Implementing A+P in the provider's IPv6-only network
draft-deng-softwire-aplusp-experiment-results-01
The information below is for an old version of the document.
| Document | Type |
This is an older version of an Internet-Draft whose latest revision state is "Expired".
|
|
|---|---|---|---|
| Authors | Xiaohong Deng , Yiu Lee , Tao Zheng , Xiaohong Huang , Qin Zhao , Yan Ma | ||
| Last updated | 2011-10-31 (Latest revision 2011-09-25) | ||
| RFC stream | (None) | ||
| Formats | |||
| Stream | Stream state | (No stream defined) | |
| Consensus boilerplate | Unknown | ||
| RFC Editor Note | (None) | ||
| IESG | IESG state | I-D Exists | |
| Telechat date | (None) | ||
| Responsible AD | (None) | ||
| Send notices to | (None) |
draft-deng-softwire-aplusp-experiment-results-01
Internet Engineering Task Force X.Deng
Internet Draft M.Boucadair
Intended status: Informational France Telecom
Expires: May 1, 2012 Y.Lee
Comcast
T.Zheng
L.Wang
France Telecom
X.Huang
Q.Zhao
Y.Ma
BUPT
Oct 31, 2011
Implementing A+P in the provider's IPv6-only network
draft-deng-softwire-aplusp-experiment-results-01.txt
Abstract
This memo focuses on the IPv6 flavor of A+P.
This memo describes an implementation of A+P in a provider's IPv6-
only network. It provides details of the implementation, network
elements, configurations and test results as well. Besides
traditional port range A+P, a scattered port sets flavor of A+P is
also implemented to verify feasibility of offering non-continuous
port sets with A+P approach, and to investigate possibility and
efforts of making UPnP 1.0 work with A+P.
The test results consist of the application compatibility test, UPnP
1.0 extensions and UPnP 1.0 friendly port allocation for A+P, port
usage and BitTorrent behaviors with A+P.
This memo focuses on the IPv6 flavor of A+P.
Status of this Memo
This Internet-Draft is submitted in full conformance with the
provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF). Note that other groups may also distribute
working documents as Internet-Drafts. The list of current Internet-
Drafts is at http://datatracker.ietf.org/drafts/current/.
Deng, et al. Expires May 1, 2012 [Page 1]
Internet-Draft A+P implementation October 2011
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 to cite them other than as "work in progress."
This Internet-Draft will expire on May 1, 2012.
Copyright Notice
Copyright (c) 2011 IETF Trust and the persons identified as the
document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal
Provisions Relating to IETF Documents
(http://trustee.ietf.org/license-info) in effect on the date of
publication of this document. Please review these documents
carefully, as they describe your rights and restrictions with respect
to this document. Code Components extracted from this document must
include Simplified BSD License text as described in Section 4.e of
the Trust Legal Provisions and are provided without warranty as
described in the Simplified BSD License.
Table of Contents
1. Introduction....................................................3
2. Terminology.....................................................3
3. Implementation environment......................................4
3.1. Environment Overview.......................................4
3.2. Implementation and Configuration of A+P....................5
3.2.1. IPv4-Embedded IPv6 Address Format For A+P CPE.........6
3.2.2. DHCPv6 Configurations.................................6
3.2.3. Avoiding Fragmentation................................6
3.3. Implementing non-continuous Port Sets for A+P..............7
3.3.1. Non-continuousPort Sets allocation mechanism..........7
3.3.2. IPv4-Embedded IPv6 Address Format for Non-
continuousPort Sets A+P CPE.................................10
3.3.3. Customize a non-continuousPorts Set A+P NAT on Linux.11
4. Application Tests and Experiments in A+P Environment...........12
4.1. A+P Impacts on Applications...............................12
4.2. UPnP extension experiment.................................13
4.2.1. UPnP 1.0 extension...................................13
4.2.2. Evaluation of non-continuous port allocation taking
UPnP 1.0 friendliness into account..........................15
4.3. Port Usage of Applications................................17
Deng, et al. Expires May 1, 2012 [Page 2]
Internet-Draft A+P implementation October 2011
4.4. BitTorrent Behaviour in A+P...............................19
5. Security Considerations........................................19
6. IANA Considerations............................................20
7. Conclusion.....................................................20
8. References.....................................................20
8.1. Normative References......................................20
8.2. Informative References....................................20
9. Acknowledgments................................................21
1. Introduction
A+P [RFC6346] is a technique to share IPv4 addresses during the IPv6
transition period without requiring a NAT function in the provider's
network. The main idea of A+P is treating some bits from the port
number in the TCP/UDP header as additional end point identifiers to
extend the address field, thereby leaving a range of ports available
to applications. This feature facilitates migration of networks to
IPv6-only while offering the IPv4 connectivity services to customers,
because the IPv4 address and the significant bits from the port range
can be encoded in an IPv6 address and therefore transporting IPv4
traffic over IPv6 network by stateless IPv6 routing.
We have implemented A+P in a residential ADSL access network, where
IPv6-only access network is provided over PPPoE. In this document, we
describe the implementation environment including A+P IPv6 prefix
format and network elements configurations, and results of
application tests as well. The document focuses on the implementation
of the SMAP function specified in [RFC6346]:
o Implement DHCPv6 options to retrieve an IPv4-embedded IPv6 address
and a port range.
o Support of those DHCPv6 options in both the DHCPv6 server side and
the DHCPv6 client side.
o Support of those DHCPv6 options in both the DHCPv6 server side and
the DHCPv6 client side.
For extensive application tests results in A+P environment, please
refer to [draft-boucadair-behave-bittorrent-portrange-02] and [draft-
boucadair-port-range-01].
2. Terminology
This document makes use of the following terms:
o PRR: Port Range Router
Deng, et al. Expires May 1, 2012 [Page 3]
Internet-Draft A+P implementation October 2011
o A+P CPE: A+P aware Customer Premise Equipment
3. Implementation environment
3.1. Environment Overview
public
addresses +----------+
realm | PRR |
| |
=== +----------+
IPv4 ^ ^ ^
| | |
| v v
| +--------------+
| | PPPoE/DHCPv6 |
over | | Server |
| +--------------+
| === ^ ^
| IPv6 ^ | |
| over | | |
IPv6 | PPPoE | | |
V v | |
=== === v v
^ +----------+
| | A+P |
| | CPE |
| +----------+
Private | ^ ^
RFC1918 | | |
realm | v v
| +----------+
| | Host |
| | |
V +----------+
Figure 1 : Implementation Environment
We had developed both A+P home gate way function and Port Range
Router (PRR) function on Linux platform and ported the home gate way
function to a Linksys wrt 54G CPE, on which an openwrt 2.6.32 (based
on Linux kernel) is running.
Deng, et al. Expires May 1, 2012 [Page 4]
Internet-Draft A+P implementation October 2011
Figure 2 shows the Parameters of A+P CPE. IPv6 is provisioning over
PPPoE to CPE while DHCPv6 server offers IPv6 prefix and A+P
parameters by extended options defined in [draft-boucadair-dhcpv6-
shared-address-option].
+--------+------------+-------+-----+------------+-----------+------+
| Model | CPU Speed | Flash | RAM | Wireless | Wireless | Wired|
| | (MHz) | (MB) | (MB)| NIC | Standard | Ports|
+--------+---------- -+-------+-----+------------+-----------+------+
| Linksys| 200 | 8 | 32 | Broadcom | 11g | 5 |
| WRT54GS| | | |(integrated)| | |
+--------+------------+-------+-----+------------+-----------+------+
Figure 2 :Parameters of A+P CPE
3.2. Implementation and Configuration of A+P
A+P CPE, using Netfilter framework, the IPv4 port restricted NAT
operation performed by CPE has been implemented by simply rules
through iptables tool on Linux. After the port restricted NAT
operation, the IPv4 packets are sent to a TUN interface which is
described as a virtual network interface in Linux. Using the IPv4-
Embedded IPv6 address format defined in section 3.2.1, an IPv4-in-
IPv6 encapsulation/decapsulation is performed by the TUN interface
handler.
