Cloud reselling options - Design

 

 It is interesting topic when we try to understand this concept from application design point of view. Few technology companies never re invented storage cloud technology but by design the storage for them is iCloud.

Initially for me it is quite surprising but it can reduce lot of design effort w.r.t storage . We can look at some good examples , So how this cloud resetting works  in general and what kind of benefits it can bring to Application design.  When I was with yahoo and managing video CDN , we got interesting security issue that some one using cloud CDN infrastructure in their web application without paying any thing and this  costs lot of money as well. But what is happening there is Video CDN like YouTube player with YouTube branding which tells that content hosted at YouTube. But yahoo CDN at that time no branding attaching when playing content on a player. This is very interesting case of cloud reselling. even though it is cyber issue but the idea is the reseller able to get lot of benefit without  recreating cloud infrastructure.

Using cloud as SAAS based integration with application some times may result in reselling options but not always.We can see this trend in B2C applications rather than B2B.

Lot of the concepts I write in my blog  also I will test them in my demo web site. www.talash.azurewebsites.net .  I do refer this as Talash cloud as if I do have my own cloud infra. But  my application will share image/video/pdf who ever want to bookmark and keep in their own storage without really subscribing to any cloud storage. This is what most of the image/ video/ pdf content related sites do. they don't need to tell end user what cloud

their content hosted and shared but the end user will just happy that content  can be retrieved easily on any device without  playing around with password expiration and protection strategies and reading those marketing mails from those  cloud services. In my view if any technology not for common people just leave it for corporates. No need to play with and prove that one is tech genius. Some times it hurts.

Scalable application Design - Example app EasyWin

 In this application design important design consideration discussed mainly to make sure at least 1k concurrent users can use this app and also required reporting regarding review of content and dashboards. Major design objective is to scale this app to work with more users and more content.

AI First approach is new Design paradigm . Always I will check the patterns from Microsoft and one of their training on Modernizing web apps, the search using vector db and getting data from pdfs related to  hotels and  facilities in hotels  and accepting user voice commands and in some cases text to sql conversion.  The present app design also follow same approach to adopt AI in its design and my Favorite repo  for design reference is "Eshop container" for API/Micro service  design.

Work flow: Ingest & Reports

Review component design:

Contest component Design:

Work flow:  Contest creation

High level Design Goals:

1. Like other Talash web sites Talash Contest app - EasyWin will have components

•  UI developed for both desktop  and Mobile

• BFF to accept graphql requests from UI. This helps to in both aggregating services and                  also other benefits like flexibility of response data fields based on UI requirement .                        Which can reduce number of end points required . Apollo graphql client can add lot of      inbuild capabilities as Talash has many apps so even it is possible to achieve graphql            federation      and sharing schemas across apps.

• Major objective is to make it as framework based on micro services so that all Talash apps can share the services quickly and reusability will be better.

• Operability point of view azure  logging  and using  cloud for deployment can help to scale this app. Come across a tool where logs will show sql fired for that functionality and params passed to that sql Even though the code developed with ORM. that is quite innovative and reduce time to resolve issues quite quickly. These kind of  tools integrations makes operability better. My favourite is serilog for logging as it has lot of integrations and data dog which can show response time of each api end point quickly. At some point I am using data dog to compare response time of each api between two releases and find out any performance issues quickly.

• Scalability wise either event driven or service bus processing should be implemented to handle high volumes of request processing wherever possible.

• Lot of analytics tools integration like data bricks , power BI helps to understand the user analytics and trends on the app and can help lot in enhancing the geo based or demographic based  useful and prime content  which can increase user engagement.

• Objective of all talash apps is not hosting huge content but just hosting prime content and user should be easy to grab prime content from web and can share and save for day to day use.

Innovation 

Framework design  to support multiple apps. Framework  will have well defined API later and based on micro services .

Content classification algorithms are unique and  and can quickly filter the spam. Here AI will help lot compared to previous algorithms used in youtube or  other content hosting sites.

Third one is the prime content like flicker.com . It is site for photographers and people who enjoy photography. It is able to achieve with 100% accuracy. Same way Talash apps are unique in hosting

content and classifying content.

Content Sharing : Apps helps to identify  interesting content quickly and share with others.

Request Queuing and processing:

BFF  end point post data to Queue for further processing which can decouple any of bottle neck

handling huge requests . It is queue processing need to be with in the sla so that the data uploaded

can be retrieved with in 30 sec time.

 This was named as "Content Agility" where Hadoop used to processing queue  and quickly made consent available for end users with in 30 sec. So Data bricks can be considered for parallel

 extraction of data from files and updating DB in case of app service scaling costs more for its hardware.

