Prepared for Prof. Javed I. Khan
Department of Computer Science, Kent State University
Date: November 2001
Recent Advances
Internet Measurement Insfrustructure
Public Insfrustructure
Private InsfrustructureFuture Challenges
Backbone engineering and planning are among the most pressing needs for reliable forms of traffic data and analyses. Key elements of these analyses are aggregate traffic data at the IP layer, including port and protocol statistics (packets and bytes per port and per protocol) and traffic matrix statistics (how many packets and bytes were sent from network A to network B).
One participant of ISMA'97 from ANS described what forms of measurement data that he and other backbone engineers need most, suggesting that critical data not readily available include:
General Classes of Measurement Approaches
Aggregation-based approaches are deterministic functions of the observed data [3] . They usually compute the sum or the maximum of some metric over the dataset (e.g. sum of packets traversing over a link during an interval, or the maxiend-to-end round trip delay for a set of packets.)Whereas sampling based approaches extract a random subset of all the
possible observations. This sample subset is supposed to be representative
of all of the whole.
Classification Table
[4] shows differents projects which are now underway
targetting different groups.
|
|||
|
|
|
|
CoralReef | Traffic Engr, Internet Researchers | Passive | Workload |
IEPM | Network For HEP Community | Active | Performance |
I2 (Abilence) | High-availability Baclbone For Academic Researchers | Passive & Active | Workload & Performance |
Mantra | Internet Researchers | MBGP Routing | Multicast Performance |
MAWI (WIDE) | Internet Researchers | Passive | Workload, Performance |
NIMI | Global Internet Comminity, Internet Researchers | Active | Workload, Performance |
NLANR (MOAT) AMP | Internet Researchers | Active | Performance |
NLANR (MOAT) PMA | Internet Researchers | Passive | Performance |
NI ACI NWS | PACI High_Performance Application Users and Developers | Active & Passive | Workload, Performance |
PPNCG | UK Particle Physics Community | Active | Performance |
RIPE-RIS | European Internet Comminity | Passive | Performance |
skitter | Global Internet Comminity, Internet Researchers | Acitve | Topology, Routing |
Surveyor | US Higher Education Community | Acitve | Topology, Performance |
TRIUMP | Canadian Particle Physics Community | Acitve | Topology, Performance |
U-Oregon Route Views | Traffic Engr, Internet Researchers | Passive | Topology, Routing |
WAND | Internet Researchers | Passive | Workload, Performance |
SPAND (Shared Passive Network PerformanceDiscovery) is a system that facilitates the development of adaptive network applications. In each domain, applications make passive application-specific measurements of the network and store them in a local centralized repository of network performance information. Other applications may retrieve this information from the repository and use the shared experiences of all hosts in a domain to predict future performance. In this way, applications can make informed decisions about adaptation choices as they communicate with distant hosts. All messages between the components of SPAND system use a format similar to Active Messages [20] . A SPAND message contains a handler string, an active flag, a data length, and a message-specific payload.
[21] looked at both of the problems of determining the setting of the retransmission timer (RTO) for a reliable protocol, and estimating the bandwidth available to a connection in the context of TCP, using a large TCP measurement set [22] for trace-driven simulations. For RTO estimation, they evaluated a number of different algorithms, finding that the performance of the estimators is dominated by their minimum values, and to a lesser extent, the timer granularity, while being virtually unaffected by how often round-trip time measurements are made or the settings of the parameters in the exponentially-weighted moving average estimators commonly used. For bandwidth estimation, they explored techniques previously sketched in the literature [23, 24] and found that in practice they perform less well than anticipated. [21] developed a receiver-side algorithm that performed significantly better.
Knowledge of the up-to-date bandwidth utilizations and path latencies is critical for numerous important network management tasks, including application and user profiling, proactive and reactive resource management and traffic engineering, as well as providing and verifying QoS guarantees for end-user applications. Indeed, these observations have led to a recent flurry of both research and industrial activity in the area of developing novel tools and infrastructures for measuring network bandwidth and latency parameters [2] . Examples include SNMP and RMON measurement probes [1] , Cisco’s NetFlow tools [25] , the IDMaps [26] , [27] and Network Distance Maps [28] efforts for measuring endto-end network latencies, the pathchar tool for estimating Internet link characteristics [29] , [30] , and packet-pair algorithms for measuring link bandwidth [31] , [32] . A crucial requirement for such monitoring tools is that they be deployed in an intelligent manner in order to avoid placing undue strain on the shared resources of the production network
Topology-d is a service that estimates the state of networked resources by periodically computing the end-to-end latency and available bandwidth. Using its delay and bandwidth estimates, topology-d computes a fault tolerant, minimum-cost spanning tree connecting participating site.
