Network Working Group Mikael Degermark, Lulea University
INTERNET-DRAFT Hans Hannu, Ericsson
Expires: June 2000 Lars-Erik Jonsson, Ericsson
Krister Svanbro, Ericsson
Sweden
December 10, 1999
CRTP over cellular radio links
<draft-degermark-crtp-cellular-01.txt>
Status of this memo
This document is an Internet-Draft and is in full conformance with
all provisions of Section 10 of RFC2026.
Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF), its areas, and its working groups. Note that other
groups may also distribute working documents as Internet-Drafts.
Internet-Drafts are draft documents valid for a maximum of six months
and may be updated, replaced, or obsoleted by other documents at any
time. It is inappropriate to use Internet-Drafts as reference
material or cite them other than as "work in progress".
The list of current Internet-Drafts can be accessed at
http://www.ietf.org/ietf/lid-abstracts.txt
The list of Internet-Draft Shadow Directories can be accessed at
http://www.ietf.org/shadow.html
This document is an individual submission to the IETF. Comments
should be directed to the authors.
Abstract
This document evaluates the performance of a header compression
protocol for RTP, CRTP [RFC-2508], over links built on cellular radio
access technology. The key characteristics affecting CRTP performance
over such links are the high error rates and the relatively long
roundtrip time over the link.
Bandwidth is typically expensive in cellular radio access networks,
saving a single octet per voice packet can be equivalent to saving
many billion dollars in deployment since fewer base-stations are
Degermark, Hannu, Jonsson, Svanbro [Page 1]
INTERNET-DRAFT CRTP over cellular radio links December 10, 1999
needed. This is beneficial for operators as well as end-users who can
get cheaper wireless IP telephony service.
CRTP performance is evaluated for two kinds of link layers operating
over a realistic radio channel with high bit-error rates. Two main
conclusions are drawn. The first is that CRTP does not perform well
for this type of link. The second is that in high-error environments
it is very beneficial to have a checksum covering the compressed
header only, not the payload, so that the decompressor sees all non-
damaged headers. When a strong checksum covers the entire link layer
frame, header compression performs badly since too many headers are
discarded due to damaged payloads.
TABLE OF CONTENTS
1. Introduction..................................................3
2. Header compression............................................3
3. Link layers...................................................5
3.1. PPP in HDLC-like framing..............................5
3.2. Link layer with partial checksum......................6
4. Description of simulations....................................6
4.1. Simulated scenario....................................6
4.2. The cellular link and the back channel................7
5. Frame error rates (FER).......................................8
6. Evaluation of CRTP for cellular radio links...................8
6.1. An ideal header compression scheme....................9
6.2. CRTP without Twice...................................10
6.3. CRTP with Twice......................................12
6.4. Loss patterns........................................13
6.5. Using only COMPRESSED_NON_TCP packets................16
6.6. Using periodic refreshes instead of requests.........17
7. Conclusions..................................................19
7.1. How to improve CRTP performance......................20
8. Authors addresses............................................21
9. References...................................................21
Degermark, Hannu, Jonsson, Svanbro [Page 2]
INTERNET-DRAFT CRTP over cellular radio links December 10, 1999
1. Introduction
With IP telephony gaining momentum and cellular telephony having
hundreds of millions of users, it seems inevitable that some future
wireless telephony systems will be based on IP technology. What we
today know as cellular phones may in addition to telephony and video
have IP stacks, web browsers, email clients, networked games, etc. If
based on IP, the telephony service will be much more flexible than
today. This document concentrates on the problem of providing a good
IP solution for speech, but it is clear that applications for video,
games, etc, will also have to be supported.
It is vital for cellular phone systems to use the radio resources
efficiently in order to support a sufficient number of users per
cell. Only then can deployment costs be kept low enough. It will also
be important to provide sufficiently high quality voice and video. In
particular the voice service should be as good as what users expect
from the cellular phone systems of today. A lower quality may only be
accepted if costs are significantly lower than today.
