|
Previous
|
Content
|
Next
|
|
| 2.4.1 The router queues |
|
|
|
| Next figure would help to understand how packet forwarding process takes place on the router. |
|
|
| The router is represented by the big box. For our scenario it has two interfaces. This time we are interested on flows going from network connected to interface IF-0 (network A) to network connected to interface IF-1 (network B). After packets enter from interface IF-0 they are placed on queue number 0 which is called the ingress queue. Next, inside the router, the forwarding process takes place. Based on destination address on each packet header, some packets entering from interface IF-0 will be placed in the queue number 1 corresponding to the interface IF-1. This queue is called the egress queue. |
| It's very important to note that ingress queue will be most of its time empty. This is true because the forwarding process is very fast (made by the cpu processor drived by the routing protocol and operating system software). Rate of forwarding is always greater than the rate at which the interface IF-0 can accept packets coming from network A. On the contrary, egress queue will be most of its time busy, depending mainly, of course, of the incoming rate of packets entering by the interface IF-0 and the outgoing rate of packets leaving to network B through the output interface IF-1. |
| Being the incoming rate lower than the outgoing rate (for example, if interface IF-0 is a serial interface connected to a 1.5 Mbps link and interface IF-1 is an ethernet interface connected to a 100 Mbps, and not so much congested, ethernet network) the egress queue will be almost empty. The forwarding process goes nice and everybody is happy. On the contrary, being the incoming rate higher than the outgoing rate (if the interface IF-0 is the high throughput 100 Mbps ethernet interface and the interface IF-1 is the low throughput 1.5 Mbps serial interface), then the egress queue will be almost full, and when overflows, packets have to
be dropped not having space on it to store them for preserving its survival. |
|
|
|
| Observe also that queues we are talking about and representing in our figure are just Fifo queues. Packets enter at the tail and leave the queue from the head. Let's see this in more detail. |
|
|
| This sixteen packets capacity queue is moving 20 packets. Packets 1 and 2 already entered and left the queue, packets 3 to 18 are just in the queue and packets 19 and 20 are in the process of entering the queue. Of course, before a new packet can be accepted another has to be dispatched. If packets arrive very fast, faster than they can be dispatched, the router doesn't have any other choice that dropping them. But we don't want this; we want to conserve our packets to protect protocol behavior and applications than send/receive these packets. |
| When we see all these packets competing to reach the head of the queue to be dispatched, we can leave our mind flying and think that because they are all over here together we can begin to attempt to exercise some control over them. For example, to control which of them are going to be forwarded and how fast this process will be done; which of them could be delayed in detriment of some other; and which of them, lamentably, will be dropped. When we assemble all these ideas in our brain, in fact, we are not discovering the wheel, because as this, it was
invented some time ago. As we told somewhere above, router queues is where the battle for
controlling packets takes place. Our next step in our way will be to study briefly these queues, and what type or flavor of them have been invented to help us in our wishes of controlling flows better. |
|
|
|
|
|
Previous
|
Content
|
Next
|