Open flow is a new kind of protocol designed to run on Switches and routers. It’s built on a separated base between the forwarding plane and control plane Joined through a secure channel. The controller, within the control plane, handles the management of the tables and traffic flow through updating, deleting and adding of flows while the forwarding plane handles traffic forwarding. Open Flow is designed in a way that the devices can separate research traffic from the normal operations traffic within a network.
In this protocol a user or researcher an control his/her traffic directly through attaching the routes, actions and how it should be handled. So the Which traffic’ – type of traffic we are sending I. E. Is it multicast, broadcast, echo requests etc, Where to’ – to which addresses are we sending it to ‘and how / what action’ – how should it be handled. These questions are important in running this protocol as they have to always be answered. Open Flow propagation is based on the existing inflexibility of the available network devices in leaning towards testing new operations, protocols etc.
A good example loud be someone coming up with a new protocol example would be RIP v and would want to test it. So s/he would run it through the controller and picks the routes the traffic would have to take and also the actions to be carried out on the traffic. This can be done on all switches the traffic is to traverse while separating it from the traditional traffic in the network Delving into those three basic parameters which should be well defined, its imperative to note that Open Flow devices should have a routing table, call it a Content Addressable Memory (CAM) table for the switches.
In this table all entries should have packet headers that are used to guide the flow, an action tag on how traffic should be processed I. E. Either let through or dumped and the counters that show amount of packets processed or passing through for each flow, set queues, fields etc. Secondly there must be a controller into which all commands that control traffic are put and finally a channel that connects the table to this controller. Note that, before the packets are let to the secure channel, packet headers could be encapsulated with only the packet ID visible for deciphering.
This could be true for PLAN tagged traffic which shows a specific PLAN number or an IP flow etc. 1. 1 . Open flow architecture. Figure 1. 1 showing a simple Open Flow architecture. There are three major components of the Open Flow I. E. The switch, the controller and the channel. The switches are categorized into two I. E. The traditional switches and the Open Flow enhanced switches. Both types of switches forward traffic from one port in the switch to another. Normally it’s at a line rate pace. During switching in the Open Flow enhanced switches, packets are normally encapsulated and passed ever to the controller through a secure channel.
Switches and routers can also sieve through and decide depending on the set parameters which traffic to let through and which one to drop. The switch or router should also be able to continue passing traditional or normal traffic through concurrently to personalized or research traffic. Depending on the administrator policies a little research traffic can also be allowed to traverse the network; however this could be running in a specified PLAN. Type O switches as currently known can do all these duties. More advanced switches r call them type 1 switches are in the making and will have more enhanced features.
Examples of switches in the market today supporting these enhanced features are HP Procure zloty supporting Open Flow versions 0. 8. 9 which has 48 ports expandable to over 500 with a speed of 1 Gaps. Pronto is another type of switch having similar features as the HP above but includes higher speeds of up to gasps ports. Its supports Open Flow version 1 . NECK PAYOFF is another example of these switches with similar features as the HP but carries an extra 2 ports running at 10 Gaps speeds. Controllers are another component of Open Flow. They act as the brains behind the operations.
They can remove and act entries to the flow tables and they can also be static or dynamic in their operations. Controllers depending on their complexity can be accessed or used to handle a number of operations from various researchers. Each user can be having a different kind of traffic, set different control rules and various actions attached to their traffic. Examples of controllers are Big Switch which uses a beacon in production enterprises and uses a command line interface in managing networks. Hellos a C based open flow, Maestro which is Java based, SNACK based on KNOX and many others.
In comparison to history, Open Flow has given a new platform to network research. Before one could run any new protocols, they had to liaise with vendors to allow work on their platforms or building their own equipment out of the normal hardware, but with Open Flow this is possible since enterprises and network operators can now design their own software and implement features they want without waiting for new vendor products. Open Flow can easily align business objectives with the networks.
Research can easily be directed and tested concurrently with normal networks operations. 1. 4 Disadvantages of Open Flow. The biggest disadvantage with Open Flow is its age or infancy. The latest version having been released in 2011, most of its new features are still in the trial errors and have not yet been standardized. So its scalability in terms of usage is still low. Most vendors are still reluctant to adopt it in their devices. Most of the applications being developed still need to be tested to see how far they can scale which is a big hurdle that needs to be crossed.