Agilent Technologies, Inc. · NYSE:A
Agilent Technologies — Technical Presentation (March 2002)
24 slides from Agilent Technologies's technical presentation, Mar 26, 2002
Slide 1: On Active Measurements in QoS-Enabled IP Networks
On Active Measurements in QoS-Enabled IP Networks Graham Pollock, Agilent Labs 3/26/2002 Passive & Active Measurement Slide 1 Workshop 2002
Slide 2: Overview
Overview • Qos Measurement in IP networks • The Configuration Mismatch • Strategies for Matching Configuration • Prototype Experience • Conclusions 3/26/2002 Passive & Active Measurement Slide 2 Workshop 2002
Slide 3: QoS Measurement in IP Networks
Qos Measurement in IP Networks • Network service providers (NSPs) use service level agreements (SLAs) to specify the service levels they commit to provide to their customers • SLAs typically specify service levels using measurable quality of service (QoS) parameters – Packet Loss – Delay – Jitter •
Slide 4: QoS Measurement in IP Networks
Qos Measurement in IP Networks • Network monitoring applications monitor service performance against SLA requirements, and generate alerts when actual performance drops below guaranteed levels • These alerts often trigger troubleshooting activities on the part of network operators • These applicatio
Slide 5: The Configuration Mismatch
The Configuration Mismatch • Network and service management applications often rely on active measurement techniques • In QoS networks traffic is handled according to traffic handling policies configured into network elements • Unless the application can determine the network’s QoS configuration (e.
Slide 6: The Configuration Mismatch
The Configuration Mismatch • QoS mechanisms establish classes of traffic that can be handled differently by network elements • In each case the basic concept is the same: network elements classify and handle packets differently using information in the packet header • In the absence of QoS configura
Slide 7: Configuration Mismatch – An Illustration
Configuration Mismatch – An Illustration CQ Q1 R Q2 M N1 N2 QoS Network Configuration • Router R connects the network. • Nodes Q1 and Q 2 are defined to have a QoS relationship • QoS is provided using the Type of Service (TOS) bits for priority handling through R. • The network monitoring applicatio
Slide 8: Configuration Mismatch – An Illustration
Configuration Mismatch – An Illustration • The iperf tool from NLANR was used to generate traffic and to report its throughput between the various nodes • The TOS bits on packets sent from Q1 to Q2 were set to non-zero values. Traffic between N1 and N 2 did not set the TOS bits • The network monitor
Slide 9: Configuration Mismatch – An Illustration
Configuration Mismatch – An Illustration 3/26/2002 Passive & Active Measurement Slide 9 Workshop 2002
Slide 10: Configuration Mismatch – An Illustration
Configuration Mismatch – An Illustration The Figure above shows iperf output for CQ, generated at one- minute intervals, showing the amount of data transferred and the bandwidth. Client data are shown on the left, and the corresponding server data on the right. It is clear that CQ is well within the
Slide 11: Configuration Mismatch – An Illustration
Configuration Mismatch – An Illustration • The obvious flaw in this illustration is that the monitoring application is not configured to measure the correct path: iperf packets on CQ effectively traverse a separate path from monitoring packets destined to Q2 3/26/2002 Passive & Active Measurement Sl
Slide 12: Other QoS Configuration Issues
Other QoS Configuration Issues • Packet Classification – In order for active measurement traffic to accurately measure a particular class of traffic, measurement packets must be classified and handled the same as the traffic they intend to measure. – TOS is just one method - Classification can be ba
Slide 13: Other QoS Configuration Issues
Other QoS Configuration Issues • Dynamic QoS Configuration – Dynamic SLAs allow on-demand provisioning to accommodate variable QoS requirements. – Dynamic SLAs require a signaling mechanism (such as RSVP) in order to pass along changing QoS configuration data. – The impact of such changes on active
Slide 14: Strategies for Matching QoS Configuration
Strategies for Matching QoS Configuration • Several strategies for matching active measurement configuration with QoS configuration were considered. • Match Measurements at the Source • Match Measurements at the Access Router 3/26/2002 Passive & Active Measurement Slide 14 Workshop 2002
Slide 15: Matching QoS Configuration
Matching QoS Configuration • Match Measurements at the Source – The basic idea behind this approach is to configure measurement packets to have the same characteristics as the packets of SLA traffic that is being measured. – E.g. if TOS bits are used to differentiate traffic, measurement packets wou
Slide 16: Matching QoS Configuration
Matching QoS Configuration • Network Stack – In this variation, packets headers from the monitoring application or from applications being utilized by the monitoring applications (e.g., ftp) are modified by “shim” software inserted into the network stack – Modifications to existing measurement softw
Slide 17: Matching QoS Configuration
Matching QoS Configuration • Application Level – In this variation, the monitoring application creates well-formed measurement packets that contain the same QoS configuration as the measured traffic. – A significant issue with this approach is that monitoring applications often utilize existing appl
Slide 18: Matching QoS Configuration
Matching QoS Configuration • Match Measurements at the Access Router – The basic idea behind this approach is to configure the access router to recognize packets from the monitoring application and then classify them so that they are handled the same as the SLA traffic being measured. – Two variatio
Slide 19: Prototype Experience
Prototype Experience • We focused our prototype efforts on the router configuration approaches. • We began with SNMP access router configuration. • The document entitled “Management Information Base For the Differentiated Services Architecture” is an attempt by the Internet Engineering Task Force (I
Slide 20: Prototype Experience
Prototype Experience • The Differentiated Services MIB contains a number of objects for modeling how traffic should be handled by a device • These objects include Classifiers, Meters, Actions, Algorithmic Droppers, Queues, and Schedulers. • These various object types give great flexibility over how
Slide 21: Prototype Experience
Prototype Experience • Unfortunately, the set function was not implemented in several of the tables in the MIB implementation which we had chosen, thus we could not complete our prototype using this approach. • As a fallback, we implemented a script-based approach that performed the necessary diffse
Slide 22: Prototype Experience
Prototype Experience • Conceptually these scripts carry out the same functionality as was envisaged with the SNMP approach, albeit in a more proprietary and device-dependant manner. • Once the scripts that query and configure the router to handle measurement traffic had been executed, the bulk trans
Slide 23: Conclusions
Conclusions • Network monitoring applications are often used by service providers to monitor compliance with service level agreements made with customers and to trigger troubleshooting activities when service drops below guaranteed level. • Many of these applications rely on active measurement techn
Slide 24: Conclusions
Conclusions • The lack of standardization of QoS methods makes it difficult to come up with a general solution for matching active measurements in monitoring applications with customer SLAs. • In the area of Differentiated Services, support for the proposed IETF MIB should make QoS-aware monitoring
Agilent Technologies — Technical Presentation (March 2002) on 6ix