PRR, located in the interconnection point of the IPv6 network and
IPv4 network, has been implemented with two main functions: 1) IPv4-
in-IPv6 encapsulation/decapsulation; Like CPE, TUN driver is also
used in PRR to achieve function IPv4-in-IPv6
encapsulation/decapsulation. 2) Destination port based routing
function, which is responsible for routing the IPv4 traffic
originated from the IPv4 Internet to the Port Range restricted A+P
CPE. Destination port based routing is implemented by generating IPv6
destination address, pre-assigned from IPv4 address and port range to
each CPE, according to IPv4-Embedded IPv6 address format defined in
Section 3.2.1.
Deng, et al. Expires May 1, 2012 [Page 5]
Internet-Draft A+P implementation October 2011
3.2.1. IPv4-Embedded IPv6 Address Format For A+P CPE
|31bits|1bit| 32bits|8 bits|16bits|4bits|1bit|1bit|1bit|1bit|32 bits|
+------+----+-------+------+------+-----+----+----+----+----+-------+
|A+P |flag|Public | EUI64| port |Port |flag|flag|flag|flag|Public |
|Prefix| 0 |IPv4 | | Range|Range| 1 | 2 | 3 | 4 |IPv4 |
| | |Address| | |Size | | | | |Address|
+------+----+-------+------+------+-----+----+----+----+----+-------+
Figure 3 :IPv4-Embedded IPv6 address format
flag0: Is this address used by CPE or PRR?
flag1: Is address shared?
flag2: Is length of invariable present?
flag3: Is port range identifying sub network?
flag4: Reserved?
To facilitate test and experiment on A+P solution, recently, we are
considering release this A+P implementation under open source
license. For more implementation details, please refer to
[Implementing A+P]
3.2.2. DHCPv6 Configurations
DHCPv6 options defined in [draft-boucadair-dhcpv6-shared-address-
option] have been implemented. These options allow to configure a
shared address together with a port range using DHCPv6.
3.2.3. Avoiding Fragmentation
Normally the TCP protocol stack will employ Maximum Segment Size
(MSS) negotiation and/or Path Maximum Transmission Unit Discovery
(PMTUD) to determine
Deng, et al. Expires May 1, 2012 [Page 6]
Internet-Draft A+P implementation October 2011
the maximum packet size, and then try to send as large as possible
datagram to achieve better throughput. However the IPv4-in-IPv6
encapsulation and the PPPoE header is very likely to cause a larger
packet that exceeds the maximum MTU of the wire, and result in
undesired fragmentation processing and decrease transmission
efficiency.
A simple solution is to enable iptables on A+P CPE to modify the MSS
value of TCP session, using the command like "iptables -t mangle -A
FORWARD -p tcp --tcp-flags SYN,RST SYN -j TCPMSS --set-mss
DESIRED_MSS_VALUE". Here the DESIRED_MSS_VALUE is taken into account
of common size of IPv4 header without options, common size of TCP
header and size of basic IPv6 header and PPPoE header as well.
3.3. Implementing non-continuous Port Sets for A+P
3.3.1. Non-continuousPort Sets allocation mechanism
As described in [I-D.ietf-intarea-shared-addressing-issues], a bulk
of incoming ports can be reserved as a centralized resource shared by
all subscribers using a given restricted IPv4 address. In order to
distribute incoming ports as non-continuous as possible among
subscribers sharing the same restricted IPv4 address, other than
allocating a continuous range of ports to per subscriber, a solution
to distribute bulks of non-continuous ports among subscribers, which
also takes port randomization of CPE NAT into account, because port
randomization is one protection among others against blind attacks,
is elaborated thereby.
Note that the non-continuous port sets allocation mechanism
implemented here is just one possible solution among others to offer
non-continuous port provisioning. The implementation itself is to
address two targets: 1) proving feasibility of non-continuous ports
with A+P approach; 2) Evaluate efforts and investigate possibility of
making UPnP 1.0 applications still work with this approach, with
which experiments results will be described in Section 4.2.2.
On every restricted IPv4 address, according to port set size N,
log2(N)bits are randomly chose as subscribers identification bits(s
bit) among 1st and 16th bits. Take a sharing ration 1:32 for
example, Figure 4 shows an example of 5bits (2nd, 5th, 7th, 9th,
11th) being chose as s bit.
|1st |2nd |3rd |4th |5th |6th |7th | 8th|
Deng, et al. Expires May 1, 2012 [Page 7]
Internet-Draft A+P implementation October 2011
+----+----+----+----+----+----+----+----+
| 0 | s | 0 | 0 | s | 0 | s | 0 |
+----+----+----+----+----+----+----+----+
|9th |10th|11th|12th|13th|14th|15th|16th|
+----+----+----+----+----+----+----+----+
| s | 0 | s | 0 | 0 | 0 | 0 | 0 |
+----+----+----+----+----+----+----+----+
Figure 4 : An s bit selection example (on a sharing ration 1:32
address).
