Oracle GoldenGate 26ai Training – Day 2 (Aug 2026) Masterclass

Hi everyone, welcome to our day 2 training session of world's first weekday live online training of GoldenGate26 AI. My name is Ashish, Ashish Agarwal and I'll be delivering this lecture to you. So before commencing the session just a quick round of update regarding my youtube channel which is youtube.com slash ashishagarwal underscore gg. So if you haven't visited the channel I would request you to do visit the channel and do like share and subscribe the channel is dedicated to Oracle GoldenGate related videos. Thank you. So in today's session which is day 2 we are going to discuss about Oracle GoldenGate architecture's classic and microservices how process data flow happens in GoldenGate 26 AI and difference between GoldenGate 21c and GoldenGate 26 AI. We are going to discuss about that. So first of all what is classic architecture and how it is different from microservices. So many of you or some of you have already attended the classic architecture training and they have already worked on classic architecture. So you might be aware but for those who are directly starting with microservices. So just to give you an overview of classic so overall there is not much difference but yes microservices is REST API based architecture which one of the feature is GUI. So let us see what how does classic architecture looks like and in comparison to microservices what is the difference. So in classic architecture so what GoldenGate does is it replicates the data from source to target. So you have a source database and a target database. So source database can be any Oracle or any non-Oracle supported database. Same way your target database can be Oracle or any non-Oracle supported database. So each database has its own transaction locks. So in case of Oracle those transaction locks are called as redo locks or backup of redo locks which are called as archive locks. So what happens is you configure GoldenGate extract process which is also known as capture. So the extract process receives the changes from the transaction locks and remember one thing it captures only committed transactions. So in the previous session we discussed that the transaction locks the redo locks and archive locks contains both committed and uncommitted transactions. The extract process of GoldenGate. So remember one thing guys capture and extract these are the terms used interchangeably. Same way going forward you will also notice that replicate and delivery they are also used interchangeably. So when we say capture or extract it means same when we say replicate and delivery it also means same. So the extract process in the database it receives the changes from the transaction locks of the database. Once it receives those changes it writes them to the trail file on local machine. These trail file resides on the operating system. Once the data is returned into the local trail file. So trail file residing on source machine are called as local trail file. Once the data is available into the local trail file the data pump process reads the data from the local trail sends it over TCP IP network and writes it to the remote trail file. So here one of the background process runs which is called as collector process. So there is a collector process which runs I'm naming it as CUL. So what happens is pump process when it sends the data to target the collector process on target receives it and then it writes it to the remote trail file on target. So the trail file on local machine are called as local trail file. The trail file on remote machine or target machine are called as remote trail file. So the extract process receives the changes captures only committed transaction this is really very important. So in these trail files you have only committed transactions you don't have any uncommitted transactions. Once the data is available into the local trail file the data pump process reads the data from the local trail sends it over TCP IP network and writes it to the remote trail on the target side. Once the data is available into the remote trail on target once the data is available into the remote trail on target the replicate process reads the data from remote trail and apply it to the target database. Okay so overall architectures look like is extract process receives the changes captures only committed transactions writes it to local trail pump process reads the data from the local trail sends it over TCP IP network and writes it to remote trail. Once the data is available into the remote trail file the replicate process reads the data from remote trail and apply it to the target database. This is how your data flow happens in golden gate classic architecture. So golden gate also supports bi-directional. So in bi-directional both the sites access source and both the sites access target. So in bi-directional what happens is the data from site A to site B gets replicated in the same way whatever application transaction application transaction happens on site B they will be replicated to site A database. Okay so in bi-directional what happens is now the extract process configured over here receives the changes from the site B database. Once it receives the changes it writes them to the local trail. Once the data is available into the local trail the data pump process configured on site A will the data pump process configured on site A will send the data over TCP IP network and writes it to the remote trail. So here also you will have collector process running separate collector process because pump will establish the connectivity with the collector once the collector process starts then it writes data to remote trail. Once the data is available into the remote trail file the replicate process applies the data to the target database. The replicate process applies the data to the target database. So now in bi-directional what happens is your replication happens from site A to site B and site