TReno meaures the throughput of a given link independent of the particular TCP implementation on the end host. bprobr and cprobe probe the network (bandwidth) by sending several pairs (bprobr) or a short train of packets (cprobr) .
Several projects with measurement infrastructures for monitoring Internet
Traffic are currently in place. These utilize either public or private
infrastructure. CAIDA provides short
summaries of those current measurement projects offering public reports.
Public Measurement Infrastructure
CoralReef is a comprehensive software suite providing a set of drivers,
librarie s, utilities, and analysis software for passive network measurement
of workload characteristics. These reports characterize workload on a high-speed
link between UCSD and the commodity Internet.
Internet2 (I2) Abilene is an advanced backbone network that connects
regional network aggregation points, called gigaPoPs, to support the work
of Internet2 universities as they develop advanced Internet applications.
The Abilene Project complements other high-performance research networks.
Abilene enables the testing of advanced network capabilities prior to their
introduction into the application development network. These services are
expected to include Quality of Service (QoS) standards, multicasting and
advanced security and authentication protocols.
SLAC/DOE/ESnet, High Energy and Nuclear Physics uses pingER tools
on 31 monitoring sites to monitor network performance for over 3000 links
in 72 countries. Monitoring includes many major national networks (including
ESNet, vBNS, Internet2-Abilene, CALREN2, NTON, and MREN) as well as networks
in South America, Canada, Europe, the former Soviet Union, New Zealand,
and Africa. Many sites are also part of SLAC's related BaBar
Wide-Area Network Monitoring effort.
The Measurement and Analysis of Wide-area Internet (MAWI) Working
Group studies performance of networks and networking protocols in Japanese
wide-area networks. Sponsored by the Widely Integrated Distributed Environment
(WIDE) project, MAWI is a joint effort of Japanese network research and
academic institutions with corporate sponsorship.
Monitor and Analysis of Traffic in Multicast Routers (Mantra) monitors
various aspects of global Internet multicast behavior at the router level.
Visualization snapshots and accompanying tables are updated every 15-30
minutes.
NIMI is a project, begun by the National Science Foundation and
currently funded by DARPA, to measure the global Internet. Based on Vern
Paxson's Network Probe Daemon, NIMI was designed to be scalable and dynamic.
NIMI scalability comes from its ability to delegate NIMI probes to administration
managers for configuration and information and measurement coordination.
NIMI is dynamic in that the measurement tools employed are treated as third
party packages that can be added or removed as needed. For example, the
MINC (Multicast
Inference of Network Characteristics) measurement methodology for determinng
performance characteristics of the interior of a network from edge measurements
has been tested and validated using the NIMI infrastructure.
NLANR's Measurement and Operations Analysis Team (MOAT) is creating a Network Analysis Infrastructure (NAI) to derive a better understanding of system service models and metrics of the Internet. This includes passive measurements based on analysis of packet header traces (link to PMA above); active measurements (link to AMP above); SNMP information from participating servers; and Internet routing related information based on BGP data.
Distributed system to periodically monitor and dynamically forecast
performance available from various network and computational resources
over a given time interval. Service operates a distributed set of performance
sensors (network monitors, CPU monitors, etc.) from which it gathers readings
of instantaneous conditions. It then uses numerical models to generate
forecasts of what the conditions will be for a given time frame. While
the forecasting methods are general, the focus is on the ability to predict
the TCP/IP end-to-end throughput and latency attainable from applications
using systems located at different sites. Such forecasts are needed both
to support wide-area scheduling of large scale computation, and by the
metacomputing software infrastructure to develop quality-of-service guarantees.
The PPNCG (Particle Physics Network Coordinating Group) runs network
monitoring processes on machines at several sites throughout Europe. Its
goal is to gather end-to-end performance information for links of specific
interest to particle physics researchers, and use the information to highlight
problems and help the PPNCG to make recommendations to the appropriate
bodies to optimise the networking available to the UK particle physics
community. This project uses Traceping
Route Monitoring Statistics .
