09.04.2003 Decus Bonn 7. Bonn 7.- -11. April 2003 11. April 2003 Decus Einführung in Quality of Service Eva Heinold CCCSC München eva.heinold@hp.com Eva Heinold , DECUS Bonn 2003 09.04.2003 Eva Heinold, HP Supportcenter München 1
09.04.2003 Agenda . . What is QoS What is QoS? ? Factors Factors, , influencing Quality influencing Quality of of each Application each Application - - Packetloss, Packetloss , Delay Delay, , Delay Delay- -variations variations QoS Demands for QoS Demands for - - Voice Voice, , Video Video, , Data Data QoS Tools QoS Tools - - Classification Classification - - Marking Marking - - Trust Boundaries Trust Boundaries - Scheduling - Scheduling - - Provisioning Provisioning - - Congestion avoidance, Policing, Schaping Schaping Congestion avoidance, Policing, - - Management Management QoS QoS Design Design Considerations Considerations What is Quality of Service? “ The Pragmatic Answer: QoS is Advanced Resource Management The Technical Answer: The Resources!! Set of techniques to manage: • Delay • Delay Variation (Jitter) • Packet Loss • Bandwidth ” Eva Heinold , DECUS Bonn 2003 09.04.2003 Eva Heinold, HP Supportcenter München 2
09.04.2003 But…but… Bandwidth…... “ “Money and sex, storage and bandwidth: only too much is ever enough” • Arno Penzias - Former Head of Bell Labs, and Nobel prizewinner ” Eva Heinold , DECUS Bonn 2003 09.04.2003 QoS is Fundamental to Voice/Video (and Data) • Standards-based Applications • Personalized • Scalable Voice and Video Telephony • Flexible • QoS-Enabled Network Infrastructure • Resilient • Security Eva Heinold , DECUS Bonn 2003 09.04.2003 Eva Heinold, HP Supportcenter München 3
09.04.2003 Factors that Affect the Quality of Any Application IP Telephony SAP 4.6 HP UDC Microsoft EDC Oracle 11i Video Conferencing Delay- - Delay Packet Packet Delay Delay Variation Variation Loss Loss (Latency) (Latency) (Jitter) (Jitter) Eva Heinold , DECUS Bonn 2003 09.04.2003 . . What is What is QoS QoS? ? Factors Factors, , influencing Quality influencing Quality of of each Application each Application - Packetloss, , Delay Delay, , Delay Delay- -variations variations - Packetloss QoS Demands for QoS Demands for - - Voice, Voice , Video Video, , Data Data QoS QoS Tools Tools - - Classification Classification - - Marking Marking - - Trust Boundaries Trust Boundaries - Scheduling - Scheduling - - Provisioning Provisioning - - Congestion avoidance, Policing, Congestion avoidance, Policing, Schaping Schaping - - Management Management - - QoS Design QoS Design Considerations Considerations Eva Heinold, HP Supportcenter München 4
09.04.2003 QoS Requirements for Voice Voice Voice • Latency ≤ 150 ms One-way • Jitter ≤ 30 ms requirements • Loss ≤ 1% • 17-106 kbps guaranteed Smooth • priority bandwidth per call Easy to Provision • • 150 bps (+ layer 2 overhead) Drop Sensitive • guaranteed bandwidth for Delay Sensitive • Voice-Control traffic per call UDP Priority • Eva Heinold , DECUS Bonn 2003 09.04.2003 VoIP Delay Budget Cumulative Transmission Path Delay Satellite Quality Satellite Quality High Quality High Quality Fax Relay, Broadcast Fax Relay, Broadcast 0 0 100 100 200 200 300 300 400 400 500 500 600 600 700 700 800 800 Time (msec) Time (msec) 0 Delay Target (max) Delay Target (max) ITU’s G.114 Recommendation = 0–150 msec 1-Way Delay Eva Heinold , DECUS Bonn 2003 09.04.2003 Eva Heinold, HP Supportcenter München 5
09.04.2003 QoS Requirements for Video-Conferencing Video Video • Latency ≤ 150 ms One-way • Jitter ≤ 30 ms requirements • Loss ≤ 1% • Minimum bandwidth guarantee required is: Bursty Bursty • • Video-Stream + 20% Greedy Greedy • • – e.g. for a 384 kbps stream a Drop Sensitive reserved 460 kbps of priority • bandwidth is recommended Delay Sensitive • UDP Priority • Eva Heinold , DECUS Bonn 2003 09.04.2003 QoS Requirements for Data Data Data • Different applications have different traffic characteristics • Different versions of the same application can have different traffic characteristics • Classify Data into relative-priority model with a maximum of four classes Gold: Mission-Critical Apps (ERP Apps, Transactions) Silver: Guaranteed-Bandwidth (Intranet, Messaging) Bronze: Best-Effort (Email, Internet) Less-Than-Best-Effort: Scavenger (FTP, Backups, Napster/Kazaa) Eva Heinold , DECUS Bonn 2003 09.04.2003 Eva Heinold, HP Supportcenter München 6
09.04.2003 Different Classification Different Classification Equant does classify as follows: Real Time Class -- optimized for toll quality Voice over IP and time- sensitive applications. Interactive Class -- designed to give quick response for business critical applications. Standard Business Class -- suitable for day-to-day business applications, client server traffic and corporate web traffic. General Class -- ideal for email, Internet http traffic and Notes replication. For further information..http://www.equant.com Eva Heinold , DECUS Bonn 2003 09.04.2003 Provisioning for Data: Application Differences Oracle SAP R/3 65-127 0-64 Bytes 1024-1518 Bytes Bytes 128-252 Bytes 512-1023 0-64 Bytes Bytes 253-511 Bytes 253-511 Bytes 512-1023 Bytes 1024-1518 Bytes 128-252 Bytes 65-127 Bytes Almost 75% of Oracle packets are Over 50% of SAP R/3 packets are greater than 1024 Bytes less than 128 Bytes Eva Heinold , DECUS Bonn 2003 09.04.2003 Eva Heinold, HP Supportcenter München 7
09.04.2003 Provisioning for Data: Version Differences 500,000 400,000 300,000 200,000 100,000 0 SAP GUI, SAP GUI, SAP GUI, SAP GUI Release Release Release (HTML), 3.0F 4.6C, w ith 4.6C, no Release Cache Cache 4.6C The same transaction takes over 35 times more traffic from one version of an application to another Eva Heinold , DECUS Bonn 2003 09.04.2003 Provisioning for Data: Sample Exercise (Provisioning for Email) • Average employee mailspool is 10 MB/day • 8 hour workday – ! 2.8 kbps per employee (for each second of workday) • Remote-branch of 50 employees – ! 140 kbps daily average (for each second of workday) • Email is cyclical; 8-10:30 am use is significantly higher than rest of day (assume double) – ! 280 kbps average (from 8-10:30am only) Eva Heinold , DECUS Bonn 2003 09.04.2003 Eva Heinold, HP Supportcenter München 8
09.04.2003 TCP Flow Statistics • >90% of sessions have ten packets each way or less • Transaction mode (mail, small web page) • >10% of all TCP sessions produce 80% of the traffic • with high rate bursts – It is these that we worry about managing Eva Heinold , DECUS Bonn 2003 09.04.2003 Behavior of a High-Throughput / Bulk-Transfer TCP Session 40 35 Congestion Avoidance Phase 30 Linear Growth 25 20 15 10 Slow Start 5 Exponential Growth 0 50 10 20 30 40 0 • TCP will keep at most a certain amount of traffic in flight – We say it is “elastic”—rate is proportional to latency • Voice will send only and exactly as fast as the coding algorithm permits (Also Video to an extent) – We say it is “inelastic” Eva Heinold , DECUS Bonn 2003 09.04.2003 Eva Heinold, HP Supportcenter München 9
09.04.2003 Application QoS Requirements Voice FTP ERP and Voice FTP ERP and Mission-Critical Mission-Critical Bandwidth Low to Moderate Low Bandwidth Low to Moderate Low Moderate Moderate to High to High Random Drop Sensitive Low High Moderate Random Drop Sensitive Low High Moderate To High To High Delay Sensitive Delay Sensitive High High Low Low Low to Low to Moderate Moderate Jitter Sensitive Jitter Sensitive High High Low Low Moderate Moderate Eva Heinold , DECUS Bonn 2003 09.04.2003 . . What is What is QoS QoS? ? Factors Factors, , influencing Quality influencing Quality of of each Application each Application - Packetloss, , Delay Delay, , Delay Delay- -variations variations - Packetloss QoS Demands for QoS Demands for - - Voice Voice, , Video Video, , Data Data QoS QoS Tools Tools - - Classification Classification - - Marking Marking - - Trust Boundaries Trust Boundaries - - Scheduling Scheduling - - Provisioning Provisioning - - Congestion avoidance, Policing, Schaping Schaping Congestion avoidance, Policing, - - Management Management QoS QoS Design Design Considerations Considerations Eva Heinold, HP Supportcenter München 10
Recommend
More recommend
