in the Future
• 99th Percentile Throughput over Time
• Throughput/Loss for all E2E paths using a
  specific link
• Commonalities among first mile analyzers
• Sum of Partial Paths vs. Whole Path

12
Data Analysis

• Analysis Today
• Throughput over Time
• Latency over Time
• Loss over Time
• Worrisome Tests? (Any bad apples in Worst Ten?)
• Not the Network (If Worst Ten is good enough)

• Analysis in the Future
• Latency vs. Loss
• How good is the network?
• Do common first mile problems exist?
• Does a link have problems that only manifest in
  the long-haul?
• Is the network delivering the performance
  required by a funded project?

13
Data Discovery / Interoperability

• Discovery in the Future
• Where are the measurement nodes corresponding to
  a specific node?
• Where are the test results for a specific partial
  path?

• Interoperability in the Future
• Can I have a test within or to another
  measurement framework?
• Can I have a measurement result from within or to
  another measurement framework?

14
Overview

• E2E piPEs Overview
• What is piPEs?
• Goals
• E2E piPEs Measurement Infrastructure
• Abilene Measurement Domain
• Data Analysis Status
• Preliminary Data Discovery Approach
• AMI Web Service

15
Preliminary Data Discovery Approach

• Good for first/last mile analysis only
• Does not support multi-homing
• A collection of testing servers are placed at the
  networks edge
• Test requests come from outside the network
• Users can test to the ingress or egress point of
  the network
• Testing service will support both, but would
  prefer ingress point testing

16
Connectivity
17
Ingress Point Benefits

• Closer to client (users desktop)
• Shorter network path, fewer links to analyze
• Reduces test traffic over network core
• Better for finding configuration problems with
  client host/network

18
Egress Point Benefits

• Closer to destination
• Approximates the path an application will use
• Better for finding E2E performance problems

19
Traceroute Tree Map

• All servers run traceroute to all other servers
  in the cloud
• Routes are stored in a tree-based map

20
Sample Traceroute Tree Map
21
Client Test Request

• Client connects to server S1 requesting that a
  test be run
• Server S1 runs traceroute back to client IP
• Server S1 compares client route to core map

22
Sample Map/Path Comparison

• Traceroute Map
• S1
• R1
• R4
• R5
• S6

• Traceroute to Cient
• S1
• R1
• R4
• R5
• Ra
• Rb
• Rc
• Client

23
Ingress Server Selection

• Comparison showed that server S6 is the ingress
  server, redirect client to S6
• Prefer configuration testing
• Allow client to automatically select egress point
  server
• Allow performance testing
• Allow client to manually select any server in the
  cloud

24
Overview

• E2E piPEs Overview
• What is piPEs?
• Goals
• E2E piPEs Measurement Infrastructure
• Abilene Measurement Domain
• Data Analysis Status
• Preliminary Data Discovery Approach
• AMI Web Service

25
AMI Web Service

• AMI measurements published as a web service.
• Implementation of 2004-01-15 NMWG response schema
• Older implementation still exists in different
  namespace
• Only supported for Advisor and Monalisa
• Until end of March.

26
Data Service Implementation

• Data served using perl SOAPLite
• Both SOAP and XMLRPC services
• Same back-end perl module
• OGSI/WSRF server under development
• Using OGSILite
• Also working on .NET/C

27
Database and Clients

• Data pulled from same mySQL database as AMI
  graphs and other information on web pages.
• Multiple clients have been tested
• Perl, python for SOAP
• Perl, Java (Advisor, Monalisa) for XMLRPC
• Php for XMLRPC coming soon
• Others?

28
Bumps on the Road

• Perl is untyped.
• SOAPLite performs autotyping
• Sometimes doesnt do what we want!
• Caused unexpected errors in java
• Can over-ride but not yet done so (other
  priorities)
• XML encoding
• Defaults to RPC
• Can over-ride to force Literal (again, not yet)

29
Uphill Struggles

• The hardest part of the implementation was
  decoding the XML.
• There are some tools but mostly it was a
  time-consuming manual process.
• XMLParser
• Prefer old-style Profile document or even a list
  of objects.

30
Client Side

• New client requires characteristic to be
  specified.
• Partial implementation of request schema for
  requesting data not just measurements
• Future deployment will follow request schema to
  request data and measurements
• Only return what is asked for, previously
  everything available was returned.

31
Supported Characteristics

• Measured by owamp
• path.delay.oneWay
• path.loss.oneway
• Measured by bwctl
• path.bandwidth.achievable.TCP
• path.bandwidth.achievable.UDP
• path.jitter.oneWay

32
Other Inputs

• Measurement Nodes by nickname (ATLA, CHIN, DNVR,
  HSTN, IPLS, KSCY, LOSA, NYCM, SNVA, STTL, WASH).
• Type is IPv4or IPv6 at all nodes. Not case
  sensitive. 4 or 6 will suffice.

33
Other Inputs

• startTime/endTime
• iso8601 is supported
• Dan Gunter - Timezones are a problem waiting to
  happen - very true
• But users are looking for trouble
• Currently no support for epochSeconds
• Too much opportunity for timezone confusion

34
Example Client

• In perl

my request ( networkCharacteristic gt
"path.delay.oneway", subject gt
source gt address gt name gt "ATLA", , ,
destination gt
address gt name gt "LOSA", , ,
type gt "IPv4", ,
startTime gt 2004-02-01T000000-500
", endTime gt 2004-02-01T235959
-500", ) my webService
SOAPLite -gt uri('http//abilene.internet
2.edu/PIPES') -gt proxy('http//usernamepa
ssword_at_abilene.internet2.edu/ami/services/soap_ser
ver.cgi') -gt NetworkMeasurementSet(\reque
st) -gt result
Wrap the request up in an object
Send it to the webservice
35
Example Client Output

• print webService-gtversion
• 2004-01-15
• print webService-gtnetworkMeasurement-gtcharacte
  ristic
• path.bandwidth.achievable.TCP

36
Supported Objects

• Could add Memory, disk details (both can have a
  significant effect on performance).
• Last updated field would be a useful sanity
  check.
• Some system information can be retrieved by snmp,
  sysctl, currently hard coded.
• Methodology has separate source and destination
  objects instead of single endpoint.

37
Non-Standard Objects

• NetworkMeasurementSet gt Help or
  NetworkMeasurementSet gt Message
• e.g. No data matching query or no response
  from db
• Paola suggested using SOAP Failure
• But not really a SOAP failure if there is no data
  matching the query.
• Same applies for other SOAP messages

38
Future Work

• Deployment on CENIC, NCNI, SOX
• Interconnecting measurement domains
• Other Characteristics
• Path.delay.roundtrip - AMP data
• Continued development of Request schema, future
  Response schema
• Development of AAA (certificate based?)

39
More Information

• piPEs/AMI Web Service home page
• http//abilene.internet2.edu/ami/webservices.html
• Monalisa
• http//vinci.cacr.caltech.edu8080
• Advisor
• http//dast.nlanr.net/Projects/Advisor
• Or send questions/comments to
• Warren.Matthews_at_oit.gatech.edu

40
(No Transcript)