Subscriber ID pattern is then formed by setting all the s bits to 1
and other trivial bits to 0. Figure 5 illustrates an example of
subscriber ID pattern which follows the s bit selection of figure 4.
Note that the subscriber ID pattern can be different, ensured by the
random s bit selection, per restricted IP address no matter whether
the sharing ratio varies.
|1st |2nd |3rd |4th |5th |6th |7th | 8th|
+----+----+----+----+----+----+----+----+
| 0 | 1 | 0 | 0 | 1 | 0 | 1 | 0 |
+----+----+----+----+----+----+----+----+
|9th |10th|11th|12th|13th|14th|15th|16th|
+----+----+----+----+----+----+----+----+
| 1 | 0 | 1 | 0 | 0 | 0 | 0 | 0 |
+----+----+----+----+----+----+----+----+
Figure 5 : A subscriber ID pattern example (on a sharing ration 1:32
address).
Deng, et al. Expires May 1, 2012 [Page 8]
Internet-Draft A+P implementation October 2011
Subscribers ID value is then assigned by setting subscriber ID
pattern bits (s bits shown in figure 4) to a unique customer value
and setting other trivial bits to 1. An example of subscriber ID
value, having a subscriber ID pattern shown in the figure 5 and a
customer value 0, is shown in the figure 6.
|1st |2nd |3rd |4th |5th |6th |7th | 8th|
+----+----+----+----+----+----+----+----+
| 1 | 0 | 1 | 1 | 0 | 1 | 0 | 1 |
+----+----+----+----+----+----+----+----+
|9th |10th|11th|12th|13th|14th|15th|16th|
+----+----+----+----+----+----+----+----+
| 0 | 1 | 0 | 1 | 1 | 1 | 1 | 1 |
+----+----+----+----+----+----+----+----+
Figure 6 : A subscriber ID value example (customer value: 0)
Subscriber ID pattern and subscriber ID value together uniquely
defines a restricted port set (Non-contiguous port sets or a
contiguous port range, depends on Subscriber ID pattern and
subscriber ID value) on a restricted IP address.
Pseudo-code shown in the figure 7 describes how to use subscriber ID
pattern and subscriber ID value to implement a random ephemeral port
selection function within the defined restricted port sets on a
customer NAT.
do{
restricted_next_ephemeral = (random()|subscriber_ID_pattern)
& subscriber_ID_value;
Deng, et al. Expires May 1, 2012 [Page 9]
Internet-Draft A+P implementation October 2011
if(five-tuple is unique)
return restricted_next_ephemeral;
}
Figure 7 : Random ephemeral port selection within the restricted port
set
3.3.2. IPv4-Embedded IPv6 Address Format for Non-continuousPort Sets
A+P CPE
|31bits|1bit| 32bits|8bits|16bits |4bits|1bit|1bit|1bit|1bit|32bits|
+------+----+-------+------+------+-----+----+----+----+----+-------+
|A+P |flag|Public | EUI64|SID_ |Reser|flag|flag|flag|flag|Public |
|Prefix| 0 |IPv4 | |Value |-ved | 1 | 2 | 3 | 4 | IPv4 |
| | |Address| | | | | | | |Address|
+------+----+-------+------+------+-----+----+----+----+----+-------+
Figure 8 :IPv4-Embedded IPv6 address format
SID Value: Subscriber_ID_Value, which is unique for per subscriber
sharing a given restricted IPv4 address. and has been allocated to
each subscriber.
flag0: Is this address used by CPE or PRR?
flag1: Is address shared?
flag2: Is length of invariable present?
flag3: Is port range identifying sub network?
flag4: Reserved?
PRR maintains a mapping table, which consists of restricted IPv4
address and it's Subscriber ID Pattern. To form an IPv6 destination
address for incoming packet, PRR could find the right SID Pattern
according to a destination IPv4 address, and then apply a simple
operation shown in the figure 9.