RIPE (Reseaux IP Europeens) is a collaborative organisation open
to organisations and individuals,operating wide area IP networks in Europe
and beyond. The objective of RIPE is to ensure the administrative and technical
coordination necessary to enable operation of a pan-European IP network.
RIPE does not operate a network of its own. Currently, more than 1000 organisations
participate in the work. The result of the RIPE coordination effort is
that an individual end-user is presented with a uniform IP service on his
or her desktop irrespective of the particular network his or her workstation
is attached to.
Surveyor is a measurement infrastructure that is being currently
deployed at participating sites around the world. Based on standards work
being done in the IETF's IPPM WG, Surveyor measures the performance of
the Internet paths among participating organizations. The project is also
developing methodologies and tools to analyze the performance data.
Canadian national research facility uses perl scripts to trace paths
toward nodes of interest to TRIUMF
. Packet los summaries and graphs are generated daily from pins made at
10 minute intervals. Traceroute data is gathered four times daily. Network
vizualization maps are generated from the traceroute data.
A collaborative endeavor to obtain real-time information about the
global routing system from the perspectives of several different backbones
and locations around the Internet. The Route Views router, route-views.oregon-ix.net
, uses multi-hop BGP peering sessions with backbones at interesting locations
(note that location should not matter if the provider is announcing consistent
routes corresponding to its policy). Route Views uses AS65534 in its peering
sessions, and routes received from neighbors are never passed on nor used
to forward traffic. Finally, route-views.oregon-ix.net itself does not
announce any prefixes.
The WAND project aims
to build models of internet traffic for statistical analysis and for the
construction of simulation models. The project builds its own measurement
hardware and collects and archives significant network traces. These are
used internally and are also made available to the Internet research community.
Traces are accurately timestamped and synchronized to GPS. Many traces
are 24 hours long, some are up to a week long, and there are plans to provide
even longer traces in the future. The WAND project is based at the University
of Waikato in New Zealand with strong collaboration from the University
of Auckland .
Measures "traffic index", response time, and packet loss by pinging
many routers along "major paths" on the Internet. (traffic index
- a score from 0 to 100 where 0 is "slow" and 100 is "fast", determined
by comparing the current response to a ping echo request with all previous
responses from the same router for past 7 days.)
Assessment of average response time for accessing and downloading
home pages of 40 Web Sites deemed most indicative to business users, as
measured Mo nday through Friday every 15 minutes between 6 am and noon
Pacific time by Keyno te software measurement agents located in metropolitan
areas of the United State s. Keynote's methodology
The MIDS Internet Average is a high-level summary of Internet performance
measured from hosts all around the world. It provides one baseline against
which more specialized Internet performance data might be compared, serving
a similar role as the Dow Jones Industrial Average does in the financial
world.
The MIDS IWR presents ongoing animated scans of macroscopic conditions
across the Internet. IWR displays geographical maps that show ping-based
RTT latency from MIDS offices in Austin, Texas to thousands of Internet
domains worldwide. Data is currently updated every four hours, six times
a day, seven days a week. Java applets support nonstop viewing and single-stepping
frame by frame. Single GIF images are also available for each of the most
current maps.
MIDS monitors thousands of sites worldwide every 15 minutes to map
the data flow of the Internet. Statistical analyses of this data to determine
network performance form the basis for their Matrix Internet Quality (MIQ)
products. Only a small fraction of the information capable of being provided
by MIQ is used in this public ratings page.
As we enter the new decade, organizations engaged in analyzing macroscopic, infrastructure-wide traffic behavior [5] suggests that we must focus on
Unlike many other fields of engineering, Internet data analysis is no longer justifiable as an isolated activity. The ecosystem under study has grown too large and is under the auspices of too many independent, uncoordinated entities. Nonetheless, as the system continues to evolve rapidly, the depth and breadth of our understanding of it should follow in close pursuit.
IP
Providers Metrics (IPPM) - subgroup of the IETF's Bench marking Working
Group
(BMWG
IP Netwrk Management and Performance Department, AT&T Labs - Research
The Cooperative Association for Internet Data Analysis ( CAIDA )