Monday, August 18, 2014

CATV distribution and Conditional Multicast

"Looking through the glass always does not help building vision, vision is built when one actually steps out of the door to the open land and experiences it with all the senses that he/she possesses."


My Dear Friends of Transmission Fraternity, 

How are all of you? I hope you have read the last article on Multicast data transmission that I had put forward. The matter that I am going to present in this post is actually a continuation of the same.

1. Recap on the Multicast Traffic:

A multicast traffic is a stream of Data that has a destination address which actually when entering a switching device floods all the ports with the same stream except the port from where it is injected.
The behavior of the traffic is same as broadcast as per the flow is concerned however there are somethings that are different in the multicast traffic.
Multicast traffic can actually be grouped by means of IGMP and can actually be controlled by user based policy as compared to a broadcast traffic that can only be restricted by the means of BSC (Broadcast Storm Control).
Multicast transmission actually exploits the idea of the flooding state of the switch to its advantage and thus optimizes resources in the network to envision a drop and continue architecture of the traffic delivery. This is how more an more traffic can be pumped to more and more locations.
The Network reservation of the resources does not depend on the number of users but actually depends on the number of channels/streams that are do be delivered in the multicast delivery model.

2. Optimizing more resources on the Multicast delivery scheme:

Previously we saw the very basic of the transmission of Multicast traffic in the system, however this kind of plain multicast traffic is very good when it comes to the upper distribution layer. In a normal CATV delivery system there is a upper distribution layer and then there is a lower distribution layer or can also be called as the access layer. Generally the upper distribution layer is full of resources and there is not much of conditional access that is deployed over there. It is however the lower distribution layer that is devoid of the BW and where a conditional multicast can do wonders in the delivery scheme.

Let us see the actual model of the Multicast delivery scheme in a CATV network in ideal scenarios.

Distribution in a CATV network system
As we can see that a CATV entity is not just one entity. A CATV, the person who is actually your cable operator is not the person who is owning the content of the channel. So there will be a particular content owner who is having the access of the channels. This will be a big company. They have a master location from where the content is fed into their main router and this starts a multicast drop and continue across all the locations.

So then there is a content owner, which is actually the owner of the channels then there is this upper layer Distribution, who is called as the content distributor. Big Distribution companies, ISPs or Managed CDNs actually to this at their layer. Then there are SUB-Operators or Local Cable Operators (LCOs) who actually have their own infrastructure in certain areas of the city to feed to their subscribers directly. The LCO model can also have a SUB-LCO entity that actually owns one single point of distribution to the customer, but such kind of deployments and arrangements are really less in number.

2.1 Distribution done by the Content Distributor:

As explained before the content distributor does a plain multicast delivery as the probability of repeat-ability in that layer is very bleak. In that layer all the sub-operator would actually demand an access to all the channel feed at the same time in order to distribute to their subscribers from their end whenever required. In this case the availability of all the channels should be there in the Upper distribution Network all the time. So this is a kind of an unconditional plain multicast that is there in the network with drop and continue arrangement.  The picture remains the same as shown below.

Drop and Continue Distribution and Upper Distribution Layer
As we can see in the picture clearly that there is a proper load sharing of the drop and continue architecture that is present in the upper distribution layer. This ensures proper flow of traffic to all the points with least processing load on the routing entities.
Assuming there are about 200 Channels in the system and each channel has on an average BW of 5 Mb/s this is a total infrastructure of 1Gb/s that is provisioned. There can be more BW provisioned if there are more number of channels. to be very precise how the BW engineering can be done in the Upper distribution layer will be explained later as it will have considerations for SD, HD, 4K and Audio channels.

2.2 Distribution done at the Local Cable Operator Layer:

While the distributor may have the luxury of having a 1G or a 10G BW provisioning to provide all the channels in a plain multicast fashion in the system, same may not be the case for the access network provider or the local cable operator. The local cable operator has the feed of the entire channel in his DSLAM box from which he multicast the content to the access. However, there is a difference over here. The part that is played by the local cable operator in this case is a bit advanced as he does a kind of Conditional Multicast in the system. The Local cable operator uses IGMP to actually do the distribution.
To have theoritical details about IGMP please read the following links

IGMP Wikipedia Link
IGMP V2 (RFC 2236)
IGMP V3 (RFC 3376)

2.2.1 Traffic in the Head end:

The head end of the LCO acts as the IGMP Root of the system and takes in all the channels from the distribution layer of the distributor. Let us see the figure for this.

The Local Cable Operator Distribution topology
As we can see the local cable operator distribution topology actually takes in all the channel feed from the Upper layer distribution, however it does not feed in all the channels at the same time to the access network. 

2.2.2 Repeatability/Probability factor and conditional distribution:

The local cable operator is working in a particular region of the city and so there is quite a possibility that all channels may not be in demand at the same time in the network. Let us take the example of a particular region in the afternoon. At this time the demand may be only of the channels that are showing movie and some news channels. 

Let us say that across this region that we have depicted in our diagram there is a request for CNN, HBO and FTV Only from two regions. 

Users requesting from channel as they are pressing the Channel button

As we can see over here that the users have requested the channels and these requests go from the leaf DSLAM to the Root DSLAM  as channel requests. The Root DSLAM will register the fact that the request has only come for CNN, HBO and FTV in the network and so will only populate the network resources with these three Channels.

Thus effectively the network will not take the entire load of all the channels but will only take the load of 3 channels that have been requested by the network so the load decreases to 15Mb/s from a mammoth 1Gb/s.

The Root also takes care of the fact that from which side or which direction of the network the requests are coming so as to provide the Feed only to that direction. In this case the bottom leaf is not demanding any content and so there is no need to populate the bottom legs with the content. However, there are demands of three channels from left and two from the right. So the Root ensures to send the stream in a way that 15Mb/s flow towards the left part of the ring and 10 Mb/s flow at the right side of the ring. This further optimizes the Bandwidth.

Let us have a look at the picture below for a clear understanding.

On the Left there is consumption of 15Mb/s and on the right there is consumption of 10Mb/s

As and when new request from new channels will come more resources will be filled.

2.2.3 New Leaf Demanding the same Channel:

In our case above the leaf a the bottom was not demanding any channel. However, let us understand that what happens if the user now in the bottom leaf turns on the Set Top Box and starts demanding for a content that is already flowing in the system. Say in this case the user is demanding HBO content that is already flowing in the system. In this case it is seen that the HBO content is already been sent to the left side leaf and so the bottom leaf immediately gets the content from the left side leaf.

New Leaf demanding the content that is already flowing the system
Now, seeing this there may be a very interesting question that what if the leaf would have demanded a content that is available from both the ends, say FTV. Well, in that case the thing that comes into play is the link cost. A link cost is a complex function of Link BW, quality and available resources. The lowest link cost will actually bear the traffic in that case. Of course when there is a failure there will be a different scenario.

Also there may be another question over here in all of your minds. There is a content say Fox news that is being only demanded by the bottom Leaf and that is being served by the Root, now the left side leaf one of the user requests for Fox news then what happens? Will the stream be sourced back from the botoom leaf? The answer is no. The stream when it is already flowing to the bottom leaf , if,  is making a transit from the left DSLAM , then the stream is just made to drop on the left DSLAM and continue to the bottom leaf. However if the path of the stream to the bottom leaf is from the right then depending on the link cost the stream can be sourced from the bottom to the left or directly from the root.

2.3 How does the Cable operator benefit from this? 

Due to this mechanism of the conditional distribution using IGMP there is a heavy optimization that the cable operator does on the network resources that it owns. In this case he may not need to provision a full 1Gb/s bandwidth or he may provision but use the same bandwidth also to provide Internet connectivity and voice. This is called as the triple play which will see in later posts. Assuming that he does provide Broad band also on the same resource then a separate engineering need not be done on providing the TV and the broadband. He can easily put the broadband on the same infrastructure albeit on a lower CoS than the TV. So when there is less space occupied by the TV then there will be more BW to play with in the broadband. Also in the same link infra there can be a Video on Demand service.

The main idea is that as and when people watch the same channels over the same access network more and more BW is optimized for better distribution and thus OPEX decreases considerably. So say, for instance, when a match is on most of the people would be hooked to the match and there will be more resource savings and thus providing better quality of the system and more efficiency in the broadband.

So friends, from this we understand that Conditional Multicast in the access network with IGMP (V2/V3) is very much beneficial for the optimization of the whole network. Especially in the cable operator sector where there is very less infrastructure to play with, this comes as a boon in which the operator can graduate to a small scale ISP. 

We will cover more and more details on this thing called as Multicast. This is very interesting. 

Till then Bye and Take care, 



No comments:

Post a Comment