WEBVTT

0:00:02.880000 --> 0:00:07.260000
 Hello and welcome to this video titled,
 congestion avoidance with policing

0:00:07.260000 --> 0:00:14.160000
 and shaping. In this video I'm going
 to cover an overview of what is meant

0:00:14.160000 --> 0:00:18.520000
 by the term or phrase
 congestion avoidance.

0:00:18.520000 --> 0:00:20.820000
 And by the time we're done with this
 video you'll understand what these

0:00:20.820000 --> 0:00:25.500000
 terms mean of policing,
 shaping and markdown.

0:00:25.500000 --> 0:00:30.180000
 And I'll compare the differences
 between policing and shaping.

0:00:30.180000 --> 0:00:35.400000
 So what is congestion avoidance?

0:00:35.400000 --> 0:00:40.100000
 So we know that the whole reason why
 we need quality of service in the

0:00:40.100000 --> 0:00:43.720000
 first place is to deal with
 links that are congested.

0:00:43.720000 --> 0:00:47.620000
 If there was no congestion anywhere
 in any interface we wouldn't need

0:00:47.620000 --> 0:00:51.800000
 QoS. As soon as the packet came in it'd
 go right out, it wouldn't delay,

0:00:51.800000 --> 0:00:53.720000
 it wouldn't be dropped,
 we'd have no problems.

0:00:53.720000 --> 0:00:57.940000
 But the problem is there is congestion
 in networks and unfortunately a

0:00:57.940000 --> 0:01:01.200000
 lot of times you don't know where
 that congestion is going to be.

0:01:01.200000 --> 0:01:06.460000
 So congestion avoidance is really talking
 about it's a general term used

0:01:06.460000 --> 0:01:10.840000
 to define a set of features that attempt
 to prevent cues from becoming

0:01:10.840000 --> 0:01:16.500000
 congested. So this is our overall objective
 here is that if I have a situation

0:01:16.500000 --> 0:01:24.980000
 like this where I have a whole bunch
 of input ports, maybe these are ports

0:01:24.980000 --> 0:01:28.800000
 leading to hosts, maybe ports leading
 to other parts of my network, they're

0:01:28.800000 --> 0:01:32.600000
 all coming into this device and all
 the traffic that's coming in those

0:01:32.600000 --> 0:01:37.660000
 ports is being funneled
 out one egress port.

0:01:37.660000 --> 0:01:42.700000
 Well this egress port might get congested
 because it just can't put those

0:01:42.700000 --> 0:01:45.420000
 things out on the
 wire fast enough.

0:01:45.420000 --> 0:01:49.040000
 And if I allow all those things to
 be directed to this egress port and

0:01:49.040000 --> 0:01:52.500000
 these are my memory buffers right here,
 those memory buffers could get

0:01:52.500000 --> 0:01:56.600000
 full. And I have no choice but to start
 dropping packets because I got

0:01:56.600000 --> 0:01:58.620000
 no place left to put them.

0:01:58.620000 --> 0:02:01.740000
 So congestion avoidance like it says
 is a bunch of features that say hey

0:02:01.740000 --> 0:02:07.100000
 you know what, let's try to stop these
 packets or maybe some of these

0:02:07.100000 --> 0:02:11.180000
 packets from getting into this queue
 in the first place, maybe the low

0:02:11.180000 --> 0:02:15.880000
 priority, the stuff that's of lesser
 importance to us to always make sure

0:02:15.880000 --> 0:02:21.720000
 there's some room in that queue
 for the higher priority stuff.

0:02:21.720000 --> 0:02:24.580000
 And that's what we're talking
 about right here.

0:02:24.580000 --> 0:02:26.560000
 So where would you do this?

0:02:26.560000 --> 0:02:30.320000
 Well there's three general places depending
 on your platform and depending

0:02:30.320000 --> 0:02:34.180000
 on what architecture have that
 will allow you to do this.

0:02:34.180000 --> 0:02:44.300000
 Number one, we could do it on the ingress
 interface queue, on the ingress

0:02:44.300000 --> 0:02:54.320000
 interface if these are all of my ingress
 ports and that's my one egress

0:02:54.320000 --> 0:02:58.640000
 port. Well we know that each one of
 these ingress ports is also going

0:02:58.640000 --> 0:03:03.080000
 to have some memory, some queues
 and buffers for storing things.