The radio channels used in cellular systems have very high bit-error
rates (BER) due to shadow fading, multipath fading, and continuous
mobility. The radio signals of one user will interfere with the radio
signals of other users, so with the desired number of users per cell,
BERs will be high. Even after error correcting channel coding, the
remaining BER can be as high as 1e-3 (one in 1000) or even 1e-2 (one
in 100) in bad environments.
The only cost efficient way to achieve sufficient voice quality over
such channels is to use clever speech encoders and decoders that can
tolerate some damage to the encoded sound data. It is not feasible to
use a link layer that delivers all data reliably through an ARQ
scheme with link-local retransmission. High delays would be the
result. If the long maximum delays caused by an ARQ scheme were
acceptable, it would be better to spread the signal over time in
order to reduce the BER, rather than using an ARQ protocol. Neither
is it feasible to have the link layer discard all damaged frames. The
large fraction of discarded frames would result in insufficient
speech quality.
Unless explicitly stated otherwise, the numbers and figures presented
in this document are for IPv4 [RFC-791], not IPv6 [RFC-1883].
2. Header compression
Speech data for IP telephony will most likely be carried by RTP [RFC-
1889]. A packet will then, in addition to link-layer framing, have
an IP header (20 octets), a UDP header (8 octets), and an RTP header
(12 octets) for a total of 40 octets. With IPv6, the IP header is 40
octets for a total of 60 octets. The size of the payload depends on
Degermark, Hannu, Jonsson, Svanbro [Page 3]
INTERNET-DRAFT CRTP over cellular radio links December 10, 1999
the speech encoding used and the packet rate; it can be as low as 15-
20 octets.
From these numbers it is obvious that the header size must be reduced
for efficiency reasons. A proposed standard for compressing
RTP/UDP/IP headers over low-speed serial links, CRTP, has recently
been approved by the IESG [RFC-2508, RFC-2507], together with a way
to negotiate parameters for header compression over PPP [RFC-2509].
With CRTP, compressed headers are as small as 2 octets if the UDP
checksum is disabled.
CRTP uses delta encoding where compressed headers carry differences
from the previous header. The decompressor maintains state, known as
the context, that represents what the header looks like, how it is
expected to change, etc. The differences carried in each compressed
packet updates the context, and thus loss of a packet will bring the
context of the decompressor out of sync with the compressor as it is
not updated correctly.
CRTPs mechanism for bringing the decompressor context in sync with
the compressor relies on messages from the decompressor reporting its
state to the compressor. Such CONTEXT_STATE messages cause the
compressor to send packets with more information in their headers to
update the context of the decompressor: either FULL_HEADER packets
with 40 octet headers (60 for IPv6), or COMPRESSED_NON_TCP packets
with compressed UDP/IP headers but a complete RTP header. Headers in
COMPRESSED_NON_TCP packets are 17 octets if the UDP checksum is
disabled, and 19 octets otherwise (15 and 17 octets for IPv6,
respectively).
CRTP uses a link sequence number, incremented by one for each packet
with a compressed header, to detect lost packets. The link sequence
number ranges between 0 and 15. Gaps in the sequence number space
triggers the context repair mechanism outlined in the previous
paragraph.
High BERs will cause the repair mechanism to be triggered often,
causing many FULL_HEADER packets or COMPRESSED_NON_TCP packets to be
sent, which consume extra bandwidth. With a long roundtrip time over
the link, each damaged packet can cause several subsequent packets to
be discarded due to mismatching contexts.
The "Twice" mechanism proposed for compressed TCP [RFC-793] headers
in [RFC-2507] and also for CRTP in [RFC-2508] can often repair the
context and avoid some of the loss caused by mismatching contexts.
The assumption behind the "Twice" mechanism is that the delta of a
lost CRTP packet is often the same as the delta of the subsequent
packet. An attempt to repair the context by applying the delta twice
will therefore often succeed. Successful repairs are detected by a
matching transport-layer checksum.
