Enterasys LAN Switching
Deploying IP Multicast Switching
Method of assuring globally unique MAC address mappings in
an IP multicast environment.
John Roese
Systems Implementation Group
Abstract:
With the advent of more IP multicast based applications
coupled with the deployment of larger switched LANs, the
possibility of non-unique MAC address mappings in an IP
multicast environment is greatly increased. The presence of
non-unique MAC multicast addresses prevents IP multicast
capable switches from building independent MAC layer multicast
delivery trees for each IP multicast group thus causing some
overlapping delivery of unnecessary IP multicast packets. In
order to maximize the efficiency of the MAC layer multicast
features in layer two switching systems, deployment of layer 3
IP multicast groups must be based on their ultimate mapping to
a unique layer two MAC address. This is achieved by following
relatively simple address selection rules documented in this
paper
Overview:
The establishment of address selection rules for IP multicast
is critical if deploying this technology over large scale
switched LANs. By choosing the IP Multicast group addresses
based on the rules defined here, each multicast group will map
to a globally unique MAC layer multicast address which allows
the layer two switched LAN to handle each group as an
independent entity. This paper first describes the process of
mapping IP multicast class D addresses to MAC layer
multicasts. With this description, the issue of non-unique MAC
addresses is identified. Once identified, a set of rules will
be defined to provide guidance for the selection and
deployment of IP multicast addresses over a switched LAN.
Part 1: IP Multicast Addressing Process
IP multicast provides a method of using a unique class D
address to transport data to multiple destination stations
using a single IP packet. Since the destination address is an
IP class D, the scope of the delivery can be specified at the
end node by its willingness to accept that particular class D
address in its IP stack. Additionally since the class D
address range includes IP addresses between 224.0.0.0 and
239.255.255.255, many concurrent multicast groups can exist
over the same IP network yet remain logically independent.
Applications that wish to utilize IP multicast must
establish class D address for their particular multicast
session or application. While some of these class D addresses
are well known (224.0.1.1 is used by NTP Network Time Protocol
specified in RFC1119 for instance), the majority are defined
when the application is activated. In most of these
applications, the user has the option of choosing their
multicast group address as a session is established. The
complete process of establishing IP multicast sessions is
fully documented in RFC 1112 "Host Extensions for IP
Multicasting" and the IGMP Version 2 draft RFC. For the
purposes of this document, only the process used to map a
selected IP multicast class D address to a specific MAC layer
multicast destination address is significant.
Once an application determines the class D IP multicast
address it will utilize, that address must be mapped into a
MAC layer multicast for delivery across any LAN based system.
This process is outlined as follows:
Step 1: Using the Class D address, identify the low order 23
bits of the class D address.
Step2: Map those 23 bits into the low order 23 bits of a
MAC address with the fixed high order 25 bits of the IANAÕs
IEEE multicast addressing space prefixed by 01:00:5E.
An example of this process is shown in figure 1. This
example assumes that the application has selected the IP class
D address 239.1.1.10. Figure 1 demonstrates the mapping of the
low order 23 bits of the class D IP address to the low order
bits of the MAC layer destination.
Figure 1: Mapping between Class D Address and Mac Layer
Multicast
The mapping of 239.1.1.10 to a MAC layer multicast was
executed by placing the low order 23 bits of the class D
address into the low order 23 bits of the reserved MAC layer
multicast address 01:00:5E:xx:xx:xx. Since only 23 bits are
mapped, the 24th significant bit is fixed at 0. The final MAC
address that is utilized by the multicast group 239.1.1.10 is
01:00:5E:01:01:0A.
This process of associating a Mac layer multicast with a
Class D address is relatively simple but does create one
significant issue. Since only 23 of the 28 significant bits
(the first 4 bits of a class D IP address are fixed at 1110)
of the class D address are mapped into the MAC layer address,
there is a definite possibility that more than one class D
address will share a common MAC layer multicast. For example
in figure 2, the address 226.129.1.10 is mapped into its
appropriate MAC address.
Figure 2: Mapping of class D 226.129.1.10 IP address to MAC
layer multicast
Note that the MAC address that the class D IP address
226.129.1.10 maps to is the same address as the MAC address
used by the 239.1.1.10 Class D IP. Both utilize the address
01:00:5E:01:01:0A. While this does not prevent the two class D
addresses from operating on the same LAN, it does prevent any
type of MAC layer based control mechanism from isolating the
traffic of one group from the others end users. This issue is
caused because the two multicast groups have inadvertently
failed to utilized globally unique MAC layer multicast
addresses simply based on their class D address selection.
