Transcript 슬라이드 제목 없음 - Asia Pacific Advanced Network

Broadband
Communication
Lab.
One Way Delay Measurement
in KOREN MPLS Networks
Sok-Hyun Jung
[email protected]
Broadband Telecommunication Lab.
Dep. of Electronics Engineering, Korea University
APAN 2003 Meeting
Experiment Description
Objectives of Experiment
To measure End-to-End One way Delay through various
Transit networks such as Conventional IP and MPLS
networks
To evaluate MPLS performance
Related project
We had a grant of foundation of Korean National
Computerization Agency
Experiment organization
Korea University (KU), Chonnam University (CHNU)
Broadband Communication Lab.
2
APAN 2003 Meeting
Overall Experiment Environment
GPS satellites
Chonna
mUniv.
(CHNU)
KOREN
MPLS
Conventional
Networks
LabelTransit
Switched
Path (LSP)
network
IP network
Korea
KU
Univ.
(KU)
One way delay
measurement
KU connects with CHNU through various transit networks
Such as Conventional IP and KOREN MPLS networks
In case of KOREN MPLS networks, LSP is set up between KU
and CHNU
Broadband Communication Lab.
3
APAN 2003 Meeting
KOREN networks
Korea
Univ.
(KU)
Constitutions of KOREN
ATM switches added
MPLS function
LER & LSR
Functions of KOREN
Currently, provides only
basic functions such as
LSP setup using the LDP
(Label Distribution
Protocol) and so on.
In the future, provides
TE, VPN, QoS and so on.
Chonna
mUniv.
(CHNU)
Broadband Communication Lab.
4
APAN 2003 Meeting
Transit Network 1 –
CHNU
Conventional IP network
Conventional
IP network
168.131.161.30
KU
203.255.252.65
One way delay
measurement
This network is compared with other transit networks
such as Packet based MPLS and ATM based MPLS
Broadband Communication Lab.
5
APAN 2003 Meeting
Transit Network 2 -
Packet based MPLS
Packet based MPLS network
Swap
Table
CHNU
168.131.161.30
X
CISCO
7204
Y
GSR
Router
Seoul
Swap
Table
Y
Z
GSR
Router
Daejon
Swap
Table
Z
X
GSR
CISCO
Router
Kwanju 7505
KU
203.255.252.65
Label Switched Path (LSP)
MPLS runs over the Ethernet as link layer protocol
LSP is set up between CHNU and KU
Broadband Communication Lab.
6
APAN 2003 Meeting
Transit Network 3 -
ATM based MPLS
ATM based MPLS network
(Fore ASX200BX)
168.131.161.30
CISCO
7204
(Nortel 15000 ) (Nortel 15000 )
Seoul
Daejon
(Nortel 15000)
Kwangju
203.255.252.65
(RealCell)
CISCO
7505
KU
CHNU
Allocating New VP
Unlike Packet based MPLS, ATM based MPLS makes
use of ATM as link layer protocol
New VP is allocated from CHNU to KU and similar to
LSP
Broadband Communication Lab.
7
APAN 2003 Meeting
How to measure ?
AMS: Active
Measurement System
GPS satellites
Sender
Host1
Receiver
Transit Networks
Host2
AMS
AMS
CHNU
KU
1. To Synchronize between two end systems using GPS
2. Host1 sends one way traffics to host2 through various
Transit networks
3. Host2 receives traffics, and measures delay
4. To compare measured results for each Transit networks
Broadband Communication Lab.
8
APAN 2003 Meeting
Result 1 :
Conventional IP network
One way delay is about 10 msec
Table1: One-way Delay Measurement for conventional IP at KU
Table2: Ping test for conventional IP at KU
Broadband Communication Lab.
9
APAN 2003 Meeting
Result 2 :
packet based MPLS
Packet based MPLS network
CHNU
168.131.161.30
Host1
CISCO
Router
(LER)
ATM
CISCO
Router
(LSR)
ATM
CISCO
Router
(LSR)
ATM
ATM
ATM
KU
CISCO
Router
(LER)
203.255.252.65
Host2
ATM
One way delay is about 5 msec
It is smaller than that of the conventional IP network
Table3: One-way Delay measurement
for packet based MPLS at KU
Table4: Ping test for
packet based MPLS
at KU
Broadband Communication Lab.
10
APAN 2003 Meeting
Result 3 :
168.131.161.30
Host1
ATM based MPLS
CHNU
KU
CISCO
Router
(LER)
CISCO
Router
(LER)
ATM based MPLS network
ATM
ATM ATM
ATM ATM
203.255.252.65
Host2
ATM
One way delay is about 3.6 msec
It is smaller than that of the Packet based MPLS
Table5: One-way Delay
Measurement for ATM based
MPLS at KU
Table6: ping test for
ATM based MPLS
at KU
Broadband Communication Lab.
11
APAN 2003 Meeting
Conclusion
Types of Transit Networks
E2E One Way Delay
Case 1
Conventional IP
10.995 msec
Case 2
Packet based MPLS
4.995 msec
Case 3
ATM based MPLS
3.578 msec
Bad
Good
Delay in ATM based MPLS is the smallest value
ATM based MPLS can reduce one way delay, as compared
the Packet based MPLS and Conventional IP network
Broadband Communication Lab.
12
APAN 2003 Meeting
Thank you
for your attention !!
Broadband Communication Lab.
13
APAN 2003 Meeting