Degermark, Hannu, Jonsson, Svanbro [Page 4]
INTERNET-DRAFT CRTP over cellular radio links December 10, 1999
3. Link layers
When evaluating CRTP, the link layer must be considered. We will use
two different link layers. One is PPP in HDLC-like framing [RFC-
1662], which has a 16/32-bit CRC covering the entire frame. This
implies that all damaged frames will be discarded at the link layer
since the checksum will fail. It is possible to change the networking
code to have such frames delivered, but then it is pointless to have
the checksum in the first place and a framing scheme without a
checksum would be a better solution.
For header compression purposes it is important that headers are not
damaged over the link. As outlined in the introduction, however,
damage to the payload is often acceptable to the (speech) decoder of
the application. It would therefore make sense to have a checksum
which only covers the header part of a packet. That should increase
the number of headers seen by the decompressor and improve header
compression performance. The second link layer we use for evaluation
purposes is an imaginary such link layer, henceforth called the Link-
Layer with Partial Checksum (LLPC).
3.1. PPP in HDLC-like framing (HDLC)
PPP typically uses HDLC-like framing [RFC-1662]. With a 16-bit
checksum and compressed Address and Control fields, frames carrying
CRTP, COMPRESSED_NON_TCP, or FULL_HEADER packets have the following
format.
1 1 2
+----------+----------+-------------+----------+----------+
| Flag | Protocol | Information | FCS | Flag |
| 01111110 | 8 bits | * | 16 bits | 01111110 |
+----------+----------+-------------+----------+----------+
The Flag only occurs once between frames if they are sent back-to-
back, so the amortized framing overhead is 4 octets per frame. The
checksum (FCS) is calculated over the Protocol field and the
Information field (payload), but not the Flags or the checksum
itself.
Any errors anywhere in the frame will cause the FCS to fail. The
frame will then be discarded.
Degermark, Hannu, Jonsson, Svanbro [Page 5]
INTERNET-DRAFT CRTP over cellular radio links December 10, 1999
3.2. Link-layer with partial checksum (LLPC)
This is an imaginary framing scheme derived from the HDLC-format in
3.1 by adding a one-octet Length field.
1 1 1 2
+----------+----------+----------+-------------+---------+----------+
| Flag | Length | Protocol | Information | FCS | Flag |
| 01111110 | 8 bits | 8 bits | * | 16 bits | 01111110 |
+----------+----------+----------+-------------+---------+----------+
The Length field indicates how many octets of the payload that are
covered by the FCS. It can have values from 0 to 255. The FCS covers
the Length and Protocol field plus as many octets in the beginning of
the Information field as indicated by the Length field. The value of
the Length field must not make the FCS extend over the FCS field.
When sending a FULL_HEADER packet, the Length field would have the
value 40, since it should protect the IP, UDP, and RTP headers. When
sending a minimal COMPRESSED_RTP packet, the Length field would have
the value 2. The amortized framing overhead for LPC is 5 octets per
frame.
Any errors in the Flag, Length, Protocol, FCS, or the initial Length
octets of the Information field will cause the FCS to fail. The frame
will then be discarded. Errors in the Information field after the
first Length octets will not affect the FCS and will not cause the
frame to be discarded.
4. Description of simulations
Section 4.1 describes the simulated scenario and 4.2 elaborates on
the properties of the cellular link and the back channel.
4.1. Simulated scenario
A source generates RTP packets containing speech data and sends them
across the Internet to an end-system. The end-system is connected to
the Internet over a cellular link over which the RTP stream is
compressed using CRTP.
Degermark, Hannu, Jonsson, Svanbro [Page 6]
INTERNET-DRAFT CRTP over cellular radio links December 10, 1999
Compression
Source point End-system
_____ ________ +-------+
/ back channel\ | |
+----+ +----+/ \+----+ |
| |--------->---------| HC |-------->--------| HD | |
+----+ Internet path +----+ Cellular link +----+ |
(loss) | |
+-------+
Figure 0: Simulated scenario
Over the Internet path there are uniformly distributed losses which
influence the efficiency of CRTP mechanisms, and especially the
"Twice" mechanism.
Over the Cellular link one of the framing protocols of section 3
carry the packets. The radio channel of the cellular link is
simulated accurately for various BERs and represents fairly bad, but
realistic, conditions. The roundtrip time can be varied.
The compressor (HC) at the compression point compresses RTP/UDP/IP
headers according to CRTP, and sends them over the cellular link to
the decompressor (HD). When HD detects that the context is out of
sync, it will send CONTEXT_STATE messages back to HC over the back
channel.
The speech source generates packets with payloads of a fixed size, 16
octets (representing the smallest reasonable payload size), at a rate
of 50 packets per second (20 ms worth of sound data per packet).
Silence suppression is used. The lengths of talk spurts and the
silent intervals between them are both exponentially distributed with
an expected length of 1 second. Loss over the Internet path due to
congestion is uniformly distributed. This loss pattern is reasonably
accurate since packet intervals are relatively long compared to
congestion related loss events.
4.2. The cellular link and the back channel
The cellular link is simulated accurately using a realistic radio
channel model [WCDMA] and adding channel coding. The reported bit
error rates, BER, are always the BERs after channel coding, i.e., the
BER seen by the link layer.
The interesting BERs for cellular systems are in the range between
1e-3 (1/1000) and 1e-6 (1/1000000). Circuit-switched cellular voice
transmission can deliver acceptable speech quality down to around
1e-2, while the systems become expensive at BERs much less than 1e-6.
Degermark, Hannu, Jonsson, Svanbro [Page 7]
INTERNET-DRAFT CRTP over cellular radio links December 10, 1999
The compressor repairs the decompressor context after feedback in the
form of a CONTEXT_STATE message from the decompressor. This means
that the roundtrip time over the link determines the speed of the
repair mechanism. The back channel used in our simulations never
damages CONTEXT_STATE messages.
5. Frame error rates (FER)
Frames can have errors due to damage over the link. This kind of
damage can be further classified into
a) header damage: damage to parts of the frame that are
important for header compression purposes. This is the
framing plus the compressed or full header.
b) payload damage: damage to other parts of the frame. Such
damage may or may not cause the frame to be unusable by the
speech decoder, depending on the coding and the location of
the damage. Also, it may or may not cause the entire frame
to be discarded depending on the framing format.
Frames can also be damaged because the decompressor fails to
reconstruct a correct header. That can of course be caused by a), but
also by
c) context damage: the context of the decompressor being out of
sync with the context of the compressor. This is caused by
delta information being lost due to a) or b).
For HDLC, both header damage and payload damage will cause the frame
to be discarded, which will increase the rate of frames discarded due
to context damage.
For LLPC, payload damage will not cause the frame to be dropped
before reaching the decompressor, which will reduce the number of
frames discarded due to context damage. Whether or not payload
damage causes the frames to be unusable for generating speech is not
related to header compression performance. We expect, however, that
most speech decoders will be able to utilize information in frames
with payload damage.
Degermark, Hannu, Jonsson, Svanbro [Page 8]
INTERNET-DRAFT CRTP over cellular radio links December 10, 1999
6. Evaluation of CRTP for cellular radio links
6.1. An ideal header compression scheme
In order to have a reference point, we first simulated an ideal
header compression scheme. The ideal header compression scheme can
always compress the header down to a total of 2 bytes and will never
fail at decompression, i.e., no frames will ever be discarded due to
context damage. Such a scheme is probably not achievable, but it
gives us something to compare the real CRTP against.
Figure 1: FER for Ideal scheme for HDLC and LLPC
As can be seen in figure 1, for a BER of 1e-3 the FER is 1-2 % for
both link layers. LLPC is marginally better. At 5e-3 there is a
significant difference between HDLC (7.5% FER) and LLPC (4% FER).
Loss over the Internet path does not affect the ideal header
compression scheme at all, and is not included in the reported FER.
Degermark, Hannu, Jonsson, Svanbro [Page 9]
INTERNET-DRAFT CRTP over cellular radio links December 10, 1999
There is no context damage for the ideal scheme. The difference
between the HDLC and LLPC curves show how many packets with payload
damage only there are: around 0.3% for a BER of 1e-3.