Data synchronization

Review of content can be handled by multiple admin and needs process to schedule content

 and update dashboards . Signal R will make sure review data reflecting properly to 

multiple admins reviewing same content.

Analytics and Reporting:

 Component BUZZ will promote content to different channels based on predefined rules. Rules r stored in DB but can be moved to rule engine in future.

Eg. simple rule to promote content to main page is if any of category likes are more than 100 it will be added to the source of landing page.  BFF will check for the  likes and post data to Image service to update URL for landing page.

Reports r generated using data bricks and power BI, dashboards will read data from Databricks processed analytics data.

Advanced AI capabilities

Content review can be automated by using azure AI capabilities like cognitive services and others. spam detection and adult content  filtering , duplicate content detection and other useful review parameters can be automated using AI capacities. which intern improve review process and accuracy of review.

Separation of Concerns

This module should be an git repo by itself if possible that will address lot of separation of concerns.

But to reuse some of the components like BFF , image viewer and upload functionality it is better to create separate library and use that one in BFF project and also make image viewer more flexible which can accept some parameters.  Whole code will follow best practices of clean architecture and microservices which intern will address lot of separation of concerns.

Non Functional Requirements:

Basic infrastructure to start with will be app service for BFF, App service for widget which is part of existing UI and Service Bus for updating data to DB. To scale out to the required performance i.e 1k current Users 

• App service for BFF can be made higher tier compared to Basic or Free tier. When I try to  use function app which process data for 20 sec and scheduled for 5 min. total free cpu time for free tier exhausted. i.e 12 jobs per hour 24 hours running for 20 sec each job. 20*24*12 /60 min. Next app service tier may cost aorund 100$. so if we use data bricks for extraction  of data and updating mongo db. It can save lot of cost on hardware.

1. Async calls and using Service bus can  increase decoupling of components.

• UI will post the upload data to Blob storage. Which can handle most of the calls and some bench marking can be done.

1. Major processing is reading the user input and parsing the content and then pushing data to DB. This could be quite time consuming processing .  Mongo DB can handle concurrent requests pretty well and the Back end component should use Tasks and update DB with bulk updates. 

• Logging should use non blocking calls as well with proper tags  so that session stitching quickly and retrieve required logs for analysis

 UI design: 

• Additional component where user can give email and upload content.
• Admin can monitor the content and can delete  based on some exceptions.
• As given above the content for review can use some of the AI features where
• the content can be analysed offline and mark content for ready for review.
• Using AI components will be costing bit more compared to manual work but
• can automate lot of review process.
  

Security

• No user data will be saved at this stage as User login functionality is not attached to
• private data . This will be major enhancement once complete functionality is working fine.

SAAS based solutions and Pricing:

Cloudanary:  Not very expensive to get quick response of content from CDN. alternative will be akamizing content.

Search API: Google search api searchapi.io is quite good option. It cost 40$ per month for 10k requests per month. 

AI TOOLS:  Azure cognitive services  has many tools. Right now using text analyzers right now it is not quite expensive. But evaluating visual tools and their cost.  Google api also provide image analyzing API like google lens. Integrating these API is quite easy.

EmailJS. it is quite good and has very easy integration of templates and send mail functionality.

It is not expensive  till 1k mails. 

Genderize.io: It is my favourite and costs around 10$ per month and Definity quite good API.

Dyte:  Video CDN and with good api for conferencing.  It has free plan till certain number of request.

Auth0: it is used for security and user management. costs 35$ . Quite good at development stage.

Right now It costs 30$ for mongo db and azure appservices. So just planning to run 3 talash web sites below 100$.

Detailed Design:

Workflow: REVIEW

POCs to validate above design:

1. AI Engine:  Objective is to evaluate different AI capabilities related to image, video and should be able to get classification data correctly so that review can be automated as much as possible. Google lense api and Azure cognitive service custom vision are used in this POC. so any url for  image should get right classification and if it matches any of the categories defined content can be promoted automatically otherwise will be left in manual review queue.

         First POC is based on   Zero-Shot Classification

       

 

In  talash.azurewebsites.com unique thing is getting some key functionality based on AI capabilities so that they can give better results.   First POC is getting trends on AI based fucntionality. Here any content viewed or liked or downloaded will be added to queue for further processing if it can match any of the trending content. Here  Azure AI computervision will do classfication and then hugging face alogorithem zero shot classfication(LLM) will do categoriezation and filtering. If every thing gives good results it will be added to the trending topics and then new trends will be calculated.

