IP Header compression in UMTS network

Thesis Work Presentation 14.02.2006

Author: Jukka Raunio

Supervisor: Prof. Raimo Kantola
Instructor: M. Sc. Antti Lehtonen
 Content
 Background
 Objectives & methodology
 Header Compression in UMTS network
 Header Compression mechanisms
 DSP processor
 Results
 Background
 The main service provided to the mobile users have
traditionally been only speech
 Currently the mobile equipment supports also the use
of the Internet
 IP telephony has also gained momentum thanks to
improved technical solutions
 The problem in packet switched connections over the
radio interface of the mobile network is the poor
bandwidth efficiency caused by the header overhead
-for example a common size of TCP segments for
bulk transfers over medium-speed links is 512
octets and when this packet tunneled over
IPv6, the IPv6/IPv6/TCP header is 100 octets
leading to header overhead of 19,5%
 Objectives & methodology
Objectives:
 Study the header compression mechanism
 Implement the Internet Protocol Header Compression
[RFC2507] mechanism on DSP
 Study the effects of the mechanism on the DSP

Methodology:
 Literature study based on the essential RFCs, the 3GPP
specifications and other available material.
 Measuring the implemented mechanism on the DSP
 Header compression fundamentals
 Header compression can be done because:
-much of the header information stays the same over the life-
time of a packet stream
-some changing fields change with a small or a predictable
value and thus incremental coding can be used e.g. TCP
sequence number
-only fields that change often and randomly need to be carried
in every packet e.g. checksums
-values of some fields can be inferred for example from the link
layer implementation e.g. the length of
the packet
 The general principle of header compression is to occasionally
send a packet with a full header; subsequent compressed
headers refer to the context established by the full header and
may contain incremental changes to the context.
 Header compression fundamentals
 The context is basically the uncompressed version of the last
header sent (compressor) or received (decompressor) over the
link.
 A context identifier (CID) is a small unique number identifying
the context that should be used to decompress a compressed
header an it is carried in full headers and compressed headers.
 Header compression benefits
 Improve interactive response time
-achieved by sending smaller packets
 Allow using small packets for bulk data with good line efficiency
-This is important when interactive (for example Telnet) and
bulk traffic (for example FTP) is mixed because the bulk data
should be carried in small packets to decrease the waiting time
when a packet with interactive data is caught behind a bulk
data packet
 Allow using small packets for delay sensitive low data-rate
traffic
-Header compression can reduce the bandwidth needed for
headers significantly
 Decrease header overhead
 Reduce packet loss rate over lossy links
-Fewer bits are sent per packet and thus the packet loss rate
will be lower for a given bit-error rate
 Header compression in UMTS network
 In UMTS network the header compression is performed between the
User Equipment and the Radio Network Controller
 The reason why the header compression is not done for example in the
node B is that the data encryption and the Macro Diversity Combining

(MDC) are done in the RNC.

 Header compression mechanisms
 UMTS network user plane protocol stack:

 3GPP specifies two different header compression

mechanisms to be used on the Packet Data Convergence
Protocol (PDCP) layer:
-Internet Protocol Header Compression (IPHC)
-RObust Header Compression (ROHC) specified
 Internet Protocol Header Compression
 Specified in the RFC 2507
 Can be applied to of IPv6 base and extension headers, IPv4
headers, TCP and UDP headers, and encapsulated IPv6 and
IPv4 headers
 More simple of the two header compression mechanisms
 The packets are classified into five categories:

-FULL_HEADER is used when a context is generated or

updated in the decompressor. The packet is similar to a
normal full header except that it contains a CID coded into
the length field of the IP header
-COMPRESSED_NON_TCP indicates a non-TCP packet with
compressed header
Internet Protocol Header Compression
-COMPRESSED_TCP packet contains a compressed
TCP packet containing a CID, a flag octet telling which
fields have changed since the previous compressed
packet, and the changed values encoded as a
difference from the previous value
-COMPRESSED_TCP_NO_DELTA indicates a packet
with a compressed TCP header where all fields that
are normally sent as the difference to the previous
value are instead sent as they are in the original
header.
-CONTEXT_STATE is a packet that may be used to
speed up the repair of the TCP streams over a links
where the decompressor can send packets to the
compressor.
 Example compression of a IPv4/TCP
packet