SID_Value = Destination_Port | (~SID_Pattern);
Deng, et al. Expires May 1, 2012 [Page 10]
Internet-Draft A+P implementation October 2011
Figure 9 :PRR calculates SID Value
3.3.3. Customize a non-continuousPorts Set A+P NAT on Linux
With a linux kernel 2.6.32.36, only one line of linux kernel code is
changed, as shown in the figure 10, and the same IPtables command
line interface is used with the only one change of semantic that the
original staring of port range becomes SID_Value and the ending port
of a port range becomes SID_Pattern. The command line with iptables
to configure a non-continuousPorts Set A+P is illustrated in the
figure 11.
bool nf_nat_proto_unique_tuple(...)
...
//The Original code:
//*portptr = htons(min + off % range_size);
// was changed to:
*portptr = htons((ntohs(off) | min ) & max );
...
Figure 10:Function of finding a unique 5-tuple for a non-
continuousport sets A+P NAT
iptables -t nat -A POSTROUTING -o eth0 -p tcp -j SNAT --to-source
a.b.c.d: SID_Value-SID_Pattern --random
iptables -t nat -A POSTROUTING -o eth0 -p udp -j SNAT --to-source
a.b.c.d: SID_Value-SID_Pattern --random
Figure 11: IPtables commands for a non-continuousports set A+P NAT
Deng, et al. Expires May 1, 2012 [Page 11]
Internet-Draft A+P implementation October 2011
4. Application Tests and Experiments in A+P Environment
A set of well-known applications have been tested in this IPv6 flavor
of A+P environment to access A+P impacts on them. The test results
show that IPv6 flavor of A+P has the same impacts on applications as
IPv4 flavor A+P does [draft-boucadair-port-range-01]. Web browsing
(IE and Firefox), Email (Outlook), Instant message(MSN),Skype, Google
Earth work normally with A+P. For more details, please refer to
[draft-boucadair-port-range-01].
4.1. A+P Impacts on Applications
Deng, et al. Expires May 1, 2012 [Page 12]
Internet-Draft A+P implementation October 2011
+------------------+--------------------------------------+
| Application | A+P impacts |
+------------------+--------------------------------------+
| IE | None |
+------------------+--------------------------------------+
| Firefox | None |
+------------------+--------------------------------------+
| FTP(Passive mode)| None |
+------------------+--------------------------------------+
| FTP(Active mode) | require opening port forwarding |
| | |
+------------------+--------------------------------------+
| Skype | None |
+------------------+--------------------------------------+
| Outlook | None |
+------------------+--------------------------------------+
| Google Earth | None |
+------------------+--------------------------------------+
| BitComet | UPnP extensions may be required, when|
| | listening port is out of A+P range; |
| | other minor effects(see Section 4.4) |
+------------------+--------------------------------------+
| uTorrent | UPnP extensions may be required, when|
| | listening port is out of A+P range; |
| | other minor effects(see Section 4.4) |
+------------------+--------------------------------------+
| Live Messenger | None |
+------------------+--------------------------------------+
Figure 12: A+P impacts on applications
For P2P (Peer-to-Peer) applications, when some of them listening on
specific port to expect inbounding connection, it is likely to fail
due to the listening port is out of A+P port range. Some UPnP
extensions may be required to make P2P applications work properly
with A+P. Other minor effects of A+P are discussed in Section 4.4.
4.2. UPnP extension experiment
4.2.1. UPnP 1.0 extension
To make P2P application work properly with port restricted NAT , we
have designed extensions including new variables, new error codes as
well as new actions to UPnP 1.0, and have them implemented with
Deng, et al. Expires May 1, 2012 [Page 13]
Internet-Draft A+P implementation October 2011
[Emule], [open source UPnP SDK 1.0.4 for Linux] and [Linux UPnP IGD
0.92].
In figure 5, a new error code is proposed for the existing
"AddPortMapping" action to explicitly indicate the situation that the
requested external port is out of range.
+----------+-----------------------+-----------------------------+
| ErrorCode| errorDescription | Description |
+----------+-----------------------+-----------------------------+
| 728 |ExternalPortOutOfRange | The external port is out |
| | | of the port range assigned |
| | | to this external interface |
+----------+-----------------------+-----------------------------+
Figure 13:New ErrorCode for "AddPortMapping" action
New state variables have been introduced to reflect the valid port
range. The definitions of these state variables are shown in figure
6.
+-------------+-------+------+----------+---------+-------+
|Variable |Req. or| Data | Allowed | Default | Eng. |
| Name | Opt.| Type | Value | Value | Units |
+-------------+-------+------+----------+---------+-------+
|PortRangeLow | O | ui2 | >=0 | 0 | N/A |
+-------------+-------+------+----------+---------+-------+
|PortRangeHigh| O | ui2 | <=65535 | 65535 | N/A |
+-------------+-------+------+----------+---------+-------+
Figure 14: New state variables for port range
Correspondingly, new actions, GetPortRangeLow and GetPortRangeHigh,
defined to retrieve port range information are illustrated in figure
Deng, et al. Expires May 1, 2012 [Page 14]
Internet-Draft A+P implementation October 2011
7. An IP address should be provided as argument to invoke the new
actions, for the port range is associated with a specific IP address.
+----------------+-----------------------+----+--------------------+
| Action Name | Argument |Dir.| Related |
| | | | StateVariable |
+----------------+-----------------------+----+--------------------+
|GetPortRangeLow | NewExternal IPAddress | IN | ExternalIPAddress |
| +-----------------------+----+--------------------+
| | NewPortRange Low | OUT| PortRangeLow |
+----------------+-----------------------+----+--------------------+
|GetPortRangeHigh| NewExternal IPAddress | IN | ExternalIPAddress |
| +-----------------------+----+--------------------+
| | NewPortRange High | OUT| PortRangeHigh |
+----------------+-----------------------+----+--------------------+
Figure 15: New actions for port range
Please refer to [UPnP Extension] for more details of UPnP extension
experiment in A+P.
4.2.2. Evaluation of non-continuous port allocation taking UPnP 1.0
friendliness into account
UPnP 1.0 applications behaviors of asking for an external port
Deng, et al. Expires May 1, 2012 [Page 15]
Internet-Draft A+P implementation October 2011
+-------------------+----------------------------------------------+
| Application | Behaviors |
| | |
+-------------------+----------------------------------------------+
| Microtorrent v2.2 | call GetSpecificPortMapping by incremental by|
| | 1 each time, |
| (also known as | until find an external port available, and |
| uTorrent) | then call AddPortMapping,or return error |
| | after five failures |
+-------------------+----------------------------------------------+
| Emule v0.50a | call AddPortMapping, after finding the |
| | external port not available return error |
| | |
+-------------------+----------------------------------------------+
| Azureus v4.6.0.2 | call AddPortMapping, after finding the |
| | external port not available, try the same |
| | port 5 more times by call AddPortMapping, |
| | then return error |
|-------------------+----------------------------------------------+
| Shareazav2.2.5.7 | call GetSpecificPortMapping, after finding |
| | the external port not available, return error|
| | without issuing AddPortMapping |
+-------------------+----------------------------------------------+
As per above typical behaviours of UPnP 1.0 applications' asking for
external port, to instance a port allocation making port sets
interval less than 5 so that some UPnP applications would probably
succeed in 5 times retrying, A non-continuous port allocation example
would be: Subscriber ID Pattern 0x02 and 2 customers share the same
IP address, where 4 times retrying would succeed.
Deng, et al. Expires May 1, 2012 [Page 16]
Internet-Draft A+P implementation October 2011
+-------------------+----------------------------------------------+
| Application | Does it work with UPnP 1.0 friendly port |
| | provisioning method? |
| | |
| | |
+-------------------+----------------------------------------------+
| Microtorrent v2.2 | Yes |
| | |
| (also known as | |
| uTorrent) | |
| | |
+-------------------+----------------------------------------------+
| Emule v0.50a | 1/5 chance of working |
| | |
| | |
+-------------------+----------------------------------------------+
| Azureus v4.6.0.2 | 1/5 chance of working |
| | |
| | |
| | |
|-------------------+----------------------------------------------+
| Shareazav2.2.5.7 | 1/5 chance of working |
| | |
| | |
+-------------------+----------------------------------------------+
The test results show that even with restricted UPnP 1.0 friendly
port allocation, where however the sharing ratio 2 may not be
applicable to most of use cases ,only one application among others
would be granted working, while with others only the chances of
success have been increased.
