Download Slides - Locator/ID Separation Protocol

1
TCM optimization and LISP, LNOG003 Meeting, 26th of June 2014
Improving efficiency by header
compression and multiplexing in
scenarios using LISP *
Jose Saldana
Julián Fernández-Navajas
José Ruiz-Mas
University of Zaragoza
Luigi Iannone
Diego R. Lopez
Telecom ParisTech
Telefonica I+D
* Research paper presented at ICC 2013, Budapest, June 2013,
“Enhancing Throughput Efficiency via Multiplexing and Header
Compression over LISP Tunnels,” From research to standards Workshop
http://diec.unizar.es/~jsaldana/personal/budapest_ICC_2013_in_proc.pdf
2
Index
1.
2.
3.
4.
5.
Problem statement
Summary of TCM-TF
Can LISP and TCM work together?
TCM-TF Signaling
Expected Bandwidth Savings
TCM-optimization and LISP, LNOG003 Meeting, 26th of June 2014
Problem statement
3
 Emerging real-time services
 High interactivity requirements
 Delay is important, so frequent information
updates are needed
TCM-optimization and LISP, LNOG003 Meeting, 26th of June 2014
Problem statement
4
 Emerging real-time services
 High rates (10 to 50 pps)
 Small packets (some tens of bytes)
 Low efficiency
40 50 60 70 80 90 100 110
bytes
0
10 20 30 40 50 60 70
ms
Packet size and inter-packet time for Counter Strike 1
TCM-optimization and LISP, LNOG003 Meeting, 26th of June 2014
Problem statement
5
 What would happen in an “All LISP World”?
One IPv4/TCP packet 1500 bytes
One IPv4/UDP/RTP VoIP packet with two samples of 10 bytes
IP RLOC
20 bytes
UDP
8 bytes
LISP
8 bytes
In a MTU-sized packet the extra
overhead is not significant
IP stub+UDP+RTP
40 bytes
VoIP: 76 header bytes
for 20 bytes payload
TCM-optimization and LISP, LNOG003 Meeting, 26th of June 2014
6
Summary of TCM
 TCM-TF (Tunneling Compressed Multiplexed
Traffic Flows) is a proposal for improving the
efficiency of small-packet flows by means of:
 Header Compression
 Multiplexing
 Tunneling
payload
payload
RTP
UDP
UDP
IP
IP
Compression layer
No compr. / ROHC / IPHC / ECRTP
Multiplexing layer
PPPMux / Other
Tunneling layer
GRE / L2TP
MPLS
Network Protocol
IP
TCM-optimization and LISP, LNOG003 Meeting, 26th of June 2014
7
Summary of TCM
 TCM-TF optimization example
One IPv4/TCP packet 1500 bytes
η=1460/1500=97%
One IPv4/UDP/RTP VoIP packet with two samples of 10 bytes
η=20/60=33%
Five IPv4/UDP/RTP VoIP packets with two samples of 10 bytes
η=100/300=33%
One IPv4 TCMTF Packet multiplexing five two sample packets
η=100/161=62%
saving
TCM-optimization and LISP, LNOG003 Meeting, 26th of June 2014
Can LISP and TCM work together?
8
 Can we find simultaneous flows between the
same pair of stub networks?
Web server
Stub 2
aggregation network of
a network operator
Internet
RLOC Address Space
Stub 3
Stub 1
Border routers
TCM-optimization and LISP, LNOG003 Meeting, 26th of June 2014
Can LISP and TCM work together?
9
 Can we find simultaneous flows between the
same pair of stub networks?
Company headquarters
Stub 2
Office in a country
Internet
RLOC Address Space
Stub 3
Stub 1
Border routers
TCM-optimization and LISP, LNOG003 Meeting, 26th of June 2014
Can LISP and TCM work together?
10
 Let’s group packets in the border router, in
order to share the overhead of the tunnel
4 IP/UDP/LISP headers
Stub 2
Internet
RLOC Address Space
Stub 3
Stub 1
Border routers
TCM-optimization and LISP, LNOG003 Meeting, 26th of June 2014
Can LISP and TCM work together?
11
 Let’s group packets in the border router, in
order to share the overhead of the tunnel
1 IP/UDP/LISP header
Stub 2
Internet
RLOC Address Space
Stub 3
Stub 1
Border routers
TCM-optimization and LISP, LNOG003 Meeting, 26th of June 2014
TCM Signaling
12
 We have to negotiate different parameters between
mux and demux
 Maximum added delay
 Header compression scheme
 LISP signalling for Endpoint ID(EID)-Routing
Locators (RLOC) mappings can be used for this aim
 Able to carry meta-information
 Which flows can be multiplexed, based on (e.g.),
 IP addresses
 ToS
 application
TCM-optimization and LISP, LNOG003 Meeting, 26th of June 2014
TCM Signaling
13
 Standard format for signalling
 Start or adapt multiplexing on demand, depending
on network traffic status
TCM-optimization and LISP, LNOG003 Meeting, 26th of June 2014
Expected Bandwidth Savings
14
 Options: only mux / mux+header compr.
Four IPv4/UDP/RTP VoIP packets with two samples of 10 bytes
TCMTF multiplex
multiplex saving
TCMTF multiplex and compress
multiplex + compress saving
Three IPv4/UDP client-to-server packets of Counter Strike
TCMTF multiplex
multiplex saving
TCMTF multiplex and compress
multiplex + compress saving
TCP ACK without payload
Four IPv4/TCP client-to-server packets of World of Warcraft. E[P]=20bytes
TCMTF multiplex
multiplex saving
TCMTF multiplex and compress
multiplex + compress saving
TCM-optimization and LISP, LNOG003 Meeting, 26th of June 2014
Expected Bandwidth Savings
15
 Asymptotic savings for each service
Bandwidth Saving
100%
90%
TCP
UDP/RTP
80%
UDP
No header
compression
70%
Bandwidth Saving
IPv4 on IPv4
IPv6 on IPv4
IPv4 on IPv6
IPv6 on IPv6
60%
50%
40%
30%
20%
10%
0%
VoIP
FPS
MMORPG
ACKs
TCM-optimization and LISP, LNOG003 Meeting, 26th of June 2014
Expected Bandwidth Savings
16
 VoIP G.729a
Bandwidth saving, VoIP, G729a, 2 samples per packet
80%
70%
Bandwidth saving
60%
Header
compression
50%
No header
compression
40%
30%
IPv4 on IPv4 only Mux
20%
One packet
from each flow
10%
IPv4 on IPv4 Mux + compr
IPv4 on IPv6 Only Mux
IPv4 on IPv6 Mux + Compr
0%
0
5
10
15
20
25
30
Number of VoIP flows
35
40
45
50
TCM-optimization and LISP, LNOG003 Meeting, 26th of June 2014
Expected Bandwidth Savings
17
 First Person Shooter game (Counter Strike 1)
Bandwidth Saving. FPS Game. IPv4 on IPv4
80%
20 players
15 players
10 players
5 players
70%
Bandwidth Saving
60%
20 players, no compr
15 players, no compr
10 players, no compr
5 players, no compr
50%
40%
30%
20%
Additional delay
is half the period
10%
0%
5
10
15
20
25
30
period (ms)
35
40
45
50
TCM-optimization and LISP, LNOG003 Meeting, 26th of June 2014
Expected Bandwidth Savings
18
 MMORPG (World of Warcraft)
Bandwidth Saving. MMORPG Game. IPv4 on IPv4
80%
70%
Bandwidth Saving
60%
56% of the
packets are ACKs
50%
40%
30%
20%
100 players
50 players
20 players
10 players
10%
100 pl, no compr
50 pl, no compr
20 pl, no compr
10 pl, no compr
0%
10
20
30
40
50
60
period (ms)
70
80
90
100
TCM-optimization and LISP, LNOG003 Meeting, 26th of June 2014
Expected Bandwidth Savings
19
 ACKs between stub networks (no compress)
Bandwidth Saving. ACKs. IPv4 on IPv4
80%
1000ACK/sec
500ACK/sec
70%
200ACK/sec
Bandwidth Saving
60%
100ACK/sec
50%
40%
30%
20%
Additional delay
has to be limited
10%
0%
5
10
15
20
25
30
period (ms)
35
40
45
50
TCM-optimization and LISP, LNOG003 Meeting, 26th of June 2014
20
Thank you very much!
More info about TCM-TF:
•
Presentation of TCMTF.
•
"Tunneling Compressed Multiplexed Traffic Flows (TCMTF)" draftsaldana-tsvwg-tcmtf.
•
"Maximum Tolerable Delays when using Tunneling Compressed
Multiplexed Traffic Flows," draft-suznjevic-tsvwg-mtd-tcmtf.
•
TCMTF list info page
Jose Saldana, Luigi Iannone, Diego R. Lopez, Julián Fernández-Navajas, José Ruiz-Mas
The paper is here:
http://diec.unizar.es/~jsaldana/personal/budapest_ICC_2013_in_proc.pdf