网络课程笔记

00:00 Okay.

00:00 So with respect to the network performance, there are a lot of things that we saw.

00:04 And let's try to remember what you need to know for your exam.

00:09 So as an exam essential understand that the MTU values first of all what is MTU.

00:15 And then the various MTU values within the VPC.

00:19 And typically when the traffic flows out of VPC.

00:21 So if the traffic stays within the VPC MTU size of 9001 byte, which is also called jumbo frames, are

00:28 supported or otherwise if the traffic is going over the internet or if it is over the site to site VPN

00:34 connection, which terminates on the Virtual Private Gateway or the Transit gateway, the jumbo frame

00:40 is not supported and the MTU size is limited to 1500 bytes.

00:45 Now, when it comes to the VPC peering connection, the Inter-region VPC peering supports jumbo frame

00:50 of size 9001 bytes, but for the Inter-region VPC peering it is 8500 bytes.

00:58 Now, understand that when the packet per second is a bottleneck, that means your CPU is having the

01:04 limited capacity and it can't process more packet per second.

01:08 In that case, increasing the MTU would give the better throughput because you are now processing the

01:14 larger packet at once.

01:16 Right.

01:16 So that's one optimization technique that you can apply now while using the enhanced networking and

01:22 placement group.

01:23 Ultimately, it lowers the latency between the EC2 instances and the hypervisor.

01:27 But if you are using Dfdc in that case it improves the packet processing at the operating system level.

01:34 So understand the difference between enhanced networking and the DP DQ.

01:39 So whenever dp DQ understand that its OS level optimization.

01:44 Okay.

01:44 Moving on.

01:45 If you talk about the network bandwidth that EC2 can get, the maximum network bandwidth it gets, it

01:51 depends on a lot of factors, right.

01:53 Those factors could be instance family whether it's M family, C family.

01:57 So every instance family would have different network bandwidths as well as the size of the instance.

02:03 That means what CPU capacity it has, what memory it has, how much disk input output IO it has.

02:11 Right.

02:12 Also whether the instance supports enhanced networking on not all this factor will ultimately decide

02:18 the maximum bandwidth that EC2 instance can get.

02:21 But in general, if you want to look at the bandwidth, then you can visit the AWS website documentation.

02:27 And then there you'll see a particular instance type.

02:30 What is the maximum bandwidth it supports.

02:33 Now also understand that the bandwidth when we talk about it's an aggregated bandwidth across multiple

02:38 flows.

02:39 Now one flow is one, uh, connection between EC2 and the other end, right, which is a five tuple,

02:47 typically protocol source port or destination port source IP and destination IP.

02:53 So when we say maximum bandwidth, we are not talking about just one single connection between the two

02:58 machines, right.

02:59 We are talking about, uh, aggregate bandwidth using multiple flows.

03:04 So uh, bandwidth of uh, aggregate flow within the region would depend on the EC2 instance.

03:12 Maximum bandwidth.

03:13 So, as I said, uh, EC2 would provide different bandwidth depend on EC2 instance sizes.

03:19 Now if you go to the AWS documentation.

03:22 So here is the documentation.

03:23 If you go to this page I have given the link as well.

03:26 It states that within the region you can get the full network bandwidth of EC2 instance.

03:31 And if you just click here and go to general purpose instances.

03:36 In that case you can see here for example, the there are different instance types first.

03:43 And then if you go to the network performance you can see T2 instances.

03:47 Maximum network performance is one GB.

03:50 But if you go down and see other instances like M6 this is M6 a it provides 12.5 GB and in some cases

03:59 you get up to 100 GB maximum bandwidth across multiple flows.

04:03 Again, remember that right with any you have to use any in that case.

04:07 Now some instances support Intel virtual functions, some support Ina.

04:12 And accordingly the bandwidth is decided.

04:16 So that means there is no straightforward answer.

04:18 What is the maximum bandwidth for a given instance?

04:21 It depends on a lot of factors.

04:23 So always visit this page and you will see the maximum bandwidth.

04:26 Similarly, there are other instance types as well, and you can go and see their bandwidth as well.

04:31 Okay.

04:32 Now if the traffic is within the AWS region, then you can get the maximum bandwidth of 100 Gbps with

04:39 the Ina that we learned about.

04:41 Also, if the S3 bucket is also within the same region, then also you can get up to 100 Gbps of bandwidth

04:48 between S2 and S3, again using the multiple flows.

04:52 Now outside of the region to the internet or direct connect, the maximum bandwidth you can get is maximum

04:58 50% of the available network bandwidth.

05:01 Now that will also depend on the generation of the instance as well as the size of the instance.

05:06 The size of the instance has to be greater than or equal to 32 vCPUs, and it should be current generation.

05:12 Then you can get up to 50%.

05:14 That means, uh, the page that we just saw, if it says, say, 20 Gbps maximum bandwidth, but if

05:20 the traffic is going outside of the region to the internet or direct connect, then you can get up to

05:25 ten Gbps as a maximum bandwidth.

05:27 However, if the instance is not the current generation or the size is less than 32 week CPU, then

05:35 the maximum bandwidth you get is five Gbps only.

05:38 So remember that.

05:40 And now the last thing you need to remember is that the single flow bandwidth is typically limited to

05:47 five Gbps.

05:48 If the traffic is between two EC2 instances, which are across VPCs or within the same VPC, and if

05:55 the traffic is going out to the internet or anywhere, then it is limited to five Gbps.

06:00 But if the instances are within the same cluster placement group now I hope you know what is the placement

06:06 group.

06:07 Then you can get the ten Gbps in a single flow.

06:10 Now this is typically required for high performance computing kind of workloads.

06:15 Like if you are running some, you know, CAD applications or simulations, then you need that level

06:22 of bandwidth.

06:22 And in that case you will use the cluster placement group.

06:26 Now in that case per flow you can get ten Gbps.

06:29 And the aggregate bandwidth again is limited to 100 Gbps with the Ina that we already talked about.

06:35 So remember these things okay.

06:37 so that covers the important points.

06:39 What you need to remember for exam.

06:41 So if I want to summarize this again, understand that the maximum network bandwidth between EC2 instances

06:47 and other networking endpoint is 100 Gbps that you get with the Ina.

06:54 And uh, also understand that within the region you can get full bandwidth of EC2 instance and per flow

07:02 bandwidth is typically five Gbps.

07:04 But then within the cluster placement group you can get up to ten Gbps as well.

07:08 Now outside the AWS region, if the traffic is going to, then the maximum bandwidth would be 50% of

07:14 the maximum available bandwidth for that instance.

07:18 And uh, otherwise, if the instance sizes are lower than 32 vCPUs, then the maximum bandwidth is five

07:25 Gbps.

07:26 Now when we talk about maximum bandwidth within the region, it could be between Ec2's.

07:30 Or it could also be between S2 and S3 as well.

07:33 That means you can get 100 Gbps bandwidth with multiple flows between S2 and S3 as well.

07:41 So that's it about this lecture.

07:43 Thanks for watching.