Jose Saldana, Luigi Iannone, Diego R. Lopez, Julian Fernandez-Navajas, Jose Ruiz-Mas, "Enhancing Throughput Efficiency via Multiplexing and Header Compression over LISP Tunnels" . In Proc. Second IEEE Workshop on Telecommunication Standards: From Research to Standards, Collocated with IEEE ICC 2013, Budapest, Hungary. ISBN 9781467357524
This article explores the possibility of using traffic optimization techniques within the context of the LISP (Locator/ Identifier Separation Protocol) framework. These techniques use Tunneling, Multiplexing and header Compression of Traffic Flows (TCMTF) in order to save bandwidth and to reduce the amount of packets per time unit. Taking into account that encapsulation is necessary in LISP, bandwidth can be drastically reduced in flows using small packets, which are typical of many real-time services. The ability of the LISP framework to manage the signaling of TCMTF options is also studied. An analytical expression of the savings, as a function of the different header sizes, is devised and used to calculate the maximum expected savings. Different services and scenarios of interest are identified, and this allows the consideration of tests with real traffic traces, showing the savings as a function of the multiplexing period, and demonstrating that the additional delays can be acceptable for real-time services.
1. Enhancing Throughput Efficiency
via Multiplexing and Header
Compression over LISP Tunnels
Second IEEE Workshop on Telecommunication Standards: From Research to Standards
IEEE ICC 2013, Budapest, Hungary, 9th of June 2013
Jose Saldana
Julián Fernández-Navajas
José Ruiz-Mas
Luigi Iannone Diego R. Lopez
2. Enhancing Throughput Efficiency with LISP. Budapest Jun 9th 2013
Index
1. Introduction
2. Context and Scenarios of Application
3. Multiplexing/Compression Signaling
4. Expected Bandwidth Savings
5. Conclusions and Future Work
3. Enhancing Throughput Efficiency with LISP. Budapest Jun 9th 2013
Index
1. Introduction
2. Context and Scenarios of Application
3. Multiplexing/Compression Signaling
4. Expected Bandwidth Savings
5. Conclusions and Future Work
4. Introduction
Enhancing Throughput Efficiency with LISP. Budapest Jun 9th 2013
Emerging real-time services
High interactivity requirements
Delay is important, so frequent information
updates are needed
5. Introduction
Enhancing Throughput Efficiency with LISP. Budapest Jun 9th 2013
Emerging real-time services
High rates (10 to 50 pps)
Small packets (some tens of bytes)
Low efficiency
Packet size and inter-packet time for Counter Strike 1
40 50 60 70 80 90 100 110
bytes
0 10 20 30 40 50 60 70
ms
6. Introduction
Enhancing Throughput Efficiency with LISP. Budapest Jun 9th 2013
TCMTF (Tunneling Compressed Multiplexed
Traffic Flows) is a proposal for improving the
efficiency of these flows by:
Header compression
Multiplexing
Tunneling
Status: IETF draft
IP IP IP
No compr. / ROHC / IPHC / ECRTP
PPPMux / Other
GRE / L2TP / Other
IP
Compression layer
Multiplexing layer
Tunneling layer
Real-time traffic
Network Protocol
UDP
RTP
payload
UDPTCP
payloadpayload
7. Introduction
Enhancing Throughput Efficiency with LISP. Budapest Jun 9th 2013
TCMTF 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%
savingOne IPv4 TCMTF Packet multiplexing five two sample packets
η=100/161=62%
8. Introduction
Enhancing Throughput Efficiency with LISP. Budapest Jun 9th 2013
2006: The IAB (Internet Architecture Board)
felt the need for new architectures able to
overcome the scalability of the routing system
LISP: Locator/ID Separation Protocol, is an
architecture designed to this aim
Getting a growing interest from the Industry
9. Introduction
Enhancing Throughput Efficiency with LISP. Budapest Jun 9th 2013
LISP distinguishes two address spaces:
Routing Locator (RLOC): border routers
Endpoint Identifiers (EID): hosts inside stub
networks
Internet
RLOC Address Space
Stub 1
Stub 2
Border routers
EID 1
EID 2
EID 3
EID Address
Space
Stub 3
EID 1
EID 2
EID 2
EID 1
EID Address
Space
EID Address
SpaceRLOC RLOC
RLOC
10. Introduction
Enhancing Throughput Efficiency with LISP. Budapest Jun 9th 2013
A stub network only routes packets to and
from itself
Internet
Stub 1
Stub 2
Border routers
EID 1
EID 2
EID 3
Stub 3
EID 1
EID 2
EID 2
EID 1
11. Introduction
Enhancing Throughput Efficiency with LISP. Budapest Jun 9th 2013
A stub network only routes packets to and
from itself
Border routers do a “map and encap” process
when sending a packet to other stub network
Internet
Stub 1
Stub 2
Border routers
EID 1
EID 2
EID 3
Stub 3
EID 1
EID 2
EID 2
EID 1
12. Introduction
Enhancing Throughput Efficiency with LISP. Budapest Jun 9th 2013
A tunnel is necessary between stub networks
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
IP stub+UDP+RTP
40 bytes
VoIP: 76 header bytes
for 20 bytes payload
In a MTU-sized packet the extra
overhead is not significant
13. Enhancing Throughput Efficiency with LISP. Budapest Jun 9th 2013
Index
1. Introduction
2. Context and Scenarios of Application
3. Multiplexing/Compression Signaling
4. Expected Bandwidth Savings
5. Conclusions and Future Work
14. Scenarios of application
Enhancing Throughput Efficiency with LISP. Budapest Jun 9th 2013
Services generating high rates of small packets:
VoIP
Multiplexing schemes exist (RFC4170)
First Person Shooter games
MMORPG games
ACKs traveling to:
Content Delivery Networks
TCP-based video streaming web
15. Scenarios of application
Enhancing Throughput Efficiency with LISP. Budapest Jun 9th 2013
Can we find simultaneous flows between the
same pair of stub networks?
