1. Measurement team Hans Ludwing Reyes Chávez Network Operation Center
Network Operation Center
May 2006
Curitiba, Brasil

2. Introduction

3. CLARA Network

4. Levels of the CLARA Network
Pop
Universidad
CLARA
LA-NREN
Server
LA-Universidad
BACKBONE

5. Architecture CLARA Backbone Commodity Internet Connections PoP Two PoP
Administration LAN
PoP
Three
PoP
One
LA-NREN
Commodity
Internet
Connections
University A
University B

6. INTERNATIONAL CONNECTIVITY
OTRAS
NRENS
CUDI
GEANT
UNAM
CLARA
USA
ABILENE- NLR
GLO
Brasil- RNP
OTRAS
NRENS
OTRAS
NRENS-LA

7. NMS

8. NMS - Network Management Station
To get the alarms and event from the CLARA Network
Used for the CLARA NOC and NEG staff
Used to make the reports asked for the directors of CLARA
HP-Open View - Network Node Manager

9. Backbone Server
On each Pop of CLARA we have two high performance server
The server are used to monitoring the performance of the backbone CLARA network
Pop to Pot
Pop to NREN/ITN
The server are using open source tools

10. CLARA
RNP
NOC-CLARA
Chile
CUDI
NMS

11. NMS
NOC-CLARA
CUDI
RNP
CLARA
Chile

12. CLARA LAYOUT
Backbone Router
WS1
OOB
CLARA
PSTN
WS2
NMS

13. Infrastructure

14. Monitoring Infrastructure
Since the begin in every Pop CLARA have two servers dedicated to have and support tools to measurement
NMS - Network Management station
Dedicated to get alarms
Traps
Reports

15. Infrastructure - Cont.
With backbone servers we can have an active/passive monitoring
With the NMS we can get alarms, traps and the reports for the CLARA directors
The software for the NMS show a topology maps view of the network what make more easy and faster get the idea of how is working the network in every moment

16. CLARA Measurement Team

17. Main Objective
Development and deployment systems to monitoring and measurement the performance that give us the information to know the real status of the network, giving the possibility to detect and correct situations that could affect the network

18. Specific Objectives Determinate the parameter to measure
Deploy and install the tools for the monitoring
Make reports of the status of the network

19. How we are Members from Clara Technical Forum NOC of the LA-NRENs
CLARA NOC / NEG
People outside CLARA
*Formed on the second CLARA- TEC meeting in Veracruz on april 2005

20. Initial Works Install and test of the more familiar tools:
Iperf/Netperf
Pchar
PathLoad
Cacti
Netflow
Logs
AMP
OWAMP

21. Status of the Team
The applications are ready to be used for the NOC, NEG and LA-NRENs
Making tests for the EELA project
Almost of the test has been done between a LA-NRENs and CLARA, last month to Spain, soon to CERN
Many application of real-time are asking us how we can monitoring and guaranty the availability of the service.
Streamings of 10Mbps or more using Multicast
Access-grids
Virtual reality and remote instrumentation

22. Initial Works - cont. Install and test of the more familiar tools:
Beacon Multicast
Beacon H323
Route proxy
Syslog
Naggios …

23. Sample 1
Using Iperf to measure the TCP transference rate from Sao Paulo to Panama.
First using the default parameters for TCP, just one TCP session
After making more bigger the TCP window

24. Sample 1 - Cont. eriko@server2-panama:~$ iperf -s -w 200k -i 10
Server listening on TCP port 5001
TCP window size: KByte (WARNING: requested KByte)
[ 4] local port 5001 connected with port 41138
[ 4] sec MBytes Mbits/sec
[ 4] sec MBytes Mbits/sec
[ 4] sec MBytes Mbits/sec
[ 4] sec MBytes Mbits/sec
[ 4] sec MBytes Mbits/sec
[ 4] sec MBytes Mbits/sec
[ 4] sec MBytes Mbits/sec
[ 4] sec MBytes Mbits/sec ….
[ 4] sec MBytes Mbits/sec
[ 4] sec MBytes Mbits/sec
[ 4] sec MBytes Mbits/sec
[ 4] sec MBytes Mbits/sec
[ 4] sec MBytes Mbits/sec

25. Sample 1 - Cont. badwidth-delay product:
traceroute
traceroute to ( ), 30 hops max, 38 byte packets
1 lan-saopaulo ( ) ms ms ms
2 saopaulo-tijuana.core.redclara.net ( ) ms ms ms
3 tijuana-panama.core.redclara.net ( ) ms ms ms
4 server2-panama ( ) ms ms ms
badwidth-delay product:
155 Mbps x 230 ms = kbits = MB

26. Sample 1 - Cont. TCP parameters
cat /proc/sys/net/core/wmem_max
131071
cat /proc/sys/net/core/rmem_max
cat /proc/sys/net/ipv4/tcp_rmem
cat /proc/sys/net/ipv4/tcp_wmem
Modify (Linux)
echo > /proc/sys/net/core/wmem_max
echo > /proc/sys/net/core/rmem_max
echo " " > /proc/sys/net/ipv4/tcp_rmem
echo " " > /proc/sys/net/ipv4/tcp_wmem

27. Sample 1 - Cont.
iperf -s -w 2000k -i 10 > iperf-tcp.log
Server listening on TCP port 5001
TCP window size: 3.91 MByte (WARNING: requested 1.95 MByte)
[ 4] local port 5001 connected with port 41265
[ 4] sec MBytes Mbits/sec
[ 4] sec MBytes Mbits/sec
[ 4] sec MBytes Mbits/sec
[ 4] sec MBytes Mbits/sec
[ 4] sec MBytes Mbits/sec
[ 4] sec MBytes Mbits/sec

28. Sample 1 - Cont.

29. Sample 2
Using Iperf to make a transference from UNAM at México city to the CLARA Pop in Tijuana at least 10Mbps
Making the same that in the first sample with out change the default TCP the transferences were below 2Mbps
After a TCP tuning we got transference of more of 10Mbps

30. Conclusions
A big issue in the CLARA are the slow transferences on TCP but UDP don´t have problem and all the BW can be used
Increasing the TCP window and Multiples session of TCP can help to make more faster file transferences
We are looking for others solutions more plug and play to solve the slow file transferences
Web100
gridFTP
Bulk transference protocol

31. Conclusions - Cont. Join to others international measurement projects
Join more LA-NRENs to be on the team

32. Thanks !!