Designing, Planning, and Prototyping a GCP Network

1. Section 1: Designing, Planning, and Prototyping a GCP Network

Designing the Overall Network Architecture (Chapter 1.1) In this chapter we are going to get into the overall design aspects of the network and those aspects include failure and disaster recovery. That is our first point. high availability DNA strategy, business requirements, how to map business requirements with SLAs or SLOs, or network meeting your SLAs, load balancing, optimising quotas, hybridnetwork identity and various security aspects, various clouds and micro segmentation Let's go ahead and start with a failover. So what does failure or disaster recovery failure mean? So if one of your services fails, what do you want to actually do with that? How can you actually make sure that the service has started? Your business is not much impacted, or not impacted at all, because of your failure. If it is an application server, whether it is a database server, whether it is a web server, your database, as an example, is filled, then how do you start a new database or your cluster field, what is that you want to actually do or take it forward? This has got a complete strategy if something goes wrong in your system. You can still continue serving your business and that is what it is called a failure or disaster recovery strategy. We are going to focus much more on the context of the cloud here and in the Google cloud. You have your customers here, whether it is a website or mobile applications. They are accessing your services from the back end, and those backends are accessed via a load balancer. So this is your load balancer here, right? And then the load balancer is toggling between sending the request based on the intelligent routing. And we are going to see in detail how the load balancer actually does this to either zone one or zone two, or region one or region two, depending on where the customer is or the consumer is. So, in the Googlecontext, these zones or regions are typically connected with their own private fibre optic cables that span the globe, okay, and all of these data centres serving your data are actually connected to your load balancer, which is also globally located. It does not actually sit within one particular region or zone. So it is a global service. It is a managed service from Google itself. What happens is, as an example, one of your own zones or the region itself. The region is a logical grouping of the geographical locations you can think of. There are multiple zones inside the region. So either your region actually goes out or a particular zone goes out of service, that's the physical location. Then your traffic is still served by another zone,another region, and that is where you can think of how you can manage the disaster recovery. But ultimately, you need to have a plan. You need to plan your services hosted in multiple zones or multiple regions so that you can still go ahead and serve your customers. And this is, you can think of a failure strategy or just a recovery strategy, but you can think of different aspects. How can you make sure that if something goes down in zone one or region one, these instances are not accessible, right? How can you still serve customers without impacting them? As an example, if there are sensations which are currently serving a customer and you have astate somewhere in your virtual machine, right? What happens if that virtual machine or one of the zones fails? right? So you'll need a strategy to keep that particular state in a persistent store somewhere, and you can use data storage or any other persistent mechanism to do so so that your customer doesn't notice. And this is very important. So for all the different levels of your disaster recovery or failure, you need to have a different strategy altogether. But this is in a high level context. You have a load balancer. The load balancer is responsible for ensuring that only healthy backend instances are served when requests are sent to them. And this is a part of the load balancer configuration. You configure health check also inside it and that's how the request is routed to the thehilti instance in the back end. So you need to have a thought and an example, such as what you want to ensure if your cloud SQL goes down, right? So one of the primary instances or master goes down in cloud scale, like MySkill primary or masterinstance goes down. What do you do? right? So in cloud scale, you have HR configuration. You can have high availability configuration in different zones altogether, so that if the master goes out or the complete zone disappears from the network, you can still toggle to the Ha instance. And this is done automatically by the Google Cloud platform. You can also have read Replica and promote reading Replica as a master, which is another way to go about it. But there are multiple ways in which you can make sure that the service is available. If, for example, back end processing or the virtual machine fails, then what is your strategy? Again, it depends on what that particular service is. As an example, if your virtual machine goes down, whatever processing that was done by that virtual machine, if that is very important to your customer and for your processing needs,then you need to store the data in either the cloud storage or any persistent storage. As an example, you want to save a game state. If your customers are playing a game and you are maintaining the game state, then you want to store that particular game state in some persistent store like cloud storage, right? And that's how if another machine comes up as a replacement for that particular virtual machine, then that machine should be able to serve the customer looking at those games. So there are different aspects of failure and disaster recovery which you can implement based on need and based on the type of service for which you are creating a failure or disaster recovery. Persistence is definitely very important. You want to store that particular state somewhere so that if another instance comes up, it can be actually processed from that particular state and onwards to your customers. So, in most cases, you'll need a disaster recovery or failure document in addition to the configurations, right? If at all, something goes wrong in your system, you need to make sure that everything is intact and your customers are still served based on your disaster recovery plan. If zone one failed and you have zone two, or one of the instances in zone one failed, and you have additional instances created out of it. We'll go over how to failover different instances in detail, both theoretically and practically. But that is a part of the network and you should be concerned about it, and you should make sure that you have written documents and you understand what services we are using. It's from the Google Cloud and you have a plan for disaster recovery as well. That's it, guys. For disaster recovery and failure, we are going to see SLA and SLA, which is like how you measure whether you're meeting business requirements. This is the agreement and how you're actually complying with the agreement. Thanks.