Internet
RLOC Address Space
Stub 1
Stub 3
Stub 2
Border routers
Web server
aggregation network of
a network operator
16. Scenarios of application
Enhancing Throughput Efficiency with LISP. Budapest Jun 9th 2013
Can we find simultaneous flows between the
same pair of stub networks?
Internet
RLOC Address Space
Stub 1
Stub 3
Stub 2
Border routers
Company headquarters
Office in a country
17. Scenarios of application
Enhancing Throughput Efficiency with LISP. Budapest Jun 9th 2013
Let’s group packets in the border router, in
order to share the overhead of the tunnel
Internet
RLOC Address Space
Stub 1
Stub 3
Stub 2
Border routers
4 IP/UDP/LISP headers
18. Scenarios of application
Enhancing Throughput Efficiency with LISP. Budapest Jun 9th 2013
Let’s group packets in the border router, in
order to share the overhead of the tunnel
Internet
RLOC Address Space
Stub 1
Stub 3
Stub 2
Border routers
1 IP/UDP/LISP header
19. Enhancing Throughput Efficiency with LISP. Budapest Jun 9th 2013
Index
1. Introduction
2. Context and Scenarios of Application
3. Multiplexing/Compression Signaling
4. Expected Bandwidth Savings
5. Conclusions and Future Work
20. Multiplexing/Compression Signaling
Enhancing Throughput Efficiency with LISP. Budapest Jun 9th 2013
We have to negotiate different parameters between
mux and demux
Maximum added delay
Header compression scheme
LISP signalling for EID-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
22. Enhancing Throughput Efficiency with LISP. Budapest Jun 9th 2013
Index
1. Introduction
2. Context and Scenarios of Application
3. Multiplexing/Compression Signaling
4. Expected Bandwidth Savings
5. Conclusions and Future Work
23. Expected Bandwidth Savings
Enhancing Throughput Efficiency with LISP. Budapest Jun 9th 2013
Let’s multiplex packets, avoiding LISP headers
Three IPv4/UDP client-to-server packets of Counter Strike
TCMTF multiplex
multiplex saving
Four IPv4/UDP/RTP VoIP packets with two samples of 10 bytes
multiplex saving
TCMTF multiplex
Four IPv4/TCP client-to-server packets of World of Warcraft. E[P]=20bytes
TCP ACK without payload
TCMTF multiplex
multiplex saving
Five IPv4/TCP ACKs
TCMTF multiplex
multiplex saving
24. Expected Bandwidth Savings
Enhancing Throughput Efficiency with LISP. Budapest Jun 9th 2013
What if the router also compresses headers?
Three IPv4/UDP client-to-server packets of Counter Strike
TCMTF multiplex and compress
multiplex + compress saving
Four IPv4/UDP/RTP VoIP packets with two samples of 10 bytes
TCMTF multiplex and compress
Four IPv4/TCP client-to-server packets of World of Warcraft. E[P]=20bytes
multiplex + compress saving
TCP ACK without payload
multiplex + compress saving
multiplex saving
TCMTF multiplex
TCMTF multiplex
multiplex saving
TCMTF multiplex and compress
TCMTF multiplex
multiplex saving
25. Expected Bandwidth Savings
Enhancing Throughput Efficiency with LISP. Budapest Jun 9th 2013
What if the router also compresses headers?