Part 2: Suggested rules for class D address selection
In switched LANs, technology exists to allow layer two control
of specific IP multicast groups based on their MAC layer
multicast address. Since the MAC layer multicast is derived
from a portion of the Class D address, this technique is an
appropriate method of executing control over each groups
traffic. This mechanism is effected by configurations that
utilize multiple class D IP addresses that share a common
layer two MAC address as shown in the previous section. This
scenario forces the switches to treat all IP groups sharing a
common MAC multicast as one logical distribution group. This
is shown in figure 3.
Figure 3: Behavior of Multicast Enabled Layer 2 Switches
given non-unique MAC layer addressing
This figure shows the result of having both the 239.1.1.10
and the 226.129.1.10 Class D address groups present on the
same switched LAN. Since both groups map to the
01:00:5E:01:01:0A MAC address, the switches at layer two are
unable to distinguish between the two groups and as such the
traffic destined to either group is delivered to members of
both groups. While this does not prevent the applications from
operating, it does result in unnecessary packets being
delivered to devices that do not wish to see them. Since newer
IP multicast applications can be used for such high bandwidth
applications as video conferencing, this lack of isolation
could result in relatively large numbers of packets being seen
by devices who are not members of a particular Class D IP
group.
In order to avoid this situation, it is recommended that
class D IP addresses used on switched LANs always map to
globally unique MAC layer addresses. To accomplish this,
several options for selecting class D IP addresses are
presented below.
Option 1: Globally Unique low order 16 bits
The key to having globally unique MAC addresses mapped from
class D addresses is based on the low order 23 bits. By
assuring that the last 16 bits of any class D address are
unique, each class D address must map to a unique MAC address.
In this scheme, addresses should be used from a pool based on
the last 16 bits (16 bits is chosen arbitrarily based on ease
of use since that number maps to the last two decimal digits
of the class D address). That pool should exclude all
addresses ending with the range 0.0 through 16.255 as many of
those addresses are reserved for specific networks (see
appendix A: Assigned Multicast Numbers). The remaining
addresses range from 17.0 through 255.255. That provides tens
of thousands of possible address groups for use.
In this scheme, the high order 16 bits are not relevant to
address calculation. They can be any value from 224.0 through
239.255 as long as no two complete class D addresses share the
same last two digits. It would be acceptable to use the
addresses 239.1.1.10 and 225.1.1.11 but not 239.1.1.10 and
225.1.1.10. The second set violates the rules of this option
as two IP class D addresses share the same last two decimal
digits.
Option 2: Use of the 239.128.X.X through 239.255.X.X class
D IP range.
Option 1 allows for the use of any first two digits in the
class D addresses as long as no two addresses share the same
last two digits. This method will provide globally unique MAC
addresses but may be difficult to track as the available
addresses cover a wide range of complete addresses. This
option attempts to simplify the range of addresses and
maximize the total number of addresses available.
To achieve this, it is recommended but not required that
the address range from 239.128.0.0 through 239.255.255.255 be
utilized. This choice is based on the assigned numbers from
the IANA (see appendix A) allocation of this range for use as
organizational and site specific scopes. These addresses are
to be used for multicast applications that are not used across
the global Internet. Since most applications will be used
within a corporate Intranet or on a single switched LAN, this
range is most appropriate.
Once the range is decided, IP class D addresses should be
assigned sequentially so that no two complete addresses within
the range are the same. By following this process over 8
million unique class D multicast addresses will be available
with each address mapping to a globally unique MAC layer
multicast address. This maximizes the available addresses and
assures independent treatment of each address by multicast
aware layer 2 switches.
Summary:
IP Multicast is an extremely useful and efficient mechanism
for multi-point distribution of information. At layer three,
IP devices are able to understand and participate in distinct
IP multicast sessions based on the use of unique IP class D
group addresses. Additionally, since a portion of the class D
IP address is used to formulate specific MAC layer multicasts
for delivery of IP multicast traffic, multicast aware layer 2
switching systems are able to control the traffic flow of
these applications. This control is jeopardized however if
incorrect class D IP addresses are used that do not map to
globally unique MAC addresses. To assure that there is a
unique IP class D to MAC layer address mapping, this document
has proposed two potential methods of selecting class D IP
addresses for use in a corporate network. Either option
provides a relatively simple distribution method of addresses
and both result in globally unique MAC and IP addresses for
each multicast group.
|
E-Mail
Related