B to site A both sites access source and both sites access target in bi-directional. Okay so bi-directional is nothing but two-way unidirectional with some considerations and we do cover this bi-directional in our topic it's a separate topic which we cover it. Okay so this is the classic architecture. So classic architecture you have all the processes the the crux of microservices also remains same you will see over there extract trail file then pump is replaced with distribution server the collector process is replaced with receiver server and then replicate process applies. So the crux remains same but there things runs as a service. So in classic architecture all the things are running as process. Now if you go to 26AI so 26AI is nothing but microservices. Okay so GoldenGate 26AI architecture overview it is composed of capture trail files and apply processes. It supports microservices architecture only. So GoldenGate 26AI supports microservices architecture only. It is AI powered automated tuning for optimal replication performance and it is integrated with oracle database 26c and autonomous database for cloud native replication. So it is integrated with oracle database 23c 26c and autonomous database as well for cloud native application. So GoldenGate 26c or in fact 23AI sorry GoldenGate 26AI or 23AI they supports they support both on-prem cloud database they support both cloud databases on-prem database for 19c 21c and then they also support your autonomous databases as well extraction as well as replication. So how does microservices architecture in a nutshell looks like. So Oracle GoldenGate microservices architecture is a REST API based architecture which enable us to configure monitor and manage Oracle GoldenGate services using web-based user interface. Okay Oracle GoldenGate 23c R3 introduces the concept of microservices architecture. So from Oracle GoldenGate 23 oh sorry from Oracle GoldenGate 12c R3 onwards microservices architecture was introduced. Okay in August 2017 basically oracle released two architectures for GoldenGate. So basically what happened is the existing traditional architecture was renamed to classic architecture and microservices architecture was introduced in 2017. So as we discussed in the previous session until GoldenGate 21c both classic and microservices coexisted. So you can pick and choose based on the support you can pick and choose what type of architecture you want. Okay then from 23AI onwards only microservices architecture is supported. Okay so one thing one line which is mentioned over here is supports microservices and traditional architecture this I mean here supports microservices architecture only. Okay so I have corrected this so now it supports only microservices architecture. So GoldenGate 26AI supports only microservices architecture it does not support classic architecture anymore. In fact 23AI also doesn't support classic architecture it only supports microservices architecture. Okay so now how does Oracle GoldenGate microservices architecture looks like. So for Oracle GoldenGate microservices architecture here what happens is things runs as a service. So in classic architecture what we saw is it is process oriented architecture. Oracle GoldenGate microservices is process oriented architecture. Okay so things runs as a service over there things runs only as a process however in microservices as the name suggests things runs as a service. So in microservices architecture you have service manager. Okay so under service manager you have four services running administration service distribution service so you can call it server in the previous version we used to call it server now Oracle calls it as service. Okay so under service manager you have got four type of services administration service or server distribution service or server performance metric service or server or receiver service or server. Okay so these are the four services which runs so I call them as a service overall because they run as a service when you do ps-efgrep as well you can see these services running and in fact from 26AI onwards now they call it service only now instead of server. Okay so under service manager so service manager is basically watchdog of all these services so you have these four services here. Okay so what is administration service so you saw that extract replicate were configured the trail files were getting generated so everything happens under admin server so whenever you configure any process in GoldenGate like any extract replicate you have to start stop the processes you have to purge the trail files or do any other task all those things happen under the administration service. Okay all those things happen under the administration service. Okay then you have distribution service so distribution service is basically in classic architecture you saw data pump right so in micro services architecture it has been replaced with distribution service. In micro services architecture it has been replaced with distribution service so distribution service as the name suggests it distributes the data so extract configuration happens under administration service then the trail files are generated once trail files are generated the distribution service reads the data you view basically what happens is under distribution service you configure multiple paths. Okay so path means sending the data from source to target. Okay so distribution service will read the data from the local trail send it over TCP IP network where it will connect to the target receiver server. Okay so in classic architecture you have data pump now in micro services you don't data pump instead you have distribution service in the same way in classic architecture you had collector process instead of that now you have receiver service in micro services architecture. So classic architecture data pump has been replaced with distribution service in micro services architecture. In the same way classic architecture collector process has been replaced with receiver service in micro services are okay. So what happens is under distribution service, you configure distribution path which will establish with target receiver server. So these two services if you are if you are sending the data from source to target, okay in that case, these services will run both on source as well as target. So source distribution service will establish connectivity with target receiver service. Once the data is available into the target server, the replicant process on target will write the data to the target the replicant process on the target side will write the data to the target. Okay, so this is the actual architecture how it looks like. So if we are installing Golden Gate micro services on both source and target servers, so this is how it would look like. You have source database and a target database. Okay. Now if you see first of all in this particular architecture, you can see that there are these four services which are running on source side in the same way target side as well. These are the four services which are there right now. One thing which you would notice is the extract process on source under administration server the extract process on source is running under source admin server in the same way on target side. The replicate process is running on target admin server. Okay, then the trade once the trail files are generated the distribution service reads the data from the local trail sends it over TCP IP network where receiver server receives it and writes it to the remote trail file. And once that data is available into the remote trail file the replicate process applies the data to the target database. So one thing which you would notice is you have service manager. So basically service manager is responsible for managing all the services which are running in the environment service manager is responsible for managing all the services which are running in the which are running in the source or target Golden Gate instead. Okay. So service manager is watchdog of all the services and when server performance metric server. So performance metric server is again a new addition which was introduced from Golden Gate microservices architecture. It is used to manage real-time performance of Oracle Golden Gate. You will notice that distribution service is running on source receiver services running on target you can so basically when you create so all these four services if I go one screen back, okay, so all these four services together are called as deployment Golden Gate deployment. So these four services together form a Golden Gate deployment. Okay, this is service manager under service manager you have served you have deployment. So deployment is formed by these four sets. So when we say create a deployment, it means that under service manager, you will be configuring admin server distribution server performance metric server and receiver server. And let me tell you one more thing under one service manager. You can have more than one deployment as well. Under one service manager, you can have more than one deployment. It means that there can be multiple admin server distribution server performance metric server and receiver server as well. Now you may say under what scenario we might need more than one deployment like you are configuring Golden Gate Hub. So you want that for multiple databases. There should be separate deployment. You don't want to merge all of them together so you can have multiple deployments in your environment. Okay, so again repeat Golden Gate microservices architecture. Let me clear all the drawings. So Golden Gate microservices architecture basically looks like this on source under admin server. You configure Golden. So the crux still remains same. If you see under classic architecture, you had the same thing. It just that you don't have data pump instead of that you have distribution service. You don't have collector instead of that. You have receiver server, but crux still remains the same crux still remains the same. So basically what happens is under admin server, you configure Golden Gate extract process which receives the changes from the source database writes it to local trail. Once the data is available into the local trail. The distribution path. So under distribution service, you can you configure distribution path. Okay, which basically creates a path from source to target. So the distribution path reads the data from the local trail file sends it over TCP IP network where receiver service receives it and writes it write it to the remote trail file. Once the data is available into the remote trail file, the replicate process applies it to the target database. So this is how your data flow happens in microservices. So microservices is restful API based architecture where different services runs. It is service oriented architecture and 26 AI. In fact from 23 AI onwards and in 26 AI Golden Gate has now shifted to microservices only. Now, what are the key features of Oracle Golden Gate 26 AI? So the key features include AI driven data replication and monitoring. Okay advanced automation for deployment and performance tuning. So there are some features which are getting introduced for advanced automation for deployment and performance tuning enhanced security with zero trust architecture. So security wise again, it has improved improved cloud integration with OCI and multi cloud support. So basically if you go with OCI procure Golden Gate over there, you will see Golden Gate microservices architecture and new APIs for seamless DevOps integration. Okay, so 26 AI onwards. There are a lot of new things which have been introduced and which will learn a lot as the session. Okay. Now one thing which I want to like I try to introduce this slide so that you can understand AI powered automation in Golden Gate 26. Yeah. So basically I tried to create an analogy so that you better understand. What is a I powered automation in Golden Gate 26 a and why is it needed? Okay, so I tried to create it with the analogy of manual. Ashes you when you are in mute. Okay. So now I have tried to create an analogy so that you understand a I powered automation in Golden Gate 26 a okay. So if I have to make you understand this I was thinking like how can I make you understand? About if I have to make I have to give you the difference between classic and microservices and what is the difference between the two? So I came up with this analog. So consider traditional Golden Gate older versions like a manual traffic control system and Golden Gate 26 a adds AI intelligence like upgrading the city to smart traffic management system. So you must have seen like when your signal goes off due to some heavy rain or any other reason due to power outage. What happens is you see traffic controllers managing the traffic and it becomes really very difficult for each one of us to understand right? So in that case, what happens is considered traditional Golden Gate works like a manual traffic control system in that case DBA act as traffic controllers manual monitor data. They manually monitor data traffic identify bottlenecks reactively. So it's not like they will be proactive. They will identify bottleneck reactively and accordingly they will take action and take time to read out of fix issues in manual way. However, Golden Gate 26 AI adds AI intelligence and you consider like you're upgrading your Golden Gate to smart traffic management system. So in this case, what happens is AI continuously analyzes the traffic. You can consider data flow. It predicts potential congestion before it happens recommends optimal routes and performance tuning. It recommends what could be the better performance tuning you can implement it can automatically adjust traffic lights as well like replication parameters. It can automatically adjust that and it improves overall flow and reduces the downtime. Okay, so benefits of data replication proactive monitoring instead of reactive predictive insights for anomalies and performance issues smart tuning suggestions for faster stable replication and higher efficiency resiliency and minimum manual intervention. So basically with this I would say in couple of lines if I have to define this with Golden Gate 26 AI it delivers, of course higher efficient. It is more efficient higher efficient. It is more resilient and manual dependency is much less so you can say minimal manual intervention. Okay, and it also allows you to do proactive monitoring instead of reactive monitoring. Okay, so this is how it works. So what are the advanced feature in 26 AI? Cloud integration. So it's seamlessly replicate data to cloud databases API integration automate processes with recipes and on security now security is one of the feature which has been added it protects data with advanced encryption is okay. So now what are the benefits of upgrading to Golden Gate 26? Now say for example, you are using classic architecture many you must have seen a lot of projects. They are now moving to. 26 year there are discussions going on for upgrading their environment to for upgrading to 26 AI. Why what are the benefits for that? So the benefits for that include improved performance with AI driven automation. So the 26 AI delivers a provides you with improved performance with AI driven automation easier cloud deployment with Advanced OTA OCI integration. Okay, so with in OCI you have Golden Gate 26 AI microservices strong security for advanced Enterprise application. So security wise also it has become more robust and more dependent is this and simplified management with new microservices and API capabilities. So with I'll tell you one thing when Golden Gate 12 CR 3 was launched with microservices initially, there were issues or I should say there were problem for me to understand it because it was completely new architecture and UI has now upgraded a lot from newer versions. These UI of Golden Gate 26 AI has improved a lot. So some of you who have worked on microservices 21 C or 19 C and now you will see 23 I 26 AI UI. It is more eye-catching pleasing to eyes and the simplification has also been introduced or I won't say it is more simplified. The thing is with with microservices architecture. Now almost 10 years old we are getting used to it and ultimately we have to get used to it because see it is the future of Golden Gate and lot of organizations are upgrading themselves to Golden Gate 26 here. And in fact one thing which I always recommend each one of you is AI is something which you cannot run away with. Okay, what we are doing right now is we are not using it that much to its capability, but whatever it is, we should use it. We should try to use this AI tools as much as possible. Like you use chat GPT Claude Gemini co-pilot, whatever model you use grok deep-seek. You should be making use of these tools a lot guys. Trust me. Whatever time you invest in learning AI tools or AI related products anywhere. Trust me your life will become much easier. You will climb the ladder of words as well. So learn AI tools. Okay and on now I'm talking about AI tools. So let me talk about one of the AI tools which I explored recently is Framer. Have you guys heard about Framer? So what Framer does basically is Framer helps you in designing your website. Okay, it helps you in designing your website and that too for free to get started for free. For example, you can sign in. How did you find us? I'll say LinkedIn. Okay, so what this does basically it is it is really very useful. It helps you in building pages. It helps you in building your website. Okay, so you can use this AI tool Framer to build your stunning website as well. So say for example, I say I'll go to Framer just a minute. I'll get Framer AI. I want to generate it using I use it today only for something. I want to type for example, I'll design it from here new design page. Ask Framer. So it is using GPT 5.6 Terra model. You can change it. There are multiple it uses. I okay. It only uses this model. It uses Sonnet 4.6 model as well GPT. This is open AI model and this is Anthropix model. So I'll say build a website about me. Prepare about me page. Describing about Ashish Agarwal. Add my phone number. I'll shape it into a personal about page. So it is asking some question which opening composition should lead the page. I'll say editorial introduction, which visual direction fits Ashish with this warm and human. How about how much should the page cover brief essential sections should this stay a one-off page or include reusable shared element. I'll do keep it one. And then you can publish it for free as well. If you can integrate your payment gateways as well like Shopify or Razorpay PayPal. It does a lot of tasks for you. Can you then have the page hosted somewhere? Okay. Now after that, of course, you can get the link for one week. I think when we get it for free after that if you have to host you'll have to buy the hosting plan domain registration all those things. Of course, you'll have to do it can do the so the advantage is if you don't want to do all that separately, you can pay to Framer and it will do all the tasks for it will buy a hosting for you. It will register the domain name for everything. It will do for you. So you can do it separately or you can ask Framer to do it. Of course, it would cost. So what you are saving here is designing as well as development costs. So see it has created this kind of page. Now, if you want to publish it, you can click on publish over here. So this is the URL. If you click on it, you should see the page see it will show at the bottom that made in Framer. I'll click on publish website is published and pre-optimized will send this link to you. Try to open it. If you want to preview it, you can preview it as well. But preview is disabled. Why it's not opening opens. Yeah, nothing is coming because it's second URL ABC. Okay, see now see. Okay. What I have done is I have created a domain here custom domain with custom domain. It works fine. So I'll ping this to you. Hope not sure if it has published. Yeah, so see this is the page it created. So are you able to open the link on your machine? Yes. Yes, right. So I think it keeps it for if I'm not but this is new feature which I see recently of creating a custom sub domain. Okay, so I don't know why this this also should work. This is the base version, but it's okay custom domain is working. So this they have introduced recently that you can create your custom domain. So you have multiple options like you can redirect your existing URLs to main one to host. So basically for those who are unaware about what is domain what is hosting. So basically consider I'll make you understand it in simple words. So domain is nothing but your domain name like google.com. It's a domain. So you have to register this domain with some registrar. There are a lot of domain websites like GoDaddy where you register your domain you buy the domain. You have to really really renew your domain. Like I have this website right at Tranquil tech.com. What happens is I yearly pay a fees to renew it around 20 US dollars. I'd be okay. So this is the domain. Okay. Why am I why am I making you understand this? It's better that I'll ask Gemini or I'll ask for example chat GPT. Okay, it's better because we are talking about AI. I want you to make use of AI as much as possible. I can make you understand everything but let us see what does day my Gemini has or chat GPT has to say. So what I'll do consider me as a newbie to the website building world. I want to build my own website. Can you make me understand what is domain? What is hosting and requirement of them? Let us see. What does it reply? So basically what it says is simple analog. Imagine you want to open a shop in real world. So domain is something your shop address. Okay hosting is physical building domain. I would say I think it has this is not correct. I would say domain would be a shop name, right? Something like when I see a tranquil tech.com. It's a website name. So domain I would say it's your shop name then hosting is your physical building your shop address basically. So basically what happens is your website is hosted somewhere on some servers, right? You have to rent those servers. You have to you have to have your own servers where those websites are. Do you have your website is hosted, right? So that website hosting is being sold by multiple players name chip is their Hostinger is their GoDaddy, of course, so there are different hosting plans. You can buy shared storage where you can use shared hosting plan like my website will be there. Your website will be like multiple people like if you're a small website now, there are VPS hosting as well in which you get your own dedicated server where only your website will be hosted. Okay, then you have SSL certificate which you need. So SSL means like HTTPS. So in terms of website, like you do HTTPS Google or HTTPS. Like if you see my website as well, it is hosted on secured HTTPS. So I have to get the SSL certificate for it so that the certificate certificate is signed by the certificate Authority. So all these things are required. So basically here in Framer, they are giving you domain for free custom domain, which has framer.website and then whole if you want to host your website somewhere. So I think they will keep it for a few days after that. They will okay. This is really very good. If you haven't tried it, right just play around with it. I would say click create your kid page like some game page for your kids something you should try and build it and share it over our WhatsApp group. Like whatever idea you come across with by building the website free website share your idea what you are doing share your link so that everyone can take benefit of it. Anyone can take idea out of it. Okay. So what are the benefits of upgrading we discussed? Now, what are difference between Golden Gate 21 C and 26 a so so in 21 C a integration I would say is not limited. It is not there in 26 a you have a I powered automation cloud support in 21 C. It supports OCI and multiple cloud in Golden Gate 26. I announced OCI and multi-cloud support security. Traditional security is there in 21 C in microservices 26 a and on security is their performance tuning you have to do manual tuning in 26 a it is a I driven to database support until oracle database oracle database 21 C from 20 in 26 a it supports oracle 23 C 26 C autonomous databases. Okay. So this is about your difference between classic microservices and 21 C and 26 a so in the next session, what we will be doing now is in the next session. We'll discuss about download of Golden Gate 26 a in and install Golden Gate 26 a this is what we are going to do in the next session. So first two sessions basically what we have covered is overview introduction and few AI tools, which I wanted to discuss so that I can make you understand about a now the real training. We are going to start from tomorrow onwards from the next session onwards and it will be completely practical and we are going to pick up pace from the next session. So in the next session, we'll discuss about download of Golden Gate 23 and we'll install Golden Gate 23 Golden Gate 26 a I as well in the next session. A small note if I can recap what we discussed in oracle Golden Gate 26 a in day two session. So we discussed about microservices architecture and classic architecture difference. We discussed about what are the components of microservices architecture. So components of microservices architecture include service manager administration server distribution server receiver server performance metric server. Now data pump of classic architecture has been replaced with distribution server in microservices architecture collector process of classic has been replaced with receiver server in microservices. Then we discussed about different benefits to upgrade to Golden Gate 26 a so basically we are doing overview and introduction and benefits then we discussed about what are the key differences between Golden Gate 21 C and 26 a then we also discussed about Golden Gate architecture. Overview like different services which are there and then how processed fill data flow happens in 26 a I extract process rights to trail file distribution server sends data over TCP IP network to the receiver server receiver server writes data to the target trail file. Then replicate process reads trail files and applies data to the target data. Okay. Yeah, someone had a question. Someone will ask you asking the question. Is there a question you can raise your hand if you have any question on either. I see you are unmuted, but it's really very difficult to hear. Yeah. Can you please explain the that what is difference between classic and Microsoft that architecture wise the architecture wise. Okay, so classic and microservices basically the major difference is the major crux remain same in classic and microservices. You will be configuring the extracts replicates trail files, etc. Everything will remain same the major differences in classic architecture. It's process driven architecture. So in classic architecture, you have extract process running. So there is no service which is running everything is running as a process extract process writing data to local trail pump process reading the data frame reading the data from local trail writing it to remote trail and then the replicate process is applying. However in microservices architecture. These processes are encapsulated under the services like under administration service. You have extract process running trail files are getting generated. Then under distribution service. So data pump is no longer exist. You have distribution path and running under distribution service same way. You have don't have collector process. You have receiver path under receiver service. Then your replicate process on target runs under administration service. So in microservices architecture things runs as a service. So it is rest API service based architecture, but internally if you see in microservices as well, it's the extract process which extracts the data writes to local trail distribution path sends the data over TCP IP network receiver server receives it on target writes it to remote trail and replicate applies it to the target data. Hi Ashish Ashish in this slide. If we check the replicate, it's fetching the data from trail, right? It's the arrow is on wrong way. If I'm not wrong. So replicant basically receives it reads the data from the remote trail and apply it to the target it. Oh, I you mean to say this arrow should be this way. Yeah. Okay. I'll correct it. Okay. And apart from that one one major difference. I'm saying in microservices architecture in classic architecture our pump was like what was optional means we can configure extract as sending data to the target site, but here the distribution service is mandatory one. It is working as a pump. Okay. Yeah, that is another important point which you raise like in classic architecture. You could remove pump process in see you could remove pump process. That's correct. But extra in that case extract process will work as extracting the data and it will work as pumping. So process is nothing but secondary extract. They can take so primary extract will work as both X data extract mechanism as well as it will pump the data to the target. So so the capture process will capture the reduce from the online to the website. It receives the changes. Yeah, so the changes needs to be returned to the transaction log. So in Oracle these transaction logs are redo logs or backup of redo logs, which are archival so correct. So in the next session, we will be covering download and installation of Golden Gate microservices 268. So next session we'll be doing the practical and you will be okay.