Three IPv4/UDP client-to-server packets of Counter Strike
TCMTF multiplex and compress
multiplex + compress saving
Four IPv4/UDP/RTP VoIP packets with two samples of 10 bytes
TCMTF multiplex and compress
Four IPv4/TCP client-to-server packets of World of Warcraft. E[P]=20bytes
multiplex + compress saving
TCP ACK without payload
multiplex + compress saving
multiplex saving
TCMTF multiplex
TCMTF multiplex
multiplex saving
TCMTF multiplex and compress
TCMTF multiplex
multiplex saving
26. Expected Bandwidth Savings
Enhancing Throughput Efficiency with LISP. Budapest Jun 9th 2013
Asymptotic savings for each service
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
VoIP FPS MMORPG ACKs
BandwidthSaving
Bandwidth Saving IPv4 on IPv4
IPv6 on IPv4
IPv4 on IPv6
IPv6 on IPv6
No header
compression
UDP/RTP
UDP
TCP
27. Expected Bandwidth Savings
Enhancing Throughput Efficiency with LISP. Budapest Jun 9th 2013
VoIP G.729a
0%
10%
20%
30%
40%
50%
60%
70%
80%
0 5 10 15 20 25 30 35 40 45 50
Bandwidthsaving
Number of VoIP flows
Bandwidth saving, VoIP, G729a, 2 samples per packet
IPv4 on IPv4 only Mux
IPv4 on IPv4 Mux + compr
IPv4 on IPv6 Only Mux
IPv4 on IPv6 Mux + Compr
28. Expected Bandwidth Savings
Enhancing Throughput Efficiency with LISP. Budapest Jun 9th 2013
VoIP G.729a
0%
10%
20%
30%
40%
50%
60%
70%
80%
0 5 10 15 20 25 30 35 40 45 50
Bandwidthsaving
Number of VoIP flows
Bandwidth saving, VoIP, G729a, 2 samples per packet
IPv4 on IPv4 only Mux
IPv4 on IPv4 Mux + compr
IPv4 on IPv6 Only Mux
IPv4 on IPv6 Mux + Compr
No header
compression
One packet
from each flow
29. Expected Bandwidth Savings
Enhancing Throughput Efficiency with LISP. Budapest Jun 9th 2013
First Person Shooter game (Counter Strike 1)
0%
10%
20%
30%
40%
50%
60%
70%
80%
5 10 15 20 25 30 35 40 45 50
BandwidthSaving
period (ms)
Bandwidth Saving. FPS Game. IPv4 on IPv4
20 players 20 players, no compr
15 players 15 players, no compr
10 players 10 players, no compr
5 players 5 players, no compr
30. Expected Bandwidth Savings
Enhancing Throughput Efficiency with LISP. Budapest Jun 9th 2013
First Person Shooter game (Counter Strike 1)
0%
10%
20%
30%
40%
50%
60%
70%
80%
5 10 15 20 25 30 35 40 45 50
BandwidthSaving
period (ms)
Bandwidth Saving. FPS Game. IPv4 on IPv4
20 players 20 players, no compr
15 players 15 players, no compr
10 players 10 players, no compr
5 players 5 players, no compr
Additional delay
is half the period
31. Expected Bandwidth Savings
Enhancing Throughput Efficiency with LISP. Budapest Jun 9th 2013
MMORPG (World of Warcraft)
0%
10%
20%
30%
40%
50%
60%
70%
80%
10 20 30 40 50 60 70 80 90 100
BandwidthSaving
period (ms)
Bandwidth Saving. MMORPG Game. IPv4 on IPv4
100 players 100 pl, no compr
50 players 50 pl, no compr
20 players 20 pl, no compr
10 players 10 pl, no compr
32. Expected Bandwidth Savings
Enhancing Throughput Efficiency with LISP. Budapest Jun 9th 2013
MMORPG (World of Warcraft)
0%
10%
20%
30%
40%
50%
60%
70%
80%
10 20 30 40 50 60 70 80 90 100
BandwidthSaving
period (ms)
Bandwidth Saving. MMORPG Game. IPv4 on IPv4
100 players 100 pl, no compr
50 players 50 pl, no compr
20 players 20 pl, no compr
10 players 10 pl, no compr
56% of the
packets are ACKs
33. Expected Bandwidth Savings
Enhancing Throughput Efficiency with LISP. Budapest Jun 9th 2013
ACKs between stub networks (no compress)
0%
10%
20%
30%
40%
50%
60%
70%
80%
5 10 15 20 25 30 35 40 45 50
BandwidthSaving
period (ms)
Bandwidth Saving. ACKs. IPv4 on IPv4
1000ACK/sec
500ACK/sec
200ACK/sec
100ACK/sec
34. Expected Bandwidth Savings
Enhancing Throughput Efficiency with LISP. Budapest Jun 9th 2013
ACKs between stub networks (no compress)
0%
10%
20%
30%
40%
50%
60%
70%
80%
5 10 15 20 25 30 35 40 45 50
BandwidthSaving
period (ms)
Bandwidth Saving. ACKs. IPv4 on IPv4
1000ACK/sec
500ACK/sec
200ACK/sec
100ACK/sec
Additional delay
has to be limited
35. Enhancing Throughput Efficiency with LISP. Budapest Jun 9th 2013
Index
1. Introduction
2. Context and Scenarios of Application
3. Multiplexing/Compression Signaling
4. Expected Bandwidth Savings
5. Conclusions and Future Work
36. Conclusions and Future Work
Enhancing Throughput Efficiency with LISP. Budapest Jun 9th 2013
Possibility of using TCMTF multiplexing and
compressing in LISP: mutual benefit
Ability of LISP signaling for negotiating
TCMTF parameters
Throughput can be highly improved by packet
grouping
Additional savings by means of compression
Depending on the capacity of the router
Future: TCMTF-able LISP border routers
37. Thank you very much!
Jose Saldana
Julián Fernández-Navajas
José Ruiz-Mas
Luigi Iannone Diego R. Lopez