Re: What are the theoretical aspects of network switching ?
Please allow me add a few comments to my friends' wonderful answers.
The different types of switching you have referenced here do have the word "switching" in their name, but they are not all comparable because they describe not only different kinds of switching but also refer to processes of principially different nature and purpose. While the word "switching" may suggest that all these are simply variations on the same subject, in reality, these all types of switching are from different areas of networking and care should be taken not to confuse them.
The first category of switching is related to how does a particular network basically operate when it provides you the service of carrying your data. Here we talk about circuit switching, packet switching, message switching and cell switching. Circuit switching describes a network technology that reserves a particular set of resources (bandwidth, sequence of links and devices along the path, timeslots, etc.) for the entire duration of your session with the opposite device, regardless of whether you really send data or not. Circuit switching is typical for telephone or ISDN networks. Packet, message and cell switching are another way of providing data transfer service by splitting the data into pieces and carrying them individually over the network, with the network sharing its resources among multiple senders without doing any permanent reservations for your session. Packet and message switching refers to technologies that use variably-sized messages (Ethernet, Token Ring, FDDI), cell switching refers to a technology using fixed-size messages (ATM).
The second category is concerned with the particular operation of Ethernet switches. The methods that belong here are store-and-forward, cut-through, fragment-free, adaptive. In short:
store-and-forward switching means that the Ethernet switch will receive the entire Ethernet frame, check its control sum, and if it is correct, it will forward it further
cut-through switching is a general term that describes Ethernet switching technologies that begin forwarding the received frame before it is received in its entirety. The cut-through category splits itself into two subclasses:
fast-forward: switching of the Ethernet frame begins after receiving the first 48 bits (6 bytes - exactly the source MAC)
fragment-free: switching of the Ethernet frame begins after receiving the first 64 bytes (the minimum size of a correct Ethernet frame)
adaptive switching is actually a hybrid of the previous two switching methods: the switch initially uses one of the cut-through methods to forward frames. However, if the number of corrupt (undersized) frames exceed a predefined threshold, the switch will move to store-and-forward method to prevent corrupt frames from forwarding
The third category is concerned about the process involved in routing an IP packet in particular. Here, the methods are process switching, interrupt context switching, fast switching, optimum switching and Cisco Express Forwarding. I am not going to dive into those; instead, I suggest you read the following document:
What are the theoretical aspects of network switching ?
There are four types how a Packet are Switched or Packet Switching:
1- Process Switching.
In this Type, all the Packet are Sent to the CPU to make the forwarding decision.
2- Fast Switching.
Its a table that created on Demand, which contains Fast Switching table (Route Cache), its vital to understand that this table doesnt contain all active switching topology.
3- ASIC (Application Specific Integrated Circuit).
ASIC is a hradware that is correctly programmed to make forwarding decision, packet doesnt need to hit the CPU or be fast Switched for forwarding. and ASIC is mostly found on a Multilayer Switches line cards.
4- CEF or (Cisco Express Forwarding).
The CEF table contains Two table a) the FIB table B) Adjacency Table.
The FIB Table contains all the Routing information base RIP and the Adjacency table contains the Layer-2 information.
CEF is considered a Topology driven Switching mechanism because it contains all Routing information and updates and the RIP table is copied into the FIB once the router boots up.
CEF is the best Switching Method that should be used today, also note , its a requirment for CEF to be enabled in an MPLS environment for Label Switching.
We are pleased to announce availability of Beta software for 16.6.3. 16.6.3 will be the second rebuild on the 16.6 release train targeted towards Catalyst 9500/9400/9300/3850/3650 switching platforms. We are looking for early feedback from custome...