BTnet - Framestream technology

Hi, one of my clients uses BTNet - Framestream as a server to 400 branch connection, I never really got the grasp of framestream, but as I understand it, it is a Frame Relay system which can carry IP data, perhaps someone can help me understand it better.  I understand the Internet and the fact that you can put an application securely on the net and it could transfer to loads of data to our branches and only bottleneck which is normally the correction of the host (100mbit/sec), so I can get 100mbit/sec / 400 branches = 0.25mbit/sec on a perfect server.

The server connection on Framestream is limited to 2mbit/sec and each branch is also 2mbit/sec.  However sending large files to each of these branches takes forever, as I understand it, you have 2mbit/sec all the time, unconditional.  So, if 10-20 of the branches requested a large file (or same file), the 2mbit/sec from the server is shared between 10-20 connections.  (slow)

So my question...

If I use a Peer2Peer type system, place multiple FTP servers on the clients network (say each area of the country), could I load balance (share out data) or would I be wasting our time?  Is the limit 2mbit/sec max or can I have lots of 2mbit/sec limits.

Who is Participating?
Forgot to mention that, if you do have a star topology (which seems to be the case) each branch can only see hq, and hq can see all branches. While a router could be configured to use it as the pivot to have each branch communicate with any given branch, it would be silly as it will clog all pipes to do this and then just DOUBLE or triple the traffic, instead of reducing it.

This is where my first suggestion comes into play, where I told you that it would be nice to have a branch have a connection with another, and this second one could have a connection to hq. Or both to hq (along with the one to each other) which is better as redundancy is concerned, but quite expensive as well.

So, putting FTP servers all over, while keeping a star topology, is not the best thing to do.
Distributing the content among the branches is a good idea, provided that:

A) Such content is static, and does not need frequent update
B) A client request could be smartly routed to the nearest idle branch

To determine the best location for download would be quite an issue, you'd have to measure the latency to all sites, from all sites, to decide which would be better.

In any case, if all branches have only one 2mbit connection (including hq) then you have a star topology, which in this case is perhaps not recommended for several reasons, let alone the present one (speed), what about redundancy? If a node goes down, it is done for? If hq goes down, all go down?.

Now, if the content you are trying to share among your WAN is not static (let's say, a CRM database) then, regardless of the distribution, the sole synchronizing of all nodes would hog up again the bandwidth.

I'd say you are better off with a mesh configuration:

     |                             |
     |                             |
                                   |------------------------- HQ
     |                             |
     |                             |

branch n ...................

and configure the routers at each branch the due routes and costs. Branch 1 could download from branch 2 without using the backbone, and branch 2 would update directly from HQ. You'd have 200 branches synchronizing with HQ thus reducing the overall frame-relay cloud usage to 50%

Hope this helps.
You can only have one 2mbit/s limit at the HQ. So it would be 2mbits/400 hosts = 5.12kb/s

Obviously it will be faster because not everyone will be requesting things at the same time, like you previously stated.

There are technologys and products that help with this.

Legato Replistor takes two (or more) sites and synchronizes a folder, but only the changes, and is very low bandwidth (you can set rates for different sites). Kind of costly though ($3500 per host).

Another idea would be DFS:
Are you using Active Directory or another network?

You should probably do what alexai was saying, but a little different. Have one HQ (or better yet, two) send out files to 20 sites. Now those 20 send out to 20 more each, and just like that all 400 have the data. And you could do it by region like you said. Replistor can be set up like this fairly easy.

Network Scalability - Handle Complex Environments

Monitor your entire network from a single platform. Free 30 Day Trial Now!

DaveRowlandAuthor Commented:
Ok thanks for that.  However, are you both familar with either BTnet - Framestream or frame-relay 'star' type networks?

I am under the impression that if i did multi-storage/server scenero throughout the network, i will then fall under a BTnet limit of 2mbit/sec maximum (or pay more money) and actually not gain much.

I believe I cannot put a server in the connection at 100mbit as it will be limited to maximum of 2mbit.

Am I right?
Relaying the data would cause the same amount of data to be transfered in the frame relay cloud as if they all took it from one site, because there is one site send and one receiving in both instances, and the data needs to cross the network for this at some point.

Basically the data has to get from point A to point B, and is going to involve the same cost with BTnet either way. The things im trying to explain are to cut down on the time required to access/distribute the data.

Is the entire cloud 2mbit/s, or is each link into the cloud 2mbit/s?
If you can get a larger link, you could speed up the transfers.

Yes, you are right. In ANY scenario, access to the server is limited by the internet connection. So your web server on gigabit is only going to be access at 1.544 mbit/s on a T1 (obviously).

You should distribute the data no matter what, and there are a few ways to do it. If you are managing 400 locations and dont know how to do this you may need to get some outside professional help.

