understand multicasting from the client/host perspective .

To respond to your inquiry from further down and to put it into a decent analogy, you've asked a forum of race car-driving pros to teach you how to corner and power brake. All racing pros of course consider this rudimentary racing knowledge. No offense but I would recommend that you learn the same way we did. Get yourself a copy of TCP/IP Illustrated Volume 1 & 2. If ever there was a "User's Guide" for TCP/IP, those books are it. Once you learn the basics of how it's supposed to work then we can better help you with the specifics of the implementation on Cisco hardware.

J

TCP/IP Illustrated Volumes 1-3 Boxed Set

TCP/IP Illustrated, Volume 1: The Protocols

TCP/IP Illustrated, Volume 2: The Implementation

TCP/IP Illustrated, Volume 3: TCP for Transactions, HTTP, NNTP, and the UNIX=AE Domain Protocols

Sincerely,

Brad Reese Cisco Repair

If you don't know or cannot understand, you don't have to say anything ..there are people can relate and try to help.

I'm a pro > April wrote:

Oh by the way, try this specifi one ...

The multicast MAC part is where I have difficulty to understand well ..

like why this is needed? As the sender, it seems to me it only needs to send to an IP ... I can think this signifies the frame is multicast,

and somehow the switch can relate the multicast MAC to the MAC of a host that joined the group, but not sure this is all it about?

Thanks > April wrote:

Well then I'm afraid I'm going to have to side with the folks who say you need to know a lot more background before we can answer your questions.

Sam

Because there could be dozens or hundreds of hosts that want to receive the data.

Its much more efficient to send a packet to *one* multicast address than to sent it to *multiple* IP addresses.

k let's talk just the last leg here .. based on what Mike said, the multicast IP is converted to the multicast MAC at the last router ...

In non multicast case, a switch will just just based on the IP and find the MAC of a host, and forward a frame to the card that has the MAC.

However, in the multicast case, once the router converted the IP to MAC, what this MAC will be used for? How does a switch know a multicast frame should be forwarded to a specific host, which is a member of a multicast group? Did the IP and MAC, and their affiliation to the multicast group is registered somewhere on the switch, say the CAM table?

Is the purpose of the multicast MAC is > > > ...

All multicast IP addresses map to corresponding mac address but there's not a one-to-one mapping. Rather than me retype it please read the info at this link, it explains how 32 different multicast IP's all map to the same Ethernet address.

BernieM

Thanks Bernie .. I understand this part - 2^5 to 1

My prob is I d> >

Just think logically. You have a hundred workstations, connected to the switch. Router knows, that somebody "subscribed" to certain "multicast stream", and did everything to bring the stream. Then it has to deliver this stream to the client. But switch is not aware, which port "subscribed" to the stream. So, it sends frames to all ports. however, if switch is smart enough, it will watch for an initial conversation between the client(s) and the router, and will remember, that port 3/15 has subscribed to the group

234.111.222.333. In this case it will send the frame directly to that port, and everybody else will not hear anything. It's like having a headphones with your radio... But if your radio has no headphones jack, you can not use this feature... :-)

Good luck,

Mike

OK. Get back to the "4th grade language" :-)

What's a difference between IP address and MAC address. Let's say, the IP address is your "Public Mailing Address", for example, 1234 Main Street, Room 234, Best Town, CA, 95001. People send you a correspondence to that address. But they don't care what color is your mailbox, what row is it located, etc. But mailmen does. He knows that your mailbox is third one on the fourth row, right over the corner.

Now, let's say some Men's magazine sends a preview issue to all male, who live in this house. They just sent a bunch of magazines to your house, and mailmen will place them into the boxes, where male tenants live. That will be "multicast".

But if mailmen is not aware, who lives where, it will place the magazine in each mailbox. He "converts" a "multicast" (Men's magazine) to "broadcast" (what if there is a men in the house...). Publisher's intent was to deliver the issue to male population only, but since mailmen was not aware, he delivered them to every mailbox.

And another situation - your town wants every person in the town to receive an invitation. They send a big box of postcards to a local post office. And mailmen instructed to put the postcard into each and every mailbox. So, he does not care, who lives there - he just delivers a mail.