 The classification of the IPv4/TCP

header fields:
 Internet Protocol Header Compression
 The summarized gain of the header compression on
the IPHC:

Packet type Header Minimum compressed Max. compression gain [%]

size header size (octets)
(octets)

IPv4/TCP 40 4 90.00
IPv4/UDP 28 2 92.86
IPv6/TCP 60 4 93.33
IPv6/UDP 68 4 91.67
 Internet Protocol Header Compression

 Summary of the pros and cons of the

IPHC
Pros Cons
Easy to implement Not robust enough in a case
Compression of IP/TCP of a packet loss
headers No direct support for the
compression of RTP packets
Relies on the use underlying
link layer to give the length of
the packet
 Robust Header Compression
 ROHC is specified in RFC 3095
 Supports currently the compression of IP/UDP,
IP/UDP/RTP and IP/ESP packets
 ROHC is meant to be used mainly with streams
containing real-time data and it is developed to be fault
tolerant even in links with higher error rates
 The robustness is achieved with efficient communication
between the compressor and the decompressor
 ROHC can be characterized as an interaction between
two state machines, the compressor machine and the
decompressor machine
 Robust Header Compression
 ROHC contains three different operational
modes: Unidirectional, Optimistic Bidirectional
and Reliable Bidirectional mode. The mode
specifies the way that the compressor and
decompressor states transitions are made.
 Compressor and decompressor each have
three different states that control what kind
of packets are transmitted in each.
 Compressor states: Initialization and Refresh
(IR), First Order (FO) and Second Order (SO).
 Decompressor states: No Context, Static
Contexts and Full Context
Robust header Compression
 Robust Header Compression
 Pros and cons of Robust Header Compression

Pros Cons
Robustness Complexity
Quick context Requires significant
resynchronization processing/memory
Improved resources (compared
encoding scheme to IPHC)
for dynamical Emerging standard
header fields
 Measurements
 The aim of the measurements was study the feasibility of the IPHC
mechanism on a Digital Signal Processor
 The implementation and the measurements of the IPHC was done on
the Freescale MSC8101 DSP
 The implementation was coded with the C-language
 Measurements
 The execution time measurements. The values of the all measurements
are scaled as a relative difference to the smallest value

3.5

2.5

2
compress
decompress
1.5

0.5

0
ipv4/udp ipv4/tcp ipv6/udp ipv6/tcp
 Measurements
 Packet size effects on the execution time

3.5

2.5
Execution Time

1.5

0.5

0
0 200 400 600 800 1000 1200 1400 1600
Packet Size [bytes]

IPv4/TCP IPv4/UDP IPv6/TCP IPv6/UDP

 Measurements
 Compiler optimization results
Level 0 code is not optimized at all and so the
code produced is linear assembly.
Level 1 option performs all target
independent optimizations, such as function
3.5 inlining.
Level 2 optimization is the first optimization
3
level that produces parallel assembly
language code. Parallel assembly code is code
2.5
where instructions are performed in parallel,
when it is possible. Parallelization is possible
2
only when there are no dependencies
Compress between the instructions. All other target
Decompress
1.5
specific optimizations are also used.
1
Level 3 uses the same optimizations as level
2, but now global register allocation is used.
0.5
Register allocation is an optimization that
allocates the virtual registers to physical
0
registers or memory slots, attempting to
0 1 2 3 minimize the transfers from memory to
Optimization Level
registers.
 Measurements
 Memory optimization is a way to optimize the code by locating parts
of the code either in the internal or the external memory.

2.5

2 1. The code and libs are completely

in the external memory
1.5
2. Libs are located in the internal
Compress
Decompress memory, code in the external
1

3. Libs and compress and

0.5 decompress files in internal
memory, others in the external
0
1 2 3 4
memory
Memory Arrangement

4. Everything in the internal

memory
Thank you!