> I understand it, you have 2mbit/sec all the time, unconditional.
Yes, you will always have 2 mbit/s AT EACH branch, thus, in the whole network

>Is the limit 2mbit/sec max or can I have lots of 2mbit/sec limits.
That is pretty much the same, since each and every branch has 2mbit/s, you would never get higher.

>....are you both familar with either BTnet - Framestream or frame-relay 'star' type networks?
Yes, I used to manage one of the largest Frame Relay WAN in my country

>I believe I cannot put a server in the connection at 100mbit as it will be limited to maximum of 2mbit.
The benefit of growing your bandwidth, say from 2048kbps to 1024mbps is that your pipe will be wider. This is, you can send more information in a shorter time. Your last hop to the frame relay cloud (maybe your router, or a PCI WAN adapter in your server) will split, encapsulate and label the TCP/IP packets into frames and send them out through your provider's network at an incredible speed, higher than your entrypoint's bandwidth since it will go through their infrastructure and by the time it arrives to the destination hop, it will reassemble the TCP/IP packet and push it through the pipe at the destination branch, which, if having the same 2mbit as the sending one, will take the same time to receive it, because Frame Relay is designed to be "transparent".

>I believe I cannot put a server in the connection at 100mbit as it will be limited to maximum of 2mbit.
You could, and this would increase your server's pipe width, shortening the time to serve the branches, but you would have to increase your telecom expense, while taking a file distribution approach would have been free, but you know your application better and you would have to put these in the balance to see what is best.

You could also provide some more details about your application without being specific or compromising, like what it is (a file, a database), can it be compressed, how frequently does it need to be synched, and so.
DaveRowlandAuthor Commented:
ok, i understand and I might just be explaining things badly.  I understand that you cannot go above a tcpip connection (like 64kb isdn, i cannot get any bigger then this).  I am a programmer and understand most network structures.

The application we have is currently in use on BTnet- Framestream, the server connection is at 2mbit/sec to 400 branches aronud the UK, it transfers media (avi, jpeg mainly) under our own communication protocol, which we are changing to FTP.  The system works slow but some of that is due to application speed, we have never got above 8 simultenaous connections, it just grinds to a halt.  

1:So, we could use a public network server and place the server on 100mbit  and use ADSL type connection to each branch to keep costs down.
2:We could get a faster/wider pipe at the server end (btnet)
3:We could put slave FTP Servers throughout the country and load balance the network.  So branch 1 could connect to it's local server and branch 2 could connect to another local server, this is my question really.  I need to confirm that I can run this type of configuration on BT-Net or other frame relay servers.  Is the Internet a frame relay? (i never thought it was)

If I understood the "cloud" term correctly, it means 2Mbit/sec is the Legal limit on the entire WAN network, (the total maximum allowed is 2mbit/sec), so what I was trying to ask is,
Is a 2mbit/sec connection (branch 1 to slave server 1) +
       2mbit/sec connection (branch2 to slave server 2) +  
       2mbit/sec connection (branch3 to slave server 3).. = lots per mbit/sec (branches to many slave servers)

Is connections on a framestream network possible?, I understand this is not possible and this is what I am trying to confirm.

Hopefully we can sum up the question, can you both give a final comment.  Thanks

OK, now that you put it that way I think I finally understand you.

2mbit + 2mbit + 2mbit = 2mbit

If you have 2 nodes (client and server) exchanging data, the fastest data rate will always that of the slowest node. Think of it this way, let's say Joe is a very bright chap and has excellent communication skills. Charlie is lucid as well, but he stutters. So, regardless of Joe's talents, he will helplessly take as much time to understand the conversation as Charlie does to speak a whole sentence.  If two Charlies were to speak to each other, they too would understand each other at the same rate. What Joe could do, is to sustain a conversation with several Charlies at the same time, because he has a wider data exchange rate.
Or consider this. You have a mailbox in your door. It can hold 3 letters. Your friend has a similar mailbox. The postman could be as fast as he can, but still, you would only get 3 letters at a time. If only you had a bigger mailbox, you could receive more letters from your friend (at the same rate though) but you could receive the letters from more than one friend. The frame relay network is extremely fast, but your connection to it, is not, adding to its transparency.


>> Is connections on a framestream network possible?, I understand this is not possible and this is what I am trying to confirm

>> transfers media (avi, jpeg mainly) under our own communication protocol, which we are changing to FTP.

>> ..... as I understand it, it is a Frame Relay system which can carry IP data, perhaps someone can help me understand it better.

On my previous explanation I was assuming several things, but let's do it this way to see if you can get the grasp of Frame Relay.

Regard the OSI model for network communications:

Layer 7, Application : Provides different services to the applications
Layer 6, Presentation : Converts the information
Layer 5, Session : Handles problems which are not communication issues
Layer 4, Transport : Provides end to end communication control
Layer 3, Network : Routes the information in the network
Layer 2, Data Link : Provides error control between adjacent nodes
Layer 1, Physical : Connects the entity to the transmission media

The model states that the end user information (at layer 7) is encapsulated through the rest of the layers (like an envelope inside an envelope, inside another and so on) and decapsulated at the other end of the network.

