Transcript CERNET`s experience

Distributed Monitoring
CERNET's experience
Xing Li
2004-07-03
1
CERNET Transport Network
2
CERNET IP Backbone
3
CERNET Statistics

38 GigaPops distributed in 36 cities, covering all the provinces in
Mainland China.
– 11 national Pops are connected via multiple 2.5Gbps DWDM links
– 27 provincial Pops are connected via multiple 155Mbps SDH links

1,500 education and research institutions connected
– 300 campus networks connect to their nearest Pops via 100Mbps-1Gbps
links.
– 15 million users
4
CERNET Distributed Measurement

Provide the high performance, reliable IP
backbone service

Survive from the network attack

Run based on self-funded model

Protect mission critical applications

produce research data
5
Comparison of Current Data and Tools

Netflow

Owamp (One-Way Latency)

Iperf

Traceroute

SNMP Interface Statistics

Internet2 Detective

Multicast Beacon

NTP Stratum 2 Server

Ping/Traceroute V6 Destination
6
Current Tools and Analysis

Throughput
– SNMP interface
– passive monitoring
– httpd+wget

Delay and Loss
– ICMP

Top 20
– passive monitoring

Multicast
– Beacon
7
Passive Monitoring
M2.1
M2.2
s
R2.1
12410
R20.1
6509
8
Performance
9
Loss vs. Delay
10
CERNET HTTP Performance
11
Top20 Histogram
12
Multicast Beacon Monitoring
13
Correlation Analysis

Delay Matrix

Loss Matrix

Combined Pop Monitoring

Committed bandwidth http performance

Top 20 Warning
14
CERNET Delay Matrix
15
CERNET Loss Matrix
16
Pop Monitoring
17
Committed bandwidth http performance
Global Internet
20M
CERNET
1G
Cx
18
Top20 Monitoring
19
Other Tools and Activities








Iperf
NTP
DNS
BGP
tcping
Out of Order Packet
Video Conference
CCERT
20
NTP
21
Address and Domains Report
22
IPv4/IPv6 BGP Analysis
23
tcping
24
Multicast Video Conference
25
CCERT
26
Comparison (1)

Netflow

Owamp (One-Way Latency)

Iperf

Traceroute

SNMP Interface Statistics

Internet2 Detective

Multicast Beacon

NTP Stratum 2 Server

Ping/Traceroute V6 Destination
27
Comparison (2)

Collection Today:
– Iperf (Throughput)
– OWAMP (1-Way Latency,
Loss)
– SNMP Data
– Anonymized Netflow Data
– Per Sender, Per Receiver, Per
Node Pair
– IPv4 and IPv6

Collection in the Future
–
–
–
–
NTP (Data)
Traceroute
BGP Data
First Mile Analysis

Correlation Today:
– “Worst 10” Throughputs
– “Worst 10” Latencies

Correlation in the Future:
– 99th Percentile Throughput
over Time
– Throughput/Loss for all E2E
paths using a specific link
– Commonalities among first
mile analyzers
– Sum of Partial Paths vs. Whole
Path
28
Comparison (3)

Analysis Today:
–
–
–
–
Throughput over Time
Latency over Time
Loss over Time
Worrisome Tests? (Any bad
apples in “Worst Ten”?)
– “Not the Network” (If
“Worst Ten” is good
enough)

Analysis in the Future:
– Latency vs. Loss
– How good is the network?
– Do common first mile
problems exist?
– Does a link have problems
that only manifest in the
long-haul?
– Is the network delivering
the performance required by
a funded project?
29
Future Work





Expend to the CERNET member universities
IPv6
New Tools
Data library
International collaboration
– APAN
– Abilene Observatory
30