0:03:03.080000 --> 0:03:08.680000
 So one thing we could do is we could
 say alright, well let's monitor that

0:03:08.680000 --> 0:03:14.380000
 and if this thing right here for this
 one port, fast ether zero slash

0:03:14.380000 --> 0:03:18.420000
 one, if that gets above a certain threshold,
 if that memory gets to like

0:03:18.420000 --> 0:03:23.900000
 80% full, let's just start preemptively
 dropping one out of every 10 packets

0:03:23.900000 --> 0:03:28.080000
 before it even gets in there, just
 drop it as it arrives at the switch

0:03:28.080000 --> 0:03:33.440000
 or router. Or let's maybe delay some
 of the traffic, you know normally

0:03:33.440000 --> 0:03:37.100000
 once it gets into that ingress queue
 it'd be looked up in microseconds

0:03:37.100000 --> 0:03:39.840000
 and then forwarded to
 the egress queue.

0:03:39.840000 --> 0:03:42.020000
 Well let's not do that.

0:03:42.020000 --> 0:03:45.900000
 Let's take some of those packets and just
 hold them back for a few milliseconds,

0:03:45.900000 --> 0:03:49.960000
 have some artificial delay induced
 in there to hopefully prevent this

0:03:49.960000 --> 0:03:53.080000
 guy from getting overwhelmed
 and congested.

0:03:53.080000 --> 0:03:56.180000
 So that's performing it at
 the ingress queue level.

0:03:56.180000 --> 0:04:01.240000
 Another place, we could actually
 do it at the forwarding engine.

0:04:01.240000 --> 0:04:05.780000
 We could say okay, as the forwarding
 engine is receiving all these packets

0:04:05.780000 --> 0:04:10.180000
 for look up, let's have the forwarding
 engine count them as they come

0:04:10.180000 --> 0:04:12.380000
 in, see what kind of rate
 they're coming in.

0:04:12.380000 --> 0:04:17.020000
 As long as they come in you know above
 a predefined threshold like at

0:04:17.020000 --> 0:04:21.380000
 five megabits per second or lower,
 we'll forward them to the outbound

0:04:21.380000 --> 0:04:23.060000
 interface and let them go.

0:04:23.060000 --> 0:04:27.240000
 But if they come in above that threshold,
 anything that's above that line,

0:04:27.240000 --> 0:04:29.320000
 we're going to kill them,
 we're going to drop them.

0:04:29.320000 --> 0:04:33.080000
 Once again we're trying to preemptively
 prevent that egress queue from

0:04:33.080000 --> 0:04:33.980000
 getting filled up.

0:04:33.980000 --> 0:04:39.620000
 Now the forwarding engine is doing
 it using something called policing.

0:04:39.620000 --> 0:04:43.940000
 Or technically we could try to do
 it at the egress queue itself.

0:04:43.940000 --> 0:04:52.240000
 This is not a very good way of, well
 let me pause and take a step back

0:04:52.240000 --> 0:04:57.280000
 on that. You can do it the egress
 queue with the egress queue.

0:04:57.280000 --> 0:05:03.320000
 What you could say is if this is my
 switch right here and here's all my

0:05:03.320000 --> 0:05:07.260000
 ingress ports, and this could also
 drive a router too, and this is my

0:05:07.260000 --> 0:05:12.360000
 egress port where stuff is being transmitted,
 and this is my memory buffer.

0:05:12.360000 --> 0:05:22.980000
 Now in the world of QoS, typically
 what you would find is that most of

0:05:22.980000 --> 0:05:27.620000
 your traffic is your default traffic,
 your low important traffic.

0:05:27.620000 --> 0:05:32.600000
 Like people just browsing the web, you
 know, every once in a while that

0:05:32.600000 --> 0:05:35.300000
 stuff gets dropped,
 not a big deal.

0:05:35.300000 --> 0:05:38.760000
 Your high priority traffic
 is like your voice, okay?

0:05:38.760000 --> 0:05:42.220000
 But in the scheme of things, if we look
 at the sum total of all packets

0:05:42.220000 --> 0:05:46.400000
 going through your network, your voice
 packets are a small percentage.

0:05:46.400000 --> 0:05:50.860000
 Your voice packets are maybe like 5
% of all the packets going through

0:05:50.860000 --> 0:05:53.000000
 your network if that.

