If you like the style of Practical Guides, let Niels help you understand Operating Systems in his newest book:
A Guide to Operating Systems: Troubleshooting and Problem Solving
Practical Guide to Frame Relay
...because it doesn't have to be complex.
by Niels Jonker
Copyright © 1997 Niels Jonker
All Rights Reserved
1. What is Frame Relay?
Frame relay is a technique used to transport data from locations to location, just like T-1 lines or ISDN connections do. In frame relay, there are a number of locations on the network that can send and receive data. These connections are known as Ports. Each location that needs access to the frame system, needs to have one of these ports.
Every port in a Frame Relay system has an Address. This address is Unique to the port at that specific location. The port is connected to the equipment that handles the Data on one side, to the Frame Relay Cloud on the other side.
The equipment that handles the data can send data out the frame relay port. This happens in the form of Packets, or Frames. Each frame is built up of two parts; the actual Data and the Control block. These frames are sent over Virtual Connections.
The frame network itself should be imagined as a cloud. Every Frame Relay Access Port runs into this Cloud. Inside the cloud, there are a lot of different ways to get from one port to another. These ways are all interconnected by Frame Switches. These switches can make informed decisions on the traffic flow over each part of each route. Together, they figure out what router to send a packet over to get it fast and reliable from the source to the destination port.
Let's for example say that we have two sites connected to a frame relay network, and we wish to send data from site A to site B. The diagram would look like this:
What happens INSIDE the Frame Cloud is not known, and does not matter to us. What we do is simply address data as Going to B or Going to A. So, at this point, things work pretty much the same as they would on a dedicated line between A and B. When we add a Location C, things get more interesting:
Now, the three sites can simply talk to each other by passing packets with the correct addresses. In a conventional setup with leased or dialup lines, this would have meant that either each site would have to have two lines, or that traffic would have to pass through two sites. Not in frame relay, the cloud takes care of all of this!
What the cloud does for us, is form 'Virtual Connections'. The way this is done, is simple. Each connection from one site to another is defined. For example, we say there is a connection from site A to site B. From site A to site C, etc. Each end of each connection is given a number that identifies it, this is called a Data link Connection Identifier, or DCLI. For everyone in the above diagram to be able to talk to everyone directly, these are the virtual connections that would be needed:
This means that everyone can talk to each other by simply stating the address and DLCI number of the line they want to talk over. This will get the data across the Virtual Connection.
Now keep in mind that all the data for all virtual connections arrives over one line, in one port! This means that the Equipment itself takes care of picking the data apart and putting the correct packets in the correct directions. these virtual connections come in two flavors, being Fixed and On Demand.
The fixed variety is referred to as a Private Virtual Circuit, or PVC. A PVC is setup by the carrier that carries the frame-data from A to B. Simply addressing the data with a port address and DLCI will get it over the Frame Cloud to the right port.
the On Demand variety is called a Switched Virtual Circuit or SVC. This type of circuit is brought up when needed, just like making a phonecall. If A wants to setup an SVC to B, it will send a control packet in LAPD format, which will bring up a SVC. Both sides will be assigned a temporary DLCI, and they are now ready to talk.
The multiple DCLI's coming into or going out of one location are all handled, as said before by one port into the frame network. This port comes in over a traditional form of media, it can be dialup, ISDN to T-1 or higher bandwidth. Let's say for example that all ports in our network have a 512 Kilobit bandwidth. If A and B start sending data to C at the same time, the total combined bandwidth will be 512+512 = 1024Kilobits. Since C only has a 512 Kbit port, we have a problem! Frame Relay uses a technique called Statistical Multiplexing to fix this; each of the two gets a fair share of the bandwidth. The rest of the data is temporarily buffered in the Frame Relay Cloud. It will be sent as soon as possible, when the dataflow lets off a little.
The risk in this is of course that the dataflow may not let off; if the buffers fill up and there is nowhere to go, the frame relay cloud will drop the data. In our uses, this is no problem, since the protocols running over frame will re-send the data anyway. It is however something to consider, since it may cause slowdowns in the network.
2. How does Frame Relay differ from other Techniques?
The biggest difference in frame Relay from other techniques is the use of Virtual Connections rather than Static Connections. As shown before, each location can have one port into a Frame Relay Network. From this port, it can have multiple Virtual Connections to various locations. It can make multiple redundant connections possible through the use of PVC's between various routers, without having to use multiple physical links.
Also, since Frame relay is not media specific, and offers a way to buffer speed differences, it can make a good interconnect medium between various devices that run at various speeds.