IGD:1 is known to be broken in shared address environment [RFC6269];
IGD:2 mitigates the issues encountered in IGD:1. The efforts,
documented above, to solve the encountered issues in IGD:1 thereby
aiming only at evaluate the amount of requirement modifications and
assess the validity of the approach.
4.3. Port Usage of Applications
Port consumptions of applications not only impact the deployment
factor (i.e., port range size) for A+P solution but also play an
important role in determining the port limitation of per customer on
AFTR for Dual-Stack Lite.
Deng, et al. Expires May 1, 2012 [Page 17]
Internet-Draft A+P implementation October 2011
Therefore we have also developed and deployed a Service Probe in our
IPv6 network, which use IPv6 TCP socket to ask A+P CPE for NAT
session usage, and store A+P NAT statistics in a Mysql database for
further analysis of application behaviours in terms of port and
session consumptions.
In figure 8, the maximum port usage of each application is the peak
number of port consumption per second during the whole communication
process. The duration time represents the total time from the first
NAT binding entry being established to the last one being destroyed.
+-----------+--------------------------+--------------+----------+
|Application| Test case | Maximum | Duration |
| | | port usage | (seconds)|
+-----------+--------------------------+--------------+----------+
| | browsing a news website | 20-25 | 200 |
| IE +--------------------------+--------------+----------+
| | browsing a video website | 40-50 | 337 |
+-----------+--- ----------------------+--------------+----------+
| | browsing a news website | 25-30 | 240 |
| Firefox +--------------------------+--------------+----------+
| | browsing a video website | 80-90 | 230 |
+-----------+--------------------------+--------------+----------+
| | browsing a news website | 50-60 | 340 |
| Chrome +--------------------------+--------------+----------+
| | browsing a video website | 80-90 | 360 |
+-----------+--------------------------+--------------+----------+
| Android | browsing a news website | 40-50 | 300 |
| Chrome +--------------------------+--------------+----------+
| | browsing a video website | under 10 | 160 |
+-----------+--------------------------+--------------+----------+
| Google | locating a place | 30-35 | 240 |
| Earth | | | |
+-----------+--------------------------+--------------+----------+
| Android | | | |
| Google | locating a place | 10-15 | 240 |
| Earth | | | |
+-----------+--------------------------+--------------+----------+
| Skype | make a call | under 10 | N/A |
+-----------+--------------------------+--------------+----------+
| BitTorrent| downloading a file | 200 | N/A |
+-----------+--------------------------+--------------+----------+
Figure 16: Port usage of applications
Deng, et al. Expires May 1, 2012 [Page 18]
Internet-Draft A+P implementation October 2011
4.4. BitTorrent Behaviour in A+P
[draft-boucadair-behave-bittorrent-portrange] provides an exhaustive
testing report about the behaviour of BiTtorrent in an A+P
architecture. [draft-boucadair-behave-bittorrent-portrange] describes
the main behavior of BitTorrent service in an IP shared address
environment. Particularly, the tests have been carried out on a
testbed implementing [ID.boucadair-port-range] solution. The results
are, however, valid for all IP shared address based solutions.
Two limitations were experienced. The first limitation occurs when
two clients sharing the same IP address want to simultaneously
retrieve the SAME file located in a SINGLE remote peer. This
limitation is due to the default BitTorrent configuration on the
remote peer which does not permit sending the same file to multiple
ports of the same IP address. This limitation is mitigated by the
fact that clients sharing the same IP address can exchange portions
with each other, provided the clients can find each other through a
common tracker, DHT, or Peer Exchange. Even if they can not, we
observed that the remote peer would begin serving portions of the
file automatically as soon as the other client (sharing the same IP
address) finished downloading. This limitation is eliminated if the
remote peer is configured with bt.allow_same_ip == TRUE.
