1

00:00:00,240 --> 00:00:01,360

Hello and welcome back.



2

00:00:01,380 --> 00:00:07,620

In this video, we are going to discuss about throughput in the last video we have discussed about latency



3

00:00:07,980 --> 00:00:14,300

and for latency, we said that it is how long it takes for your data to traverse a system.



4

00:00:14,580 --> 00:00:18,790

So we take consideration of the time, that is latency.



5

00:00:18,930 --> 00:00:25,650

But when we talk about throughput, it is how much work a machine can perform in a given period of time.



6

00:00:25,650 --> 00:00:28,200

So throughput is focused on the work side.



7

00:00:28,590 --> 00:00:35,160

And when we talk about throughput, we are typically referring to how much data can be transferred from



8

00:00:35,160 --> 00:00:39,990

one point of a system to another point of a system in a given amount of time.



9

00:00:40,380 --> 00:00:47,100

Now typically we measure our throughput in gigabits per second or kilobits per second or megabits



10

00:00:47,100 --> 00:00:47,850

per second.



11

00:00:48,180 --> 00:00:55,380

For an example, let's say you have a network of one gigabits per second, which literally means that



12

00:00:55,380 --> 00:01:02,450

this network can support one gigabit per second, or you can say that billion bits per second.



13

00:01:02,820 --> 00:01:06,410

Now to understand this, we can take an example of an image.



14

00:01:06,780 --> 00:01:08,700

So look over this image here.



15

00:01:08,700 --> 00:01:14,430

We have four clients issuing requests to this server, but in actual time, it could be thousands of



16

00:01:14,430 --> 00:01:17,380

client all issuing requests to this server.



17

00:01:17,670 --> 00:01:20,760

So here in this case, what is going to be the throughput?



18

00:01:21,000 --> 00:01:28,080

So throughput in this case would be how many request this server can handle in a given amount of time



19

00:01:28,080 --> 00:01:30,260

or in this case, a second.



20

00:01:30,720 --> 00:01:37,500

And if I say more specifically and if I reduce these requests to bit level, then how many bits this



21

00:01:37,500 --> 00:01:41,540

server can handle or let through per second.



22

00:01:42,090 --> 00:01:43,800

So we will notice one thing.



23

00:01:44,070 --> 00:01:49,920

There will always going to be a bottleneck where you can fit so many bits during a given amount of time



24

00:01:49,920 --> 00:01:56,480

because every network has its peak capacity, you cannot fit more bits than supported.



25

00:01:56,790 --> 00:01:58,610

So to repeat that, what is throughput?



26

00:01:58,890 --> 00:02:04,460

So how many bits your server can handle or you can fit in your server in a given amount of time.



27

00:02:04,680 --> 00:02:06,300

So that is going to be your throughput.



28

00:02:06,600 --> 00:02:11,410

Not the question that is asked mostly is how to increase your throughput.



29

00:02:12,210 --> 00:02:15,380

So the basic answer is you pay for it.



30

00:02:15,990 --> 00:02:21,360

Now, this is a very basic answer and not appropriate for all cases.



31

00:02:21,840 --> 00:02:22,980

So think of an example.



32

00:02:22,980 --> 00:02:31,080

Let's say your servers are hosted on AWS, or in theory you pay AWS to increase your throughput.



33

00:02:31,560 --> 00:02:32,760

And there is one reason.



34

00:02:32,760 --> 00:02:34,040

That's why I said it.



35

00:02:34,050 --> 00:02:41,460

It's a very basic thing to do and maybe not appropriate in every scenario, because just increasing your



36

00:02:41,460 --> 00:02:45,300

throughput is not going to fix a potential problem in your system.



37

00:02:45,690 --> 00:02:48,120

So let's take an example of Facebook search.



38

00:02:48,270 --> 00:02:52,530

Perhaps you will serve millions of requests per second.



39

00:02:52,830 --> 00:02:58,380

And in this case, just trying to blindly increase your throughput will not make any sense.



40

00:02:58,620 --> 00:03:05,010

And perhaps you will face the bottleneck and your server start to crash because they are not able to



41

00:03:05,040 --> 00:03:06,530

serve your request well.



42

00:03:06,750 --> 00:03:08,100

So what is the solution?



43

00:03:08,340 --> 00:03:12,660

So the better way to fix this is to have multiple servers.



44

00:03:13,290 --> 00:03:18,330

So in this case, what you are seeing, we have three servers instead of one, and the requests are



45

00:03:18,330 --> 00:03:20,460

divided between multiple servers.



46

00:03:20,790 --> 00:03:27,840

Now, this is better because each server will serve request according to their capacity, not more than



47

00:03:27,840 --> 00:03:28,200

that.



48

00:03:28,590 --> 00:03:31,530

And if you have more load, you can increase the server.



49

00:03:31,800 --> 00:03:33,960

But that's not how it is implemented.



50

00:03:34,140 --> 00:03:38,130

We need to implement some more technologies behind it to get it right.



51

00:03:38,280 --> 00:03:42,720

And that's why we have something known as load balancing and stuff like that.



52

00:03:43,020 --> 00:03:46,190

But we are going to study that in later section.



53

00:03:46,650 --> 00:03:51,750

So in this video, what we have learned, so we have learned about throughput and we have seen that



54

00:03:51,750 --> 00:03:55,440

how bottleneck can happen and how we can get through it.



55

00:03:55,830 --> 00:03:57,180

So that's it for this video.



56

00:03:57,180 --> 00:03:58,460

I hope you enjoyed it.



57

00:03:58,860 --> 00:03:59,580

See you soon.



58

00:03:59,580 --> 00:04:00,090

Goodbye.