So you are running under your own protocol for transmiting data. This is, you have a legacy system which takes information, and sends it in your own fashion. I don't know how much of this part you are related to, being a developer, but if you are familiar with sockets, then you might know that you have two options for sending data, UDP and TCP, and you only need to concern about the IP address it is going to get to.

You as a developer are responsible of the upper layers (7 to 5). The network's routing of the packet to its destination represents layer 3 (the IP protocol capsule), the safe delivery of the packet is represented in layer 2 (UDP or TCP capsules) and the last layer 1 deals with the physical transmission (i.e. Ethernet in a switch, which is QPSK modulation on a 4 pair copper wire).

Frame relay takes part in the last layer, the physical layer. You have a byte of data, open a socket, and send it. This packet goes out of your memory to the PCI bus, to the network adapter, to the cable, and very possibly, ends up in a router (typical, though, there are other ways, like a modem). Routers are level 3, so it will only care about sending the packet to the correct IP address. It will look at its interfaces and say, "ok, your packet goes for a different network, so I put it through my WAN interface", which, will not be your typical ethernet connection (CAT 5 cable with a RJ45 plug), but rather, a Frame Relay connection.

Frame Relay does not care what type of content you are sending. It will just split the packet in chunks, encapsulate each of them with its own flags, and send them over your service provider's infrastructure all the way to your branch. It will not do it in an ordered fashion, and this chunks may not even get there at all, but this scenario will be determined as a failure by upper control protocols, actually, only TCP because UDP doesn't care either if the packet ever got there or not.

So, to sum it all up. You can be on a modem, you could be on optic fiber, ethernet, frame relay, wireless, etc. It does not matter, because for the rest of the upper layers, it is all transparent. You will still know the rest of the branches by their IP address, regardless of the physical media to do so.

Now, upgrading your server's connection to 100mbps would do the trick, just so long as you stay in the Frame Relay cloud.
For cloud I mean all the PBX and exchange junctions on your telco company. As I told you before, Frame Relay is a way to take advantage of the hundreds of miles of cable a telco company already has hung nation wide so you don't have to cable all your way up from hq to a branch by yourself. In other words, you're riding on somebody else's train (side note: this doesn't mean Frame Relay is exclusive to telco companies, but it is certainly quite common to find it only there). If you hooked an ADSL modem to your server with a connection to Internet instead, you would have to make sure first that you will still be able to be inside this cloud, which I can almost assure already, you won't. The cloud is a private network. Internet is not.

So, the 100Mbps would be increased in the link you already have. Several service providers already have your NTU (the box at your premises) completely wired, though, not all interfaces may be active. You could request an upgrade to your telco company to do this.

I hope you got the grasp of it already, but still, we could look into a different solution rather than upgrading your server's connection (that's expensive), namely, file distributing. I didn't go into it here cause it is a whole different subject.
DaveRowlandAuthor Commented:
So if i got this right:-
1:  Adding mirror servers isn't going to help, the pipes on a frame stream network would not benefit from this, so p2p or 4-5 mirrors servers would not help, we would just jam up the network and gain no speed advantage as the data is repeated around the network anyway and each network point listens for the data it needs, so we would end up just putting more cars on the highway and end up with traffic jams.
2:  If we increased one connection point (server) this will increase all connection points.  Telco's only offer this by design of the cloud, I cannot put extra pipes in to the servers location?  Things like that?
3:  If a framestream network is busy with one person, that impacts on everyone else for download speed.
4:  Most ADSL city/town broadband areas share, I guess this is a mini frame stream network per city/town.  (Contention 50:1).
5:  star topology (which it is) each branch can only see hq, and hq can see all branches, I forgot about that but too and I suspect that is like subnet masking in the router.  But as you said, no point in forcing a few connections to point at the server as the utilization of the network remains the same.
6:  Our data is not the only one on the framestream, it has a dual purpose so we cannot butcher the connections in any expensive way.  Like one network for servers, then lots of others for regions of the uk (south network) etc.

1a:  We could increase the network speed for the cloud.
2a:  Invest in ADSL technologies and offload onto a few FTP servers on the net.

Thats cleared a lot of things up.  Many thanks, it was really just some outstanding questions that I needed a second opinion that we are doing things right as we have a meeting next week.  i did understand the principles but wasn't sure if I was forgetting some of the angles where mistakes could happen.  It could be embarrising.

Question has a verified solution.

Are you are experiencing a similar issue? Get a personalized answer when you ask a related question.

Have a better answer? Share it in a comment.

All Courses

From novice to tech pro — start learning today.