The second limitation occurs when a client tries to download a file
located on several seeders, when those seeders share the same IP
address. This is because the clients are enforcing bt.allow_same_ip
parameter to FALSE. The client will only be able to connect to one
sender, among those having the same IP address, to download the file
(note that the client can retrieve the file from other seeders having
distinct IP addresses). This limitation is eliminated if the local
client is configured with bt.allow_same_ip == TRUE, which is somewhat
likely as those clients will directly experience better throughput by
changing their own configuration.
Mutual file sharing between hosts having the same IP address has been
checked. Indeed, machines having the same IP address can share
files with no alteration compared to current IP architectures.
5. Security Considerations
TBD
Deng, et al. Expires May 1, 2012 [Page 19]
Internet-Draft A+P implementation October 2011
6. IANA Considerations
This document includes no request to IANA.
7. Conclusion
Despite A+P introduces some impacts on existence applications, issues
of P2P applications due to the port restricted NAT have been resolved
by UPnP extension experiment in our test bed, and other issues are
shared by other IP address sharing solutions. Therefore, from our
work, it has been proved that deploying both port range and non-
continuous port sets A+P in the Service Provider's IPv6 network
during IPv6 transition period is feasible.
8. References
8.1. Normative References
[Implementing A+P]
Xiaoyu ZHAO.,"Implementing Public IPv4 Sharing in IPv6
Environment", ICCGI 2010
[UPnP Extension]
Xiaoyu ZHAO., "UPnP Extensions for Public IPv4 Sharing in
IPv6 Environment", ICNS 2010
8.2. Informative References
[RFC6346]
R. Bush., " The Address plus Port (A+P) Approach to the
IPv4 Address Shortage", August,2011.
[draft-boucadair-dhcpv6-shared-address-option]
M. Boucadair., "Dynamic Host Configuration Protocol (DHCPv6)
Options for Shared IP Addresses Solutions", draft-
boucadair-dhcpv6-shared-address-option-01 (work in
progress), December 21, 2009
[draft-boucadair-port-range-01]
"IPv4 Connectivity Access in the Context of IPv4 Address
Exhaustion", draft-boucadair-port-range-01(work in
progress), January 30, 2009
Deng, et al. Expires May 1, 2012 [Page 20]
Internet-Draft A+P implementation October 2011
[Emule]
http://www.emule-project.net/. [Accessed October 26, 2009]
[UPnP SDK 1.0.4 for Linux]
http://upnp.sourceforge.net/. [Accessed October 26, 2009].
[Linux UPnP IGD 0.92].
http://linuxigd.sourceforge.net/. [Accessed October 26,
2009].
[draft-boucadair-behave-bittorrent-portrange]
M. Boucadair.,"Behaviour of BitTorrent service in an IP
Shared Address Environment", draft-boucadair-behave-
bittorrent-portrange-02.txt
9. Acknowledgments
The experiments and tests described in this document have been
explored, developed and implemented with help from Zhao Xiaoyu, Eric
Burgey and JACQUENET Christian.
Appreciation to Randy Bush's intitial idea of documenting these
experience results, for share the knowledge of what we have learnt
with the community.
Thanks to Jan Zorz for comments.
Deng, et al. Expires May 1, 2012 [Page 21]
Internet-Draft A+P implementation October 2011
Authors' Addresses
Xiaohong Deng
France Telecom
Hai dian district, 100190, Beijing,
China
Email: xiaohong.deng@orange-ftgroup.com
Mohamed BOUCADAIR
France Telecom
Rennes,35000 France
Email: mohamed.boucadair@orange-ftgroup.com
Yiu L. Lee
Comcast
One Comcast Center
Philadelphia, PA 19103
U.S.A.
Lan Wang
France Telecom
Hai dian district, 100190, Beijing, China
Email: lan.wang@orange-ftgroup.com
Tao Zheng
France Telecom
Hai dian district, 100190, Beijing, China
Email: tao.zheng@orange-ftgroup.com
Xiaohong Huang
Beijing University of Post and Telecommunication
Email: huangxh@bupt.edu.cn
Qin Zhao
Beijing University of Post and Telecommunication
Email: zhaoqin.bupt@gmail.com
Yan MA
Beijing University of Post and Telecommunication
Email: mayan@bupt.edu.cn
Deng, et al. Expires May 1, 2012 [Page 22]
Internet-Draft A+P implementation October 2011
Deng, et al. Expires May 1, 2012 [Page 23]