1

While studying for CCNA, I came across the information that in a network device, say a switch, the CPU is responsible for the control and management planes work, while the actual data forwarding is done by ASICs, chips built specifically for that. Thus, the forwarding is done 'in hardware'.

Recently, I read some documentation of a network switch which got me confused. It seems that the switch in question has its ports connected to PHYs, chips that implement the L1 functions. Next, the PHYs are connected to Switching Chips (Broadcom Avenger BCM53162M), which, in turn, are connected in between them, and to the CPU (TI AM355x). Moreover, the switches in the series, might have additional software capabilities (such as additional protocols) if FPGAs are present on the PCBA.

My questions are:

  • what is the purpose of the switching chip? Since they are connected to PHYs, are they the chips implementing the MAC layer functions? Or are they just switching the frames?
  • why is an FPGA necessary for additional software capabilities? Shouldn't the CPU be doing the processing of any protocol implemented in the OS?
  • which of these components (PHY, switching chip, FPGA) qualify as ASICs?

2 Answers 2

3

The purpose of the switching chip is to perform the switching functions of the device. What they specifically do is unique to the chip in question but generally yes, they perform some of the layer 2 forwarding and other functions.

They are not just forwarding frames. They do other things as well such as consistency checks, and other work but it depends on the chip in question.

FPGA are not necessary and are actually only commonly found on higher end equipment. They are as their name suggests field programmable, they can be reprogrammed to perform different functions after being deployed in the device. This makes them much more expensive than normal purpose designed chips. Usually such needs are handled either in the ASICs or via software on the general purpose CPU.

Many chips qualify as ASICs because they are Application Specific. General purpose CPUs are not application specific (they are designed to run any software compliant with their general design) so they are not ASICs. Switching chips or other network interface specific chips are considered ASICs because they are Application Specific in their design and function, you can't reuse it for other purposes than the network functions it is designed to do.

1

what is the purpose of the switching chip?

The switching chip implements the required layer-2 functions and does the actual forwarding between ports. It often has additional function, depending on the switch. There may be multiple, interconnected switching chips.

why is an FPGA necessary for additional software capabilities?

On FPGA may be necessary for additional hardware functions not present in the ASIC(s). You could also use additional ASICs but FPGAs can provide software-configured hardware functionality, without any chip development.

which of these components (PHY, switching chip, FPGA) qualify as ASICs?

Since an ASIC is application specific by definition, an FPGA doesn't qualify. The others do.

Not the answer you're looking for? Browse other questions tagged or ask your own question.