网络课程笔记

00:00 Okay, so in this lecture, we are going to talk about

00:02 a newer kind of service for AWS called

00:04 the AWS Global Accelerator.

00:05 But before we do so,

00:06 I want to introduce the problem to you

00:08 that we're trying to solve and how we're going to solve it.

00:10 So, say you have deployed and application,

00:13 and it's global and you've global users

00:15 who want to access it directly.

00:17 But our application is only deployed in one region.

00:19 So, for example here in India,

00:21 I have deployed a public application advancer.

00:24 But my users are all over the world.

00:26 They're in America, in Europe, in Australia.

00:28 And so as they access the application,

00:30 they want to go over the public internet.

00:32 And that can add a lot of latency

00:34 due to many hops through the routers. Okay?

00:36 So, I'm obviously overplaying this,

00:38 so you can see that in America,

00:40 there's five hops of routers or servers

00:42 before we get to the public ALB in India.

00:45 And that's because it goes over the public internet.

00:47 Same for Australia, many hops.

00:48 And same for Europe, many hops.

00:50 So you can see these hops I introduce a bit of risk

00:53 connection can get lost,

00:55 they also add a little bit of latency,

00:56 and they're not as direct as possible

00:58 into our amazon infrastructure.

01:01 So what we want to do is to wish to go as fast as possible

01:04 through as soon as possible the AWS US network

01:07 to minimize latency. And so before we do so

01:10 and use the global accelerator for this,

01:12 I want to introduce to you another concept

01:14 which is Unicast and Anycast IP.

01:17 So first, Unicast IP is what we know.

01:20 One server holds one IP address.

01:22 So our clients when they talk to two server

01:24 one has starting IP address is the left hand side

01:26 number 12 and the other one is 98.

01:28 Then if you refer to the IP address that begin with 12

01:30 we will be sent to the server on the left hand side.

01:33 And for the other one if you use the other IP

01:35 then we will go to the server on the right hand side.

01:38 This we know it makes a lot sense.

01:40 But for Anycast IP, is a little different.

01:43 All servers will hold the same IP address

01:45 and the client will be routed to the nearest one.

01:49 So this is contra-intuitive but this is how it works.

01:51 So our client has two servers

01:53 and these two servers as you can see in the bottom,

01:55 they have the same IP.

01:57 But our client when it tries to connect to this Anycast IP

02:00 will be sent to the server that is the closest to itself.

02:04 And as surprising as this is, this is how it works.

02:07 So Global Accelerator uses that Anycast IP concept to work.

02:11 So how does it work?

02:12 We're able to leverage the AWS internal global network

02:16 to route to our application.

02:18 So we have the same idea,

02:19 we want to route to India

02:21 and we have users all around the globe.

02:23 And is going to happen is that instead of sending it

02:26 through the public internet in America,

02:28 it's going to come to the closest edge location.

02:31 And from edge location,

02:32 it's going to go all the way straight to our ALB

02:35 through the internal AWS network.

02:38 Same for Australia, so it goes to closest edge location

02:41 near to Australia

02:42 and then it goes over the private AWS network

02:44 to get to the ALB and same for Europe.

02:48 So the idea with this is that,

02:49 we're going to use an Anycast IP

02:51 and there's actually going to be two of those

02:52 that are going to be created for your application

02:54 and they're global.

02:55 And the Anycast IP will send the traffic directly

02:58 to the closest edge location of your users.

03:01 And this is the power of Anycast IP.

03:03 So the edge location will then send the traffic

03:05 to you application, (mumbles)

03:06 For example,

03:07 through the private AWS network which is much more stable,

03:10 has less latency and so on.

03:12 So the Global Accelerator is absolutely unique

03:14 because it really allows to give two static IP addresses

03:17 all around the globe for the users

03:19 for whatever application you may have.

03:21 And right now I'm showing one ALB in one region

03:23 but it could be global as well

03:25 it could be multiple ALBs in multiple regions.

03:27 So it's a game changer to me.

03:29 So, what does it work with?

03:30 It works with Elastic IP, EC2 instances,

03:33 Application Load Balancer, Network Load Balancer

03:35 and they can either be public or private.

03:38 There is consistent performance

03:39 because we go over the network

03:41 so we have intelligent routing

03:42 to the lowest latency edge location

03:44 and we'll have fast regional failover

03:46 in case anything goes wrong.

03:47 There's no issue with client cache

03:48 because the client doesn't cache anything

03:50 the IP, the two Anycast IP we're using don't change.

03:54 It's internal AWS network to go after the edge location

03:58 so it's perfect.

03:59 And then we have health checks.

04:00 So the Global Accelerator will perform a health check

04:02 on your application and then make sure

04:04 that the application is global

04:06 and if health check fails for one ALB and one region

04:09 then there is automated failover in less than one minute

04:12 to a healthy end port.

04:13 So this is really, really great.

04:15 And it's amazing for disaster recovery,

04:17 thanks, to the health checks.

04:18 Now for security,

04:19 it's appropriately secured

04:20 because we only have two external IP

04:22 that needs to be whitelisted by your clients.

04:24 And you get DDoS protection automatically

04:26 through the Global Accelerator,

04:28 thanks to something called

04:28 AWS Shield that we'll see in the security session.

04:31 So this is great.

04:32 So you're really asking me what is the difference

04:34 between Global Accelerator and CloudFront?

04:36 And hopefully you already see the difference,

04:37 otherwise I failed but I want to outline these differences

04:40 very, very clearly.

04:41 So, Global Accelerator and CloudFront

04:44 they both use the same global network

04:47 and they will both use edge locations all around the globe

04:50 that AWS has created.

04:52 They both integrate with Shield for DDoS protection

04:55 so they both receive the same thing

04:56 but now their difference.

04:57 CloudFront is going to improve the performance

05:00 for both cacheable content such as images and video,

05:03 and also dynamic content such as,

05:05 API acceleration and dynamic site delivery.

05:07 And the content is going to be served

05:09 from the edge locations.

05:11 So once in a while the edge locations

05:13 are going to fetch the content from the origin

05:15 but most of the time hopefully CloudFront

05:18 will deliver cache content from the edges.

05:21 So here the users are getting content from the edges.

05:24 Whereas, for Global Accelerator,

05:25 it improves the performance of the wide range of application

05:28 over TCP or UDP.

05:29 But the packets are being proxyeed from the edge locations

05:33 to the applications running in two one or more AWS regions.

05:37 So in that case, all the request still make it

05:40 to our application end.

05:41 There is no caching available.

05:43 So it's a really good fit if you have a non-HTTP uses cases,

05:46 such as gaming, IoT or Voice over IP

05:49 or it's also really good if you have and HTTP use case

05:51 that require a static IP addresses globally.

05:54 Okay?

05:55 Or in case you need to have deterministic

05:57 and fast regional failover.

05:59 So, Global Accelerator is a newer kind of service

06:01 and it will appear in your exam.

06:03 I hope you like this lecture.

06:04 I will see you in the next lecture for some hands on.