With some handwaving and contemplation of packet sizes and checksum
coverage, one can argue that LLPC should give a FER which is roughly
7/23 (30%) of the FER for HDLC if errors were uniformly distributed.
They are not, however, and it seems that LLPC in fact gives FERs that
are 55-60% of the FERs for HDLC.
6.2. CRTP without Twice
With a roundtrip time over the link corresponding to around 120 ms (a
realistic value), the slowness of the context repair mechanism will
multiply link layer related loss by a large factor. Figure 2 shows
CRTP performance for HDLC, while Figure 3 shows CRTP performance for
LLPC. The ideal curves have been included for reference. The percent
numbers indicate how much loss there were over the Internet path. The
plots for CRTP with Twice are discussed in the next subsection.
In figures 2 and 3 one can see that for a BER of 1e-3, CRTP gives a
FER of 8% with HDLC while with LLPC the FER is 5%. Given the
performance of the ideal scheme, it is clear that most of this loss
is due to context damage.
The average header size will increase with increasing loss over the
Internet path, since the delta between consecutive packets will then
often be different and more data need to be sent to represent the new
delta. A single loss over the Internet path will typically cause the
following two compressed headers to have three and two extra octets,
respectively. When one out of 10 packets are lost over the Internet
path, that would add 5 octets to the remaining 9 headers. The average
header size then increases with 5/9 octets (0.56 octets).
Figure 4 shows the average header size plotted against BERs, for
varying loss over the Internet path. At low BERs, HDLC and LLPC both
give an average header size of just over 2 octets when there is no
loss over the Internet path. When there is 10% loss over the Internet
path, both give an average header size just over 2.5 octets. This is
consistent with the expected increases in header sizes due to
different deltas after losses over the Internet path.
Degermark, Hannu, Jonsson, Svanbro [Page 10]
INTERNET-DRAFT CRTP over cellular radio links December 10, 1999
Figure 2: FER for CRTP, CRTP with Twice, and Ideal for HDLC
Figure 3: FER for CRTP, CRTP with Twice, and Ideal for LLPC
Degermark, Hannu, Jonsson, Svanbro [Page 11]
INTERNET-DRAFT CRTP over cellular radio links December 10, 1999
Figure 4: Average header size for CRTP
At higher BERs the average header size is determined by the rate of
COMPRESSED_NON_TCP headers (17 octets) sent over the cellular link.
CRTP compressors update the context state by sending such headers
whenever frames have been discarded over the cellular link. The
differences between the HDLC and LLPC curves at high BERs is due to
their different FERs. For a BER of 1e-3 and no Internet loss, CRTP
with HDLC gives an average header size of 2.7 octets, while CRTP with
LLPC gives 2.5 octets. For 10% Internet loss, HDLC gives 3.2 octets
and LLPC 3.0 octets.
6.3. CRTP with Twice
The Twice algorithm is a way to repair the context quickly without
having to wait for a roundtrip over the link. Twice makes assumptions
of what the lost delta was and tries to repair the context according
to those assumptions. When using Twice there must be a way to check
whether the repair succeeded, typically the UDP checksum is used for
that purpose.
The plots in figure 2 show FERs for CRTP, CRTP with Twice, and the
Ideal scheme, when HDLC framing is used. The CRTP with Twice curves
really show how successful Twice would be in repairing the context,
we have not actually enabled the UDP checksum in our simulations, but
Degermark, Hannu, Jonsson, Svanbro [Page 12]
INTERNET-DRAFT CRTP over cellular radio links December 10, 1999
instead we determined whether Twice would have succeeded. We wanted
to try out LLPC too (figure 3), and as the UDP checksum covers the
entire payload and is fairly weak, that scenario wouldn't make much
sense using the UDP checksum. Instead we chose to investigate how
successful Twice would be if there were some other means to detect a
successful repair.
It is evident from figure 2 that Twice improves the FER
significantly, although CRTP with Twice is still much worse than the
Ideal scheme. At a BER of 1e-3, the FERs are less than 2% for Ideal,
about 4% for CRTP with Twice, and 8% for CRTP. More sophisticated
implementations of Twice might get closer to the Ideal curve.
Figure 3 shows FERs for CRTP, CRTP with Twice, and the Ideal scheme,
when LLPC framing is used. The FERs are lower than for HDLC because
fewer frames are discarded at the link layer, but the plots are
otherwise similar.
6.4. Loss patterns
For applications such as interactive voice it is not only the loss
*rate* that is interesting. Typical voice decoders will reuse earlier
frames when a frame is lost, but might decrease the intensity with
which that frame is played out. For each successive loss the
intensity is decreased such that after a few consecutive lost frames
the sound will fade out completely. When only single frames are lost,
the tolerable FER might be high. A single burst of lost frames, on
the other hand, can cause a very noticeable pause. Figure 5 shows a
histogram over the number of consecutive loss bursts of certain
lengths for CRTP, with and without Twice, for three different BERs.
It is evident from figures 5a and 5b that the majority of loss events
without Twice are such that around 7 consecutive frames are lost. The
link roundtrip time in these simulations was 120 ms and the packet
rate 50 packets per second, which means that a single discarded frame
would cause 6 additional frames to be lost due to context damage.
When there is little loss over the Internet path, Twice (or variants)
are very efficient since deltas rarely change.
At higher BERs, COMPRESSED_NON_TCP packets are sent often and thus
lengths of frame loss bursts are less regular. Updates may be
damaged, and an earlier repair may cause an update which repairs new
damage.
Degermark, Hannu, Jonsson, Svanbro [Page 13]
INTERNET-DRAFT CRTP over cellular radio links December 10, 1999
Figure 5a: Lengths of frame loss bursts, HDLC, no IP loss
Degermark, Hannu, Jonsson, Svanbro [Page 14]
INTERNET-DRAFT CRTP over cellular radio links December 10, 1999
Figure 5b: Lengths of frame loss bursts, HDLC, 10% IP loss
Degermark, Hannu, Jonsson, Svanbro [Page 15]
INTERNET-DRAFT CRTP over cellular radio links December 10, 1999
Loss bursts involving 7-8 frames are clearly noticeable for most
voice decoders. This is a major disadvantage of using CRTP over high-
loss links with nontrivial link roundtrip times. Even if the frame
rate was one per 30 ms and the link roundtrip time was only 60 ms the
typical loss burst would be 3-4 frames (one discarded at link level,
next discovers damage, update requested, update sent), which would
decrease the voice quality significantly.
6.5. Using only COMPRESSED_NON_TCP packets
The high FERs for CRTP makes it interesting to compare its
performance against sending COMPRESSED_NON_TCP packets only. Their
headers are 17 octets. No frames are discarded due to context damage,
but on the other hand it is more likely that a packet will be damaged
because it is larger.
Figure 6: FER for COMPRESSED_NON_TCP only, HDLC and LLPC
Degermark, Hannu, Jonsson, Svanbro [Page 16]
INTERNET-DRAFT CRTP over cellular radio links December 10, 1999
Figure 6 shows the FER when sending COMPRESSED_NON_TCP packets only,
for HDLC and LLPC. For HDLC, the FER when the BER is 1e-3 is 3%,
which is more than for the Ideal scheme (<2%) but less than CRTP with
Twice (5%). The FER for LLPC is just over 2% and similarly to HDLC,
it is more than the Ideal scheme but less than CRTP with Twice.
6.6. Using periodic refreshes instead of requests
One alternative to the use of context updates on request could be to
periodically refresh the context as suggested in CRTP [RFC-2508] for
simplex links or links with high delay. However, to decrease the
packet loss rate due to context invalidation, the periodic refresh
method must update the context faster than the request based scheme,
which means that the compression slow-start mechanism described in IP
Header Compression [RFC-2507] would not be suitable. Instead, the
periodic refreshes must be sent with a shorter period than the link
round-trip time. Periodic refreshes could therefore be seen as a
solution somewhere between the ordinary request-based CRTP and the
completely difference-free solution used in 6.5, with only
COMPRESSED_NON_TCP packets.
The periodic refresh model evaluated here makes use of the
COMPRESSED_RTP and the COMPRESSED_NON_TCP packet types.
COMPRESSED_NON_TCP is used for every third and every fourth packet
respectively in two different simulations, the rest are
COMPRESSED_RTP. Figures 7 and 8 respectively show the packet loss
rates and header sizes for this scheme (both with refresh period
three and four) together with results for the ordinary CRTP and the
Ideal scheme. As shown in figure 7, the packet loss rate is
significantly decreased to about half as much as for the ordinary
solution, but it is still much higher than for the Ideal scheme. The
average header size on the other hand is increased about three times
to between six and seven octets.
A conclusion that could be drawn from this experiment is that a
periodic refreshing scheme would be costly in terms of header size if
it is supposed to improve the packet loss rate over links with a
round-trip time of 100-150 ms. With even longer RTT's, periodic
refreshes could be suitable, while for shorter RTT's the solution
would have no advantages over the request based scheme.
Degermark, Hannu, Jonsson, Svanbro [Page 17]
INTERNET-DRAFT CRTP over cellular radio links December 10, 1999
Figure 7: FER for CRTP with periodic refreshes, LLPC
Figure 8: Header sizes for CRTP with periodic refreshes, LLPC
Degermark, Hannu, Jonsson, Svanbro [Page 18]
INTERNET-DRAFT CRTP over cellular radio links December 10, 1999
7. Conclusions
The packet loss rate of CRTP, CRTP with Twice, and the Ideal header
compression scheme is summarized in table 1 for various error rates.
The numbers are for HDLC-like framing, i.e., errors in any part of a
packet means that it is discarded. The payload is 16 octets. The
Ideal scheme discards packets only when the packet itself is damaged.
Bit-error rate 1e-5 1e-4 1e-3 1e-2
-------------------------------------------
Ideal 0 0.4% 1.8% 11%
CRTP+Twice 0 1.0% 5.0% 24%
CRTP 0 1.5% 8.0% 40%
-------------------------------------------
Table 1: Frame loss rates of header compression schemes (HDLC).
It is evident from table 1 that CRTP performs well for BERs less than
1e-5, but not so well for BERs higher than 1e-4. If one considers a
scenario where the path of an IP telephony conversation has a
cellular link at both ends, the packet loss rates of CRTP and
CRTP+Twice become intolerable at high BERs.
The major cause of CRTPs bad performance is that many packets are
discarded due to context damage while waiting a link roundtrip time
for the repair mechanism.
Twice is a way to repair the context locally. It requires two extra
octets of header (the UDP checksum) to verify its repair attempts.
These two extra octets make it a less attractive solution. Moreover,
the straightforward Twice used in this evaluation does not have a
sufficiently high success rate. Combinations of link-loss at a first
cellular link and congestion-related loss in the rest of the path
will ensure that the compressor at the last cellular link will see
many holes in the packet stream. Twice will then fail often.
Moreover, the UDP checksum is too weak to reliably determine the
success or failure of attempted repairs.
The losses induced by CRTP and its variations are problematic not
only because they are high. The loss patterns are such that losses
occur in groups longer than a link roundtrip time. This is
problematic for low-bandwidth voice codecs, who cannot mask such long
loss events well. Hence, the speech quality will suffer. It is a
major disadvantage of CRTP that it causes such long loss events.
Degermark, Hannu, Jonsson, Svanbro [Page 19]
INTERNET-DRAFT CRTP over cellular radio links December 10, 1999
It is worth noticing that link layers which protect the header with
strong checksums, but not the payload, will decrease the packet-loss
rate significantly. Such link-layers will deliver more headers to the
decompressor and context damage will be less frequent. Table 2
summarizes the results for such a link-layer.
Bit-error rate 1e-5 1e-4 1e-3 1e-2
-------------------------------------------
Ideal 0 0.3% 1.4% 7%
CRTP+Twice 0 0.8% 4.2% 18%
CRTP 0 1.1% 5.0% 25%
-------------------------------------------
Table 2: Frame loss rates of header compression schemes (LLPC).
Overall, the improvement in FER were around 40% with a payload of 16
octets. When the payload is larger, the improvement will be higher.
In addition to the benefits for header compression, speech codecs for
lossy links can utilize information in damaged payloads and will
deliver higher quality speech when they have access to damaged
frames.
To summarize, CRTP does not perform well over lossy links with long
roundtrip times. Twice can improve the situation somewhat but the
loss is still too high. A disadvantage of using Twice is that it
requires that the UDP checksum is enabled, which will double the size
of the compressed header. CRTP with Twice performs much worse than
the Ideal scheme in terms of packet loss. Because the UDP checksum is
fairly weak, Twice should not be extended to attempt a large number
of repairs. Because of this, CRTP with Twice cannot approach the
performance of the Ideal scheme.
7.1. How to improve CRTP performance.
The link roundtrip time should be kept low. When it is high, local
repairs of the contex (without going over the link) is essential.
Sophisticated versions of Twice should be considered, which implies
that the UDP checksum must be enabled. Unfortunately, that adds 2
octets to the compressed header.
Degermark, Hannu, Jonsson, Svanbro [Page 20]
INTERNET-DRAFT CRTP over cellular radio links December 10, 1999
8. Author's Addresses
Mikael Degermark Tel: +46 920 911 88
Dept of CS & EE, Lulea Mobile: +46 70 833 89 33
University of Technology EMail: micke@sm.luth.se
Hans Hannu Tel: +46 920 20 21 84
Ericsson Erisoft AB Mobile: +46 70 378 04 73
Lulea, Sweden EMail: hans.hannu@ericsson.com
Lars-Erik Jonsson Tel: +46 920 20 21 07
Ericsson Erisoft AB Mobile: +46 70 365 20 58
Lulea, Sweden EMail: lars-erik.jonsson@ericsson.com
Krister Svanbro Tel: +46 920 20 20 77
Ericsson Erisoft AB Mobile: +46 70 531 25 08
Lulea, Sweden EMail: krister.svanbro@lu.erisoft.se
9. References
[RFC-768] J. Postel, User Datagram Protocol, RFC 768, August 1980.
[RFC-791] J. Postel, Internet Protocol, RFC 791, September 1981.
[RFC-793] J. Postel, Transmission Control Protocol, RFC 793,
September 1981.
[RFC-1144] V. Jacobson, Compressing TCP/IP Headers for Low-Speed
Serial Links, RFC 1144, February 1990.
[RFC-1662] W. Simpson, PPP in HDLC-like framing, RFC 1662, 1994.
[RFC-1883] S. Deering, R. Hinden, Internet Protocol, Version 6
(IPv6) Specification, RFC 1883, December 1995.
[RFC-1889] Henning Schulzrinne, Stephen Casner, Ron Frederick, Van
Jacobson, RTP: A Transport Protocol for Real-Time
Applications, RFC 1889, January 1996.
[RFC-2507] M. Degermark, B. Nordgren, S. Pink, IP header
compression, RFC 2507, February 1999.
[RFC-2508] S. Casner, V. Jacobson, Compressing IP/UDP/RTP Headers
for Low-Speed Serial Links, RFC 2508, February 1999.
[RFC-2509] M. Engan, S. Casner, C. Bormann, IP Header Compression
for PPP, RFC 2509, February 1999.
[WCDMA] Procedures for Evaluation of Transmission Technologies
for FPLMTS, ITU-R TG8-1, 8-1/TEMP/233-E, September 1995.
Degermark, Hannu, Jonsson, Svanbro [Page 21]
INTERNET-DRAFT CRTP over cellular radio links December 10, 1999
This Internet-Draft expires in June 2000.
Degermark, Hannu, Jonsson, Svanbro [Page 22]