At 02:46 AM 12/28/00, [EMAIL PROTECTED] wrote:
>an aspect i did not notice to be covered in the response found below is the
>following:
>
>it appears to me that multicast technology incorporates control of traffic
>distribution into the addressing scheme: all
>multicast-rfc-compliant-devices will forward traffic to multiple hosts if
>the destination address warrants it.
At the risk of sounding like a school teacher with a little bun on top of
her head &;-), I'd like to suggest the use of capitalization to start
sentences. It makes the messages easier to read, which is the goal after
all. Thanks.
I agree with you that multicast offers more control. My understanding is
that a router uses a multicast routing protocol, such as DVMRP, MOSPF or
PIM, to learn the paths to multicast receivers. If there are no receivers
on a path, the router does not forward the frame. Also, the Internet Group
Management Protocol (IGMP) allows a host to join a group and inform routers
of the need to receive a particular multicast data stream. IP hosts use
IGMP to report their multicast group memberships to immediately-neighboring
multicast routers. More on this below. First, some basics:
Broadcasts, as we all know, go to FF-FF-FF-FF-FF-FF at the data-link layer
and 255.255.255.255 at the network layer. Although, theoretically, an all
ones IP address means send to all networks, my experience has been that
routers do not forward such a frame.
A directed broadcast goes to network.255. Directed broadcasts can be used
to send to all devices on a particular network. I think for quite a while
Cisco routers have defaulted to "no ip directed-broadcast," because there
are security risks with forwarding directed broadcasts. But you can
reconfigure this. You can even add an access list that specifies which
directed broadcasts to forward.
A multicast goes to a subset of devices. IP multicasting transmits data to
a group of hosts that are identified by a Class-D IP address. In
dotted-decimal notation, host group addresses range from 224.0.0.0 to
239.255.255.255. Network stations recognize an address as being a Class-D
address because the first four bits must be 1110 in binary.
A multicast group is also identified by a MAC-layer multicast address.
Using a MAC-layer multicast address optimizes network performance by
allowing network interface cards (NICs) that are not part of a group to
ignore a data stream that is not for them, as Jeff Kell discussed below.
The Internet Assigned Numbers Authority (IANA) owns a block of MAC-layer
addresses that are used for group multicast addresses. The range of
addresses for Ethernet is 01:00:5E:00:00:00 -- 01:00:5E:7F:FF:FF. When a
station sends a frame to an IP group that is identified by a Class-D
address, the station inserts the low-order 23 bits of the Class-D address
into the low-order 23 bits of the MAC-layer destination address. The top 9
bits of the Class-D address are not used. The top 25 bits of the MAC
address are 01:00:5E followed by a zero, or, in binary:
00000001 00000000 01011110 0
In addition to defining group addresses, the IETF also defines the Internet
Group Management Protocol (IGMP), as mentioned above. When a user starts an
application that requires a host to join a multicast group, the host
transmits a membership-report message to inform routers on the segment that
traffic for the group should be multicast to the host's segment.
In addition to allowing hosts to join groups, IGMP specifies that a
multicast router sends an IGMP query out every interface at regular
intervals to see if any hosts belong to the group. A host responds by
sending an IGMP membership-report message for each group in which it is
still a member, (based on the applications running on the host).
RFC 2236 defines IGMPv2. The main feature of IGMPv2 is the ability for a
router to more quickly learn that the last host has left a group, which is
important for high-bandwidth multicast groups and subnets with highly
volatile group membership. Paul Werner wrote some good responses on this
topic a while back. Search the archives.
In addition to determining which local network segments should receive
traffic for particular multicast groups, a router must also learn how to
route multicast traffic across an internetwork. Multicast routing protocols
provide this function. These include the Distance-Vector Multicast Routing
Protocol (DVMRP), which I think is obsolescent now, MOSPF (which may be
heading that way?), and PIM.
MOSPF, defined in RFC 1584, complements OSPF's capability to develop a
link-state database that describes a network topology. MOSPF supplements
the database with an additional type of link-state record for group
memberships.
Protocol-Independent Multicast (PIM) is another option for multicast
routing. Like MOSPF, PIM works in tandem with IGMP, and it also works with
a unicast routing protocol, such as OSPF, RIP, Enhanced IGRP, and so on.
PIM has two modes -- dense mode and sparse mode. The adjectives "dense" and
"sparse" refer to the density of group members.
Many books cover PIM. See:
"Top-Down Network Design," Chapter 12, by me &;-)
"Designing Routing and Switching Architectures," by Howard Berkowitz
"Developing IP Multicast Networks" by Beau Williamson
OK, that's enough for now!
Priscilla
>it also appears to me that directed broadcast is dependent upon the
>configuration of each device in the potential path of a given wayward ip
>packet: if a device is configured to forward directed broadcast, the packet
>ambles on. if not, it expires in the fashion of most packets, within the
>confines of the circuitry of the device that last received the packet.
>
>Jeff Kell <[EMAIL PROTECTED]>@groupstudy.com on 12/28/2000 01:23:00 AM
>
>Hunt wrote:
> >
> > What is the difference between Directed Boradcast and Mulitcast?
>
>A multicast is an IP address 224.x.x.x/4 (224-239.x.x.x) while a
>directed broadcast can be to any class A-B-C subnet; the directed
>broadcast having all ones in the host part of the subnet address.
>
>A "directed" broadcast is one which "may" be routed. Similarly,
>multicasts "may" be routed, and in the absence of any other parameters
>it will be flooded across all router interfaces, while a directed
>broadcast goes to a specific interface.
>
>At the layer 2 level, broadcasts use the all ones MAC address of
>FF-FF-FF-FF-FF-FF. Multicasts have the second hex digit "odd" so that
>the "little-endian" nature of the ethernet results in the low-order bit
>of the second hex digit being transmitted first. The NIC hardware
>will detect this and signal an interrupt in anticipation of a
>broadcast. Smarter NICs will continue to receive and match the source
>MAC (if not a broadcast) with designated multicast addresses it has been
>told are "interesting".
>
>There is also an interplay between multicast MACs and IPs (left as an
>exercise for the reader). For the first three bytes of the MAC, the
>vendor ID (in the first three bytes) is preserved but the low order bit
>of the first byte is set, making it odd (the second hex digit). For
>example, an HP manufactured NIC might have an 08-00-09 prefix; but
>HP specific multicasts will have an 09-00-09 prefix.
>
>Jeff Kell <[EMAIL PROTECTED]>
________________________
Priscilla Oppenheimer
http://www.priscilla.com
_________________________________
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