And now back to "technical language"... Each network device has it's own MAC address. Your switch knows, which MAC address available thriugh which port. So, in case on unicast, it knows, that specific freame should be delivered to the specific port. Also there is a "Layer 2 broadcast" MAC address (often named Ethernet broadcast). Any frame with the bradcast MAC address as the destination, will be delivered to each and every port on the switch. And third "group" of MAC addresses is defined to be a Multicast MAC addresses. You can look at this MAC address as on "conditional broadcast". It will reach the port only if certain condition met (port subscribed to specific Multicast group). Switch does not care what was the IP address of the group, nor what was the IP address of the sender. It knows, that it has to deliver the frame based on the destination MAC address.

Good luck,

Mike

Mike, you are the best!

I think IGMP has to bee used for sending to hosts in a specific multicast group, my teacher told me that ...

I think my remaining question is on the conversion of the last leg multicast IP to MAC, why that is required, and how it is being used (by the switch?)?

I th> >

It's the same reason why a unicast IP address has to be translated to a MAC address when sending to the final destination. Everything on a LAN has to be sent to a MAC address, that's how LANs work.

In the case of unicast we we ARP to find out the MAC address that corresponds to a particular IP address. But for multicast you can't use ARP because each member of the group has a different MAC address. So instead there's a direct translation algorithm, where a piece of the multicast IP address is appended to a standard MAC prefix.

So, at the last leg, the router translate the multicast IP to the multicast MAC (according to Mike), and forward to the switch. The switch then delivers the multicast frames to the multicast MAC. Although this multicast MAC does not exist on the subnet - not relate to any physical NIC, the NIC belongs to the host that joined the multicast group knows it, and picks up the frames that sent to the multicast MAC, is this the picture I should get?

It seems to me the switch should play a more active role here?

Barry Margol> >

April skrev:

Hasn't this just been answered? A switch has no concept of IP addresses it does not know what to do with them or even care about them. The only thing a switch cares about are MAC addresses.

Thats why the router which cares about both IP addresses and MAC addresses has to convert the IP address to a address that the switch can understand.

-SAto

Let me nitpick - the word "conversion" seems slightly inappropriate here. The IP addresses remain the same but the IP packet is encapsulated in a frame with a multicast destination MAC address derived from the multicast destination IP address.

For the avoidance of doubt, YMMV etc etc.

Sam

The switch doesn't really do a whole lot with Multicast other than deal with the multicast address and identify which ports belong to that multicast group. Nothing more. Its reality a simple concept - I believe you're probably over thinking this just a bit.

Think of the layers of your OSI model. If the multicast mac is assigned to your PC (and possibly hundreds of others and the network propagates) then data destined for your mac will be sent to you via the switch. The router encapsulates the MAC frame into an IP Packet with IP addressing and the process goes on and on. Just like a Unicast. In fact, most of the network looks at Multicast as a Unicast, even though its in all reality a type of broadcast. Its a hybrid so to speak.

But since Layer 2 requires a MAC to communicate you couldn't possibly map the multicast IP to your real MAC, because there'd have to be a separate ARP mapping for each host involved. And you can't have an ARP entry that is 1 to many mapping. Otherwise you'd be saying that the real MAC of your PC translates to multiple IP's and vise versa (one IP mapping to multiple MAC). Since that's not the way ARP works, you need sort of a broadcast method involved. Its similar to your network broadcast (i.e. 255.255.255.255) which goes to all systems on the network and all systems choose to listen to that broadcast. With Multicast, its the same concept but only the systems that needs to receive the packet/frame get it.

So end result it you're talking last leg, layer 2. Layer 2 needs a mapping, normally done through ARP. Since ARP doesn't work with 1 to many, you have to do it slightly differently and assign a MAC to IP mapping. This is an RFC standard and is the same from vendor to vendor.

Ultimately, multicast doesn't work without a mechanism to use it though

- your original post mentioned how to use it from the end user perspective. You need an application or something to say what multicast addresses to use to communicate with (whether it be the client [receive] or the server [sending]). Without the application in place, Multicast won't function.

There are entire courses set aside just for Multicast - it can get very detailed, but the basic concept isn't much different than from your basic IP knowledge, just with a little twist on it.

Ryan

April wrote: