The 3rd International HEP DataGrid Workshop KOREN E2E Performance:

Presentation on theme: "The 3rd International HEP DataGrid Workshop KOREN E2E Performance:"— Presentation transcript:

1 2004.8.26 Jaehwa Lee/KOREN jhlee@noc.kr.apan.net
The 3rd International HEP DataGrid Workshop KOREN E2E Performance: Network Engineering Activity Jaehwa Lee/KOREN Presented by Joonbok Lee/KAIST

2 Contents Goal Activity Concluding Remark 2.1 etop Visualization
2.2 Network WeatherMap 2.3 Traffic Engineering System Concluding Remark

3 1. Goal End-to-End Performance from the Network’s point of view.
End-to-End Performance Monitoring Traffic Monitoring and Traffic Engineering

4 Goal Activity Concluding Remark 2.1 etop Visualization
2.2 Network WeatherMap 2.3 Traffic Engineering System Concluding Remark

5 etop: Edge-to-edge performance system
System Architecture

6 Metrics and Tools ICMP TCP UDP Tool ping Iperf owamp * Metrics Packet Loss (%) Round Trip Time (ms) Bandwidth(kbps) Packet loss(%) Jitter(ms) One-way latency *owamp: implementation of the OWAMP protocol that is currently going through the standardization process within IETF.

7 Measurement and Data collecting
6 machines were deployed. Each machine send probe packets by using ping, iperf and owamp They report result to main DB server in Daejeon. CRON Iperf TCP Iperf UDP CRON CRON Iperf TCP Iperf UDP Iperf TCP Iperf UDP CRON Iperf TCP Iperf UDP CRON Iperf TCP Iperf UDP

8 Data Presentation (TCP Throughput)

9 Data Presentation (TCP Throughput)

10 Data Presentation (UDP Throughput)

11 Data Presentation (UDP Throughput)

12 Data Presentation (Jitter)
Jitter(ms)

13 Data Presentation (RTT)

14 Goal Activity Concluding Remark 2.1 etop Visualization
2.2 Network WeatherMap 2.3 Traffic Engineering System Concluding Remark

15 2.2 Network WeatherMap To show current link utilization. Netmap v1.3
based on existing images (only support PNG format) It uses SNMP to collect data from other devices Multiple Network WeatherMaps Multiple MRTG Style Graph {Host, CPU, Memory} Information

16 KOREN Netmap Deployment

17 UI – Devices List

18 UI – Traffic data

19 Goal Activity Concluding Remark 2.1 etop Visualization
2.2 Network WeatherMap 2.3 Traffic Engineering System Concluding Remark

20 2.3 Traffic Engineering System
Background users and traffic have grown on the Internet both users and network managers need to have an understanding of the network performance Traffic Engineering System A software system to manage and measure various values which indicate network performance To optimize the utilization of existing network resources for load balance and failure recovery To manage the network of KOREN

21 Data Collection Module
TES Architecture TES Server Web Server User Manager Server Process Data Collection Module Analysis Module DB SNMP Diagnosis Module 사용자는 웹을 통해 망의 상태를 모니터링 할 수 있고 관리자는 일반 사용자의 경우처럼 웹을 통해 정보를 얻을 수 있으며, 서버 프로세스에 의해 진단 된 결과를 관리자에게 경고 창이나, 메일로 이러한 사실을 통보 한다. - Traffic Engineering System Architecture -

22 Internal TES Architecture
서버 프로세스에는 진단 모듈, 분석 모듈, 데이터 수집 모듈로 구성된다. - Traffic Engineering System Architecture -

23 Data Collection Module
Collect traffic information about each Node of the KOREN Using active measurement data, SNMP MIB, flow information, and, … ifSpeed ifInOctets Data Collection Module ifOutOctets ifInErrors

24 Analysis Module Calculate the performance measurement metric with the collected data Insert parsed measurement metric into database Reporting to diagnosis module or network manager Throughput Utilization Link error Available bandwidth DB Data Collection Module Diagnosis Module Analysis Module 분석모듈의 두 가지 기능 데이터 수집 모듈로부터 4가지 metric에 대한 정보를 얻어 DB에 값을 저장 Throughput, Utilization, Link error에 대해 각각의 설정된 임계치와 비교하여 높으면 -> 원인을 찾기 위해 진단 모듈로… 낮으면 -> 네트워크 관리자에게 단순 통보… Threshold < Load Threshold > Load Network Manager

25 Diagnosis Module For the detection of abnormal situations via the observation of the network The various conditions C1: Outgoing overload : concentrated traffics from a node C2: Incoming overload : concentrated traffics towards a node C3: multicast/broadcast : The traffic which is not notified to the network manager C4: Link error : abnormal physical error Reporting a diagnosis result to network manager 원인 분석을 위해 4가지 조건 사례로 규정 비정상적인 망(설정된 임계치보다 높은 망)에 연결된 두 노드 중에 문제가 되는 노드를 찾고자 SNMP MIB정보를 이용하여 C1과 C2 조사를 통해 문제가 되는 노드를 찾는다. 또한, 문제가 되는 노드에서 MIB정보를 이용하여 C3와 같은 트래픽양을 알아 볼 수 있으며, C4는 에러율을 통해 알아 볼 수 있다. 이러한 진단 결과를 네트워크 관리자에게 보고한다.

26 3. Concluding Remarks KESG (KOREN Engineering Study Group) are working for Those systems will be used by end users as well as network managers. Contact Information: Etop and Network weathermap (prototype)