0:05:53.000000 --> 0:05:56.700000
 So keeping that in mind, we could say,
 well here's what we're going to

0:05:56.700000 --> 0:06:01.020000
 do. If we just let everything go about
 as normal, we didn't touch it,

0:06:01.020000 --> 0:06:05.280000
 this queue could get potentially filled
 up, and what would probably fill

0:06:05.280000 --> 0:06:07.840000
 it? Would it be our voice
 packets are filling it up?

0:06:07.840000 --> 0:06:11.640000
 Probably not. It'd be our low important
 stuff, like people who are doing

0:06:11.640000 --> 0:06:16.480000
 YouTube or people who are, you know,
 uploading files to Dropbox or stuff,

0:06:16.480000 --> 0:06:20.160000
 stuff that's not as critical
 as our voice traffic.

0:06:20.160000 --> 0:06:24.160000
 So in order to prevent this thing from
 getting filled to the point when

0:06:24.160000 --> 0:06:28.400000
 a voice packet does come in, uh oh,
 there's no place for it, drop.

0:06:28.400000 --> 0:06:31.700000
 In order to prevent that, we could
 actually go to the egress queue and

0:06:31.700000 --> 0:06:36.620000
 we could say, all right, let's set
 some thresholds there in the egress

0:06:36.620000 --> 0:06:41.300000
 queue. Let's tell this egress queue
 here that, you know, if you get to

0:06:41.300000 --> 0:06:47.340000
 40% full, why don't you start dropping
 maybe one out of every 10 frames

0:06:47.340000 --> 0:06:51.080000
 it comes in? You know, even if you got
 all this space right here, start

0:06:51.080000 --> 0:06:53.880000
 dropping one out of every
 10 frames anyway.

0:06:53.880000 --> 0:06:59.020000
 And if you get to like 75% full, let's
 start dropping like six out of

0:06:59.020000 --> 0:07:00.180000
 every 10 frames.

0:07:00.180000 --> 0:07:04.160000
 You see the objective, the purpose behind
 this is we're desperately trying

0:07:04.160000 --> 0:07:08.660000
 to reserve some room so that when a
 voice packet does come or whatever

0:07:08.660000 --> 0:07:13.940000
 it is that you define as being important
 to you, we have room for it.

0:07:13.940000 --> 0:07:18.820000
 So we could do it on the egress queue as
 well to try to prevent the congestion

0:07:18.820000 --> 0:07:23.520000
 from happening. So those are
 the three different places.

0:07:23.520000 --> 0:07:28.940000
 Now let's just do some terminology here
 to wrap up this video, policing,

0:07:28.940000 --> 0:07:30.500000
 shaping, and markdown.

0:07:30.500000 --> 0:07:34.300000
 And my objective here in this particular
 video is not to go into all the

0:07:34.300000 --> 0:07:38.900000
 gory details of how these features are
 configured, your various options

0:07:38.900000 --> 0:07:40.940000
 and gotchas and caveats.

0:07:40.940000 --> 0:07:44.300000
 That is for a much more advanced level
 series and we actually here at

0:07:44.300000 --> 0:07:49.160000
 INE have a lot of videos that do
 go into that level of detail.

0:07:49.160000 --> 0:07:52.980000
 So after this, after you've been introduced
 to these features, if you

0:07:52.980000 --> 0:07:57.740000
 want that type of knowledge, please
 go back to our website, my.ini.com

0:07:57.740000 --> 0:08:00.960000
 and you'll find some more advanced QS
 videos that go into that level of

0:08:00.960000 --> 0:08:05.080000
 detail. So for here, I want to sort
 of help you define what is meant by

0:08:05.080000 --> 0:08:08.100000
 these terms, policing,
 shaping, and markdown.

0:08:08.100000 --> 0:08:15.120000
 Okay. So this type of thing is these
 features are typically implemented

0:08:15.120000 --> 0:08:20.260000
 between your edge router at the edge
 of your network and the ISPs router

0:08:20.260000 --> 0:08:22.240000
 at the other end of
 that connection.

0:08:22.240000 --> 0:08:25.060000
 So between you and your service provider,
 it's providing you internet

0:08:25.060000 --> 0:08:30.820000
 service. Now, when you set up that connection
 with your ISP, one of the

0:08:30.820000 --> 0:08:33.340000
 things they're going to ask you and this
 is going to directly affect your

0:08:33.340000 --> 0:08:38.660000
 monthly bill is how much speed
 or how much bandwidth you need.

0:08:38.660000 --> 0:08:43.080000
 I mean, clearly they might have the
 capability of running a fiber optic

0:08:43.080000 --> 0:08:46.240000
 cable to you. They'll give you
 a 100 gigabits per second.

0:08:46.240000 --> 0:08:49.720000
 Of course, they're going to charge you
 like $5,000 a month for something

0:08:49.720000 --> 0:08:54.820000
 like that. And if your company doesn't
 need even a fraction of that, why

0:08:54.820000 --> 0:08:56.240000
 would you pay for that?

0:08:56.240000 --> 0:09:00.660000
 So they're going to ask you something
 like what kind of contracted rate,

0:09:00.660000 --> 0:09:04.560000
 what kind of committed information
 rate do you need from us?

0:09:04.560000 --> 0:09:08.380000
 Because we can run a cable to you that's
 capable of supporting out of

0:09:08.380000 --> 0:09:10.300000
 this world speeds.

0:09:10.300000 --> 0:09:12.460000
 What do you actually need?

0:09:12.460000 --> 0:09:15.980000
 So you're going to negotiate that with
 them and negotiate a contracted

0:09:15.980000 --> 0:09:18.100000
 rate called a CIR.

0:09:18.100000 --> 0:09:22.480000
 Now, once that's agreed upon between
 the two of you, it's up to both of

0:09:22.480000 --> 0:09:24.320000
 you to enforce that.

0:09:24.320000 --> 0:09:28.700000
 You see, on their end, for the ISP,
 they're saying, okay, that customer

0:09:28.700000 --> 0:09:32.220000
 better not send us any more
 than what we've agreed to.

0:09:32.220000 --> 0:09:35.960000
 If we've agreed to a committed information
 rate of like maybe 100 megabits

0:09:35.960000 --> 0:09:40.880000
 per second, and they start sending
 us 300 megabits per second, that's

0:09:40.880000 --> 0:09:41.860000
 going to be a problem.

0:09:41.860000 --> 0:09:45.760000
 Because we've got tens of thousands
 of other customers are also using

0:09:45.760000 --> 0:09:49.720000
 our network. In order to figure out
 how many routers we need to have,

0:09:49.720000 --> 0:09:53.160000
 what type of links we need to have,
 we need to have some predictability

0:09:53.160000 --> 0:09:55.020000
 here as far as what's going
 through our network.

0:09:55.020000 --> 0:09:58.300000
 And if all of our customers are agreeing
 to one thing, but actually doing

0:09:58.300000 --> 0:10:03.040000
 something else, that's going to be
 hard to maintain satisfaction with

0:10:03.040000 --> 0:10:04.700000
 all of our customers.

0:10:04.700000 --> 0:10:09.360000
 So on the ISP side, they will
 do what's called policing.

0:10:09.360000 --> 0:10:14.760000
 In other words, they will measure your
 traffic as it's coming in to them.

0:10:14.760000 --> 0:10:18.040000
 And if you've agreed to a committed
 information rate of, let's say, 50

0:10:18.040000 --> 0:10:21.820000
 megabits per second or whatever it
 is that you want, they will have a

0:10:21.820000 --> 0:10:24.780000
 police or setup that's
 monitoring that rate.

0:10:24.780000 --> 0:10:29.660000
 As long as you send them traffic that
 is up to but not including 50 meg,

0:10:29.660000 --> 0:10:33.140000
 if you stay beneath that
 rate, you're good.

0:10:33.140000 --> 0:10:36.780000
 They'll allow your traffic through,
 they won't change it in any way, and

0:10:36.780000 --> 0:10:38.740000
 you're happy as a clam.

0:10:38.740000 --> 0:10:42.660000
 Now, if you start sending them traffic
 that's above that, what does the

0:10:42.660000 --> 0:10:48.620000
 police or do? Well, that is considered
 what's called non-conforming traffic.

0:10:48.620000 --> 0:10:51.060000
 And policeors can do
 one of two things.

0:10:51.060000 --> 0:10:54.120000
 Usually what they do is
 they drop the traffic.

0:10:54.120000 --> 0:10:58.280000
 They say, hey, look, we're billing you
 $100 a month based on the agreement

0:10:58.280000 --> 0:11:02.360000
 that we would do 50 megabits per second,
 and you're going above that.

0:11:02.360000 --> 0:11:05.420000
 So chances are very likely that their
 police will say, hey, any traffic

0:11:05.420000 --> 0:11:10.200000
 you send us inbound that's more than
 50 meg, kill it, it's done, dead

0:11:10.200000 --> 0:11:15.580000
 on arrival. Now, an alternative action
 they could do is mark down the

0:11:15.580000 --> 0:11:21.060000
 traffic. So for example, if the traffic
 you're sending them has some DSCP,

0:11:21.060000 --> 0:11:26.320000
 some, I should say some non-default
 DSCP or IP precedence value, because

0:11:26.320000 --> 0:11:30.760000
 normally the defaults are zero, and you
 can't mark it down into a negative

0:11:30.760000 --> 0:11:32.740000
 number. That's the
 lowest you can get.

0:11:32.740000 --> 0:11:37.660000
 But if you were sending into the ISP
 traffic with higher than zero, their

0:11:37.660000 --> 0:11:43.180000
 police or could take that traffic above
 the rate, let it go through, but

0:11:43.180000 --> 0:11:45.180000
 just mark it down to zero again.

0:11:45.180000 --> 0:11:47.940000
 Basically negating the priority
 of your traffic.

0:11:47.940000 --> 0:11:49.700000
 Chances are they're probably
 not going to do that.

0:11:49.700000 --> 0:11:53.960000
 They're just going to kill it, but
 a police can do either or of those

0:11:53.960000 --> 0:11:57.160000
 things. Now, what about from you?

0:11:57.160000 --> 0:12:00.180000
 You're the customer on your
 end, like this says here.

0:12:00.180000 --> 0:12:02.380000
 You don't want your
 traffic dropped.

0:12:02.380000 --> 0:12:06.820000
 So let's say that I'm the customer
 router and I'm facing you, you are

0:12:06.820000 --> 0:12:08.680000
 the ISP, I'm the customer.

0:12:08.680000 --> 0:12:12.220000
 Now on the back of my head here, I'm
 getting traffic hitting me at like

0:12:12.220000 --> 0:12:14.400000
 55 megabits per second.

0:12:14.400000 --> 0:12:16.080000
 Trying to go to the internet.

0:12:16.080000 --> 0:12:18.520000
 And I think, uh-oh, okay,
 let's see here.

0:12:18.520000 --> 0:12:22.000000
 We've got a defined, we've got a committed
 information rate of 50 megs.

0:12:22.000000 --> 0:12:25.680000
 That's all I can send to you, but what
 I'm getting inbound is more than

0:12:25.680000 --> 0:12:28.100000
 that. 55 megabits per second.

0:12:28.100000 --> 0:12:32.800000
 Well, certainly I could police as well,
 but from my perspective, all of

0:12:32.800000 --> 0:12:34.340000
 my traffic is important.

0:12:34.340000 --> 0:12:35.580000
 I think it is anyway.

0:12:35.580000 --> 0:12:36.500000
 I think everything's important.

0:12:36.500000 --> 0:12:37.920000
 I don't want to drop anything.

0:12:37.920000 --> 0:12:42.000000
 So on my side, I could implement another
 quality of service feature called

0:12:42.000000 --> 0:12:44.060000
 traffic shaping.

0:12:44.060000 --> 0:12:47.780000
 What traffic shaping says, it says,
 okay, here's what I'm going to do.

0:12:47.780000 --> 0:12:52.900000
 Hopefully, this traffic that's coming
 in at 55 megabits per second is

0:12:52.900000 --> 0:12:56.000000
 just temporary. In other words, there's
 just a little glitch in time where

0:12:56.000000 --> 0:12:59.200000
 it's like, boop, really far, really
 high, and it's not like consistently

0:12:59.200000 --> 0:13:01.320000
 coming in that fast.

0:13:01.320000 --> 0:13:05.720000
 So just for a few moments, a few seconds,
 traffic is hitting me on the

0:13:05.720000 --> 0:13:08.160000
 back of the head at 55
 megabits per second.

0:13:08.160000 --> 0:13:09.300000
 Here's what I can do.

0:13:09.300000 --> 0:13:15.240000
 I will play it out to you, the ISP, at
 our negotiated rate of 50 megabits

0:13:15.240000 --> 0:13:18.840000
 per second. What am I going to
 do with the excess traffic?

0:13:18.840000 --> 0:13:21.600000
 I'll buffer it. I'll hold
 on to it, memory.

0:13:21.600000 --> 0:13:26.440000
 I'll delay it. And then I'll play it
 out when I can, as long as I don't

0:13:26.440000 --> 0:13:28.440000
 go above 50 megabits per second.

0:13:28.440000 --> 0:13:32.440000
 So that's what's called traffic shaping,
 is where it might be coming in

0:13:32.440000 --> 0:13:36.460000
 at this rate, fast, slow, fast, slow,
 but it's going out at a consistent

0:13:36.460000 --> 0:13:39.660000
 rate, and that's what the
 traffic shaper does.

0:13:39.660000 --> 0:13:42.920000
 It says, okay, any excess traffic, I'll
 just hold it back, and then I'll

0:13:42.920000 --> 0:13:47.120000
 play it out again when I can.

0:13:47.120000 --> 0:13:52.280000
 So shaping is typically done on the
 egress interface leading to the ISP.

0:13:52.280000 --> 0:13:55.620000
 So hopefully, now, you have a good
 idea as to what the differences are

0:13:55.620000 --> 0:13:57.960000
 between policing and shaping.

0:13:57.960000 --> 0:13:59.960000
 Both of them set a threshold.

0:13:59.960000 --> 0:14:02.880000
 Both of them set some level, and they
 start monitoring traffic either

0:14:02.880000 --> 0:14:05.160000
 inbound or outbound.

0:14:05.160000 --> 0:14:08.240000
 But what the difference is, is what they
 do to the traffic that goes above

0:14:08.240000 --> 0:14:11.960000
 that threshold. Policing typically
 says, you're dead.

0:14:11.960000 --> 0:14:13.540000
 Any traffic above that level?

0:14:13.540000 --> 0:14:15.340000
 Kill it. Don't deal with it.

0:14:15.340000 --> 0:14:18.700000
 Traffic shaping says, okay, any traffic
 above that level, I'm just going

0:14:18.700000 --> 0:14:23.500000
 to set it aside, and then I'll play
 it out again later when I can.

0:14:23.500000 --> 0:14:28.940000
 Now, this here is an example of a policing
 command, one of several you

0:14:28.940000 --> 0:14:31.160000
 could do in a Cisco router.

0:14:31.160000 --> 0:14:34.460000
 And my purpose here is not to go over
 how to configure this, but I just

0:14:34.460000 --> 0:14:38.320000
 want to draw your attention to a
 little bit about how this works.

0:14:38.320000 --> 0:14:45.880000
 Like, for example, we are matching
 some traffic, and the class of the

0:14:45.880000 --> 0:14:49.720000
 traffic, we've just given
 it a name of press three.

0:14:49.720000 --> 0:14:54.680000
 Probably traffic with an IP precedence
 of three would be my guess, but

0:14:54.680000 --> 0:14:58.600000
 that's just a descriptive name, press
 three, that could be anything.

0:14:58.600000 --> 0:15:02.120000
 But the point is, we're matching
 a class of traffic.

0:15:02.120000 --> 0:15:06.660000
 Now what we're going to do is we're
 going to monitor that traffic with

0:15:06.660000 --> 0:15:11.640000
 a policeor. So the policeor, we say, okay,
 what is the committed information

0:15:11.640000 --> 0:15:22.760000
 rate? Let's say it was 50,000, 50,000
 bits per second, okay, with a peak

0:15:22.760000 --> 0:15:26.900000
 information rate of 75,000.

0:15:26.900000 --> 0:15:29.040000
 All right, so what does this mean?


0:15:29.040000 --> 0:15:33.000000
 That means, okay, so let's
 do it like this.

0:15:33.000000 --> 0:15:40.000000
 So my peak is 75,000 bits per second,
 and I got another level just below

0:15:40.000000 --> 0:15:42.220000
 that, which is 50,000.

0:15:42.220000 --> 0:15:47.200000
 Bits per second, okay, so this is
 what my policeor is monitoring.

0:15:47.200000 --> 0:15:50.160000
 Now here's what the police
 is going to do.

0:15:50.160000 --> 0:15:52.180000
 Conform action, what
 does that mean?

0:15:52.180000 --> 0:15:57.580000
 That means for all this traffic, all
 the traffic that is below 50,000

0:15:57.580000 --> 0:16:00.360000
 bits per second, and that's
 just an example number.

0:16:00.360000 --> 0:16:01.900000
 So what am I going to
 do with that traffic?

0:16:01.900000 --> 0:16:04.040000
 Well, typically you would
 just transmit it.

0:16:04.040000 --> 0:16:07.840000
 Just let it go through, don't touch
 it, don't mess around with it.

0:16:07.840000 --> 0:16:14.280000
 Okay, what about the traffic that
 falls in this area right here?

0:16:14.280000 --> 0:16:20.840000
 So it is exceeding my committed information
 rate, but it hasn't reached

0:16:20.840000 --> 0:16:24.880000
 yet my peak information rate.

0:16:24.880000 --> 0:16:27.040000
 Well, that is my exceed action.

0:16:27.040000 --> 0:16:34.020000
 In this case, I'm saying, okay, let's set
 the precedence to zero and transmit

0:16:34.020000 --> 0:16:39.580000
 it. So between these two lines, we'll
 still let it go out, but if that

0:16:39.580000 --> 0:16:44.480000
 traffic came in with some non-zero precedence
 level or non-zero DSCP level,

0:16:44.480000 --> 0:16:47.740000
 we're just going to set it back to
 the default of zero, and then we're

0:16:47.740000 --> 0:16:49.180000
 going to send it out.

0:16:49.180000 --> 0:16:55.220000
 Now, what about traffic that is
 above this line right here?

0:16:55.220000 --> 0:16:57.900000
 Well, that is above my
 peak information rate.

0:16:57.900000 --> 0:17:02.180000
 Now that's my violate action,
 I am dropping it.

0:17:02.180000 --> 0:17:06.720000
 So you can see this particular police
 is doing two actions in one.

0:17:06.720000 --> 0:17:11.060000
 For some traffic, it's completely
 killing it, it's dropping it.

0:17:11.060000 --> 0:17:16.260000
 For other traffic, it's allowing it
 through, but just marking it down.

0:17:16.260000 --> 0:17:20.140000
 So that's just an example configuration
 of how a policeer could work.

0:17:20.140000 --> 0:17:27.860000
 As far as traffic shaping is concerned,
 okay, so just a little comparison

0:17:27.860000 --> 0:17:34.420000
 here. Like it says here, policeers can
 be applied on ingress or egress.

0:17:34.420000 --> 0:17:38.460000
 In other words, for traffic being received
 or traffic we're about to transmit

0:17:38.460000 --> 0:17:43.580000
 out, usually it's done
 on the ingress.

0:17:43.580000 --> 0:17:50.000000
 And like I said, typically ISPs will
 use policeers and you, the customer,

0:17:50.000000 --> 0:17:53.620000
 will use a shaper on
 the egress side.

0:17:53.620000 --> 0:17:57.080000
 Most Cisco switches do not
 support traffic shaping.

0:17:57.080000 --> 0:17:58.740000
 This is a little caveat
 right there.

0:17:58.740000 --> 0:18:01.680000
 And I want to conclude this video with
 just a sample configuration of

0:18:01.680000 --> 0:18:03.240000
 traffic shaping.

0:18:03.240000 --> 0:18:06.800000
 Like we did traffic policing,
 here is traffic shaping.

0:18:06.800000 --> 0:18:12.280000
 So without going too much into the details
 here, once again we're matching

0:18:12.280000 --> 0:18:13.780000
 on some class of traffic.

0:18:13.780000 --> 0:18:16.700000
 This time it's got a different
 name, press zero.

0:18:16.700000 --> 0:18:20.000000
 And this time we're applying
 traffic shaping to it.

0:18:20.000000 --> 0:18:25.700000
 And we can say shape average
 in bits per second.

0:18:25.700000 --> 0:18:34.580000
 So in my previous example, if the ISP
 and myself had negotiated a committed

0:18:34.580000 --> 0:18:38.960000
 information rate of 50 megabits per
 second, I could say shape average

0:18:38.960000 --> 0:18:43.460000
 50,000, that'd be in
 bits per second.

0:18:43.460000 --> 0:18:46.540000
 Or you see here, I can also add
 these things if I wanted to.

0:18:46.540000 --> 0:18:52.040000
 So instead I could say
 shape average 50m.

0:18:52.040000 --> 0:18:57.780000
 And because this is using the shape command,
 anything that hits this egress

0:18:57.780000 --> 0:19:01.740000
 interface that's faster than
 that will be buffered.

0:19:01.740000 --> 0:19:06.400000
 And this interface will only send out
 traffic at up to 50 megabits per

0:19:06.400000 --> 0:19:13.260000
 second. So that concludes this video
 on the differences between traffic

0:19:13.260000 --> 0:19:15.720000
 shaping, policing and markdown.

0:19:15.720000 --> 0:19:17.020000
 Thank you for watching.