2. 1.1 Designing the overall network architecture.

So let's go and understand what SLA and Slo are because those are very important. And this is the contract you can think of between your business and it as a service provider. OK? So SLA understands the business requirements for SLN. This is what it means: those expectations which your business has from it, and this is a complete agreement about your it services which are consumed by your business. As an example, you are providing or hosting your website as a business and that particular website is sold to the customer, right? And how much is the load that website can take? How much time should it be available if something goes wrong? Within what time frame should it be available for customers to serve it? All of that is contained in your SLA document, which is the written document and agreeddocument between you and your business. So, Slo is a typical measure, and you can consider its characteristics. So this is very much in the context of business and this is purely in the context of measurability. How do you measure SLA, right? And that is in the form of availability; how much your service is available, like 24 by 700% available. In the case of websites, that is definitely a requirement, but there is no service which is like 100% available. However, you must ensure that when your customer is not using the service and the service is shut down in accordance with the agreement with the business, you will be fine if at all. As an example, if you are serving say, Walmart or any other retail agency local to your area, and if that shop is actually closed at 06:00 p.m. or 09:00 p.m. in the middle of the night and opens at 6:00 a.m. In the morning from nine to 6:00 p.m., even if your services go down when no one is using them,then that is fine, that is absolutely okay, right? However, you must ensure that your services are available to customers from the time the shops open until they close, just like the other shops. So that is the measurability of your availability. Throughput throughput means if there are two shops accessing or 1000 shops accessing,your service should not go down, right? And that is your throughput. And again, this is measurability. You need to have a measure of the throughput which you are expecting and you can measure it through different parameters or different matrices, frequency, response time, and so many other measurements which you can define under Slo. And this is a very important document. Again, it has got matrices derived and you can use them. It is important to measure and ultimately report it. As an example, you can create a differentlog or trace for the response time. Whether you are still able to say the shop is able to see your service or the website, but at the same time, the response is too slow that the customers are queuing up at the counter, right to pay the bills. That is not acceptable. So you need to make sure that you have got an acceptable level of response time from your website. So all of that is measurability and slow means Actually, you can measure all those numbers in mattresses in the form of logs or any tools. There are hundreds of tools that are available that you can use it.The Google cloud platform also gives you some tools that you can use to measure all these numbers. So that is a part of all those attributes, typically. All of these attributes speak about the quality of service and the quality of your software, right? And SL is usually a major of Q, which is what SL is for, right? As an example here, right, 85% of your TCPreplies should be within 1.5 seconds of receiving a request. And this is your turnaround time for any request. Whether it is submitting a form or refreshing the page on your website or getting the data from the back end to your mobile application, It could be anything. What it says is 85%. So this is not 100%. It depends on the peak load, or this depends on one particular time. It can spike up. There are so many customers hitting your back-end services or websites because there are some promotions going on. It can actually go down. It can go down. Services may be impacted. But with this definition of SLA, what it defines is 85%. At least 85% of the TCP replies within 1.5 seconds of receiving the request. That was Slo and Slo guys. If you have any questions on that, let me know. We are going to talk about high availability. We already saw that in failure and disaster recovery. But we'll just do a deep dive into it. Thank you.

3. Designing the overall network - Options for HA

options for high availability. We already saw load balancing as an option for high availability. When one of the zone R regions fails due to a problem or one of the virtual machines, which acts as a back-end service, fails, the load balancer intelligently detects the failing VMs, zones, or regions and routes traffic to the healthy ones. The additional thought around the high availability is an additional VPN connection, and this is for the hybrid connectivity. So, if you have a data centre and one VPN connection to the Google Cloudplatform, you may want to or consider additional VPN connections if something goes wrong with the first VPN connection, and that is the additional connection or additional High Availability option. If at all you are using interconnection lines, you will definitely want to add an additional interconnection line, which is a dedicated MPLS line from your data centerto the Google Cloud platform for availability. This was purely in the context of the network. If you look at the database as such,then you can have the backup and distort strategy configured for your databases, whether it is Cloud SQL, whether it is PostgreSQL, or whether it is Cloud Spanner. Additionally, you can have a replica. Read Replica serves multiple purposes, and we are going to get into details of that which you can think of for the database. But the replica actually serves a purpose. If your master instance or primary instance is receiving a high volume of read traffic, you should route it to the Read Replica. So you create a Read Replica, and you route read traffic to the Read Replica. You can create multiple Read Replicas, but if something happens to your master node, then you can actually promote Read Replica to take the right traffic as well. You can have a high availability configuration for your Cloud SQL databases. For example, you could have a standby master running continuously in different zones or regions. Then, if the first master or the master fails, cloud scale automatically switches to the backup or Ha instance. specific to the applications. I would recommend that you begin thinking about persistent storage for your cachedata or persistent data, whether it is Cloud DataStore or cloud storage for your game state or any session management or data that is in the transientstate, right, and is dependent on your virtual machine or any instance in the back end. So try to use persistent storage for your session data that is at a high level. What is your high availability policy? Whenever you create high availability, the policies need to be written, and then those documents you want to actually test it, right? Test means you probably want to delete some of the VM instances on the back end and then you want to see whether the load balancer is distributing your traffic according to your policy or not. all of that. This should be testable enough so that if something happens, you have a plan and you have a plan which is like working in actual scenarios. That's it for the high availability. This is just a theory, but if you have any questions, please let me know. Otherwise, you can move to the next lecture, which is DNS strategy. DNS strategy We'll cover more of the DNS and DNSstrategy in the demo itself or the theory itself, in detail but at a high level. The Cloud DNS that Google Cloud Platform provides has a 100% uptime SLA and that's the only thing that you need to consider because it is a 100% uptime SLA. You do not need to worry about whether some of the hosted zones are going down or something because it is a highly available managed service provided to you. You cannot connect on-premises services using Cloud DNS because it is a Google Cloud platform service and you will not be able to connect it to any on-premises or outside service that is not on the Google Cloud platform. This may change in a short period of time,but you can keep that in mind.

4. Designing the overall network - Options for Load Balancer

load balancer. A load balancer is a service which is available right now on premises as well as your own service as well. So, what load balancer actually does load balancer actually take your traffic or your client's request? It accepts requests from mobile or desktop clients, websites, or any other TCP or HTTP client and distributes them in the back end instances. So you can think of it as distributing requests based on the type of request, the location, or any number of parameters, right? So that's the load balancer. The load balancer actually takes the request from the client and sends it to the back end to get that request fulfilled. There are multiple types of load balancers which are available on the Google Cloud platform and I have divided them into two different categories. One is the global load balancer and the second one is the regional load balancers. The global load balancer has got HTTP load balancing, SSL proxy load balancing, and TCP proxy load balancing. These are the global types of load balancers. There are regional load balancers. One is the regionalexternal load balancer, and they also call it a network load balancer in plain form. Rather than saying regional external load balancer, there is another load balancer which is called a regional internal load balancer or simply as an internal load balancer only. We are going to see in detail about this load balancer and even the demo as well, but these are the high-level load balancers which are available for you to use it.They have given us a free decision as well, in which case you want to use a specific load balancer or the other ones. As an example here, if your requirement is an external requirement,whether the client is IPV six or IPV four, traffic is coming from the internet to your GCP connections and depends on the type of service, whether it is HTTP or HTTPS, you choose HTTP or HTTPS load balancing. If your traffic is TCP traffic, that's where you need to check whether it has got SSL in it. If there is an SSL offload, then you can probably use an SSL proxy load balancer. If there is no SSL, then you can just go ahead and use the TCP proxy load balancer. If the traffic is UDP traffic, then you want to check whether the client has got a global loadbalancer requirement and the traffic is IPP six. If it is IPV six traffic, then there are currently no services which you can use from the Google Cloud platform. But if it is not IPV six, but IPV four traffic, then you can use network TCP or UDP load balancing. If your requests are internal to your consumer, internal to your GCP like any other VM, then you should be using internal load balancing. Whether it is GCP or UDP, it's internal load balancing. Here are some of the port numbers based on which you can use them for different types of load balancers, and we are going to get into details of that at a high level. If you look at the architecture diagram, yourglobal load balancer actually takes the request from the public internet, whether it is the user's residence in the US, Europe, or Asia, right? So that's your global load balancer, and this is your requirement. Typically, if you look at the internal, if your oneVM is calling another VM or another service inside, like your web tier is calling the application tier in the back end, you can use an internal load balancer. This is the high-level representation of how you can use load balancers for your high-level ability or anything you can think of. Let's go ahead and look at optimising it for latency, and we are going to see three different options that are available in the next lecture. Thank you.

ExamSnap's Google Professional Cloud Network Engineer Practice Test Questions and Exam Dumps, study guide, and video training course are complicated in premium bundle. The Exam Updated are monitored by Industry Leading IT Trainers with over 15 years of experience, Google Professional Cloud Network Engineer Exam Dumps and Practice Test Questions cover all the Exam Objectives to make sure you pass your exam easily.