The multiplexed nature of Frame Relay facilitates especially the transmission of Bursty Traffic. In a traditional fixed-bandwidth multiple connection scenario, a lot of bandwidth will be wasted at most times, since it is not actually being used. The bandwidth on frame is shared, allowing for multiple bursts to be handled sequentially, therefor allowing better utilization of bandwidth. The chance of congestion is however also greater, since the bandwidth in the Frame Relay Port may become a bottleneck.
frame Relay pricing is where things really get different: From the above discussion we know there are two elements to Frame Relay, the Access Port and the Private Virtual Circuits. Per port, we can feed as many circuits as we please (up to the equipment's limit..). The price of the port depends on it's bandwidth. A bigger port costs more money. The price of a PVC is fixed, and not dependent on bandwidth or usage. Data sent over the Frame Connection is not subject to additional charges. The interesting part about this pricing is that there is NO DISTANCE CHARGE involved! A Frame Relay Access Port is expensive. Two frame ports between two cities are usually also more expensive than a direct T-1 as long as there is only a one or two LATA Hop, say for example Knoxville to Nashville. However, when you feed multiple cities and Hop multiple Lata's, a Frame solution is an attractive one. In local traffic, Frame becomes attractive when you feed a large number of connections, at Fractional T-1 bandwidth generally 8 or more.
3. What is frame relay useful for?
Frame Relay can be used for various types of connections. It should be seen as a flexible protocol that lies on the Data Level of the Connections between routers. It can currently be used effectively for carrying all sorts of data, up to speeds of about 4 megabits per second.
Due to the implementation and cost of frame relay, it is most suitable for permanent or semi-permanent connections. It is not desirable as of yet to access frame relay on a dialup basis for economical purposes.
For a number of U.S. Internet clients, use of Frame Relay may allow for the connection of various remote sites using one technology. It is possible to connect the clients router to Frame Relay and access Internet Services over this circuit. At the same time, if the customer has other Frame Relay ports available in different locations, the same port may be used to connect to Wide Area offices the client has. This integrated approach can be attractive for both parties; from one Frame port, U.S. Internet can connect multiple customers. At the same time, the client can connect to multiple remote locations using the one port. Although there is a greater chance of connection then when dedicated lines are used, the economical advantage will in most cases compensate for this.
4. How does frame relay fit in the U.S. Internet strategy?
Initially, U.S. internet is deploying Frame Relay ports for some of the long distance connections required to connect to remote Points of Presence. The cost of these ports is substantial, and the equipment required to operate successfully in this architecture is only now (September 1995) being deployed.
Sprint has been chosen as our frame Relay provider. Initially, we will operate with roughly 1.5 megabit of access bandwidth out of two locations (Knoxville and Nashville). The POP's in Alabama, Chattanooga, Memphis, Johnson City, Clarksville and Pittsburgh, PA will all come on-line on this frame relay system. Several PVC's will be installed to allow for redundant hookups between the various locations.
In a later stage, we may start offering clients access to the network over frame Relay. In this scenario, clients in remote cities where we do not have bandwidth available to service them would be helped by getting in to the Frame cloud, and connecting into one of U.S. Internet's routers.
5. Mixing Frame Relay with Other Technologies.
Frame Relay is nearly a protocol that describes how packets of data move from one location to another; it does not concern itself with the issues of how packets get to and from the frame network. Several techniques can be used to get data in a frame port. U.S. Internet will initially be using Dedicated Fractional T-1 connections to connect to the Frame Cloud.
Next, access over High Speed Modem and ISDN will be investigated. it is technically possible to connect to a Frame cloud using an ISDN connection, then become a part of the frame network, and connect to U.S. Internet's services. Currently, cost of technology and cost of service are prohibitive factors in this scenario, but companies requiring more complicated Wide Area Solutions including internet may wish to look into this option.
6. Pro's and Cons of Frame Relay.
PRO: In many scenario's involving long haul, high speed connections, it is cheaper than dedicated lines.
PRO: There is a cheap solution to incorporate redundancy in the network.
PRO: Mixed speeds can be converted, traffic bursts can be buffered.
PRO: Less hardware is needed to for the same amount of connections
CON: There may be jams; no guaranteed bandwidth
CON: In a point-to-point scenario it is not economically feasible.
CON: In short haul, it is not economically feasible.
7. Technical Aspects of Installing and Running Frame.
The connection to the Frame Relay Network will be made over dedicated T-1 connections. These connections will terminate in a DSU at U.S. Internet, just like a T-1 does. The device used to handle the special frame signaling is called a Frame Relay Access Device, or FRAD. In our case, this function will be performed by the software in our Cisco routers.
In the POPs, 2500's or 4000's will serve as the FRAD, in the NOC's, the 4500's will take care of the FRAD functionality. For implementation of Sub-interface Routing, they will also function as a Frame Switch.
Attached is the Cisco documentation regarding Internet Technologies and Frame relay, a more general and detailed overview of the technology, and Frame Relay Configuration, which goes into the exact configuration strategies and commands in greater details. These documents, as well as many others, are also available on the World Wide Web. These attachments are Copyright © 1995 Cisco Systems Inc, All Rights Reserved.