 

Similar exercise also done for bring relevant videos from YouTube based on PDF content. Here AI text analyzer used to summarize document and more algorithms used to match YouTube videos and all scenarios first filtering of unwanted content is done before really linking the pdf to videos or content to show it as tending. 

 

Above approach basically make sure that capability based on right content instead of spending more time later to filter spam and other processing which may not be possible after some time.

All trending data will be passed to this LLM and get category and save them. Top   category tending data can be shown as top trending category to user.

 The same categorization will be used to filter content before adding it to trending to  avoid  any unwanted data. So same data will be processed on labels which can filter  spam   then actual label to capture trends in category like art, fashion or cars. 

2. Contest Wizard:  Very simple Wizard/UI where User can use it to join contest and upload content and get information related to contest details and results.

 

3. Rule engine: a simple rule engine where it can detect users activity and find some spam like duplicating data or some spam data detection or other rules related to geography or demographic.

 

4. Rule engine: to configure contest rules dynamically for each contest. Instead of adding thorough code for changes in contest.

 

5.Notification engine:  A Batch job which can send information about contests and track analytics regarding about contests.

Important new components. are the following.

Contest Creation

• Validation of email if possible one time password.
• Save file content and user details .

Content classification

• Validation of Data using AI services and manual services

Content filtering

•  Validation rules implementation.
• Major AI services including
•        Getting content description like image related data e.g. description.. 
•        Getting copy rights of the content.
•         Scanning for adult content.
•         Checking duplicates .
•         Separating premium and well classified data. 

• Review UI
• Implementing approval stages for the content  from  ingest, review, final .

Content security

Content Promotion

•  Promoting data based on criteria.

Contest Dashboards.

• Analytics and dashboards using Data bricks. 

Non Functional Requirements:

•  Enhance BFF and make it more modular by creating library for  feature.
• Loggin in to Azure. 
• Module should be pluggable into system so separate Data base, separate   angular component  and separate function app and service bus. Only BFF  will be shared component from the talash.azure.com.au components. 

Design Upgrade X1: performance , cold starts in cloud and tools

In this article we will look in to finetuning aspects of application design which are different from horizontal and vertical scaling.  

Use case is my portal www.talash.azurewebsites.net    where loading time for landing page is around 19 sec. and reduced to 500ms.

How to fine tune and optimize loading time depends on application. But there are general guide lines to achieve it as well. Like db indexing, multi threading / multi tasking , avoiding sync calls. Once we follow best coding practices above few things can be achieve. But in the above use case still I am getting 19sec as response time for the landing page. I started looking at what else required to achieve better performance. Come across "cold starts" and added latency in cloud. Try to avoid high resolution pictures in landing page  and used Cloudanary to get some help from CDN aspect and then Cache design for Graphql calls. and other techniques working as expected. Let us get in to more details of each technique and use case .

Check List: 

ref: How to Conquer Cold Starts for Better Performance - The New Stack

1. Deploy your function as a .zip (compressed) package. Minimize its size by removing unneeded files and dependencies, such as debug symbols (.pdb files) and unnecessary image files.

1. For Windows deployment, run your functions from a package file. To do this, use the WEBSITE_RUN_FROM_PACKAGE=1 app setting. If your app uses Azure Storage to store content, deploy Azure Storage in the same region as your Azure Functions app and consider using premium storage for a faster cold start.

1. When deploying .NET apps, publish with ReadyToRun to avoid additional costs from the JIT compiler.

1. In the Azure portal, navigate to your function app. Go to Diagnose and solve problems, and review any messages that appear under Risk alerts. Look for issues that may impact cold starts.

1. If your app uses an Azure Functions Premium or App Service plan, invoke warmup triggers to preload dependencies or add custom logic required to connect to external endpoints. (This option isn’t supported for apps on consumption plans.)

1. Try the “always ready instances” feature in our newest hosting option for event-driven serverless functions, Flex Consumption, which is in early access preview. This plan supports long function execution times and includes private networking, instance size selection, concurrency control, and fast and large scale-out features on a serverless model.

Design Uplift VIII - Trends and Analytics

Design Objectives:

1. User should be able to see trending content as landing page content. Which also should consider user preferences.

2. User clicks processed using service bus and function app to scale up the processing to more  clicks and also prioritizing user actions and if required process them offline. It can enhance scalability of processing User clicks.( likes, delete, comment , share etc)

3. Data bricks to bring near real time analytics and trends. 

4. Signal R for real time notification of content.

5. Spam detection done by audit user clicks and then  promote content or block content.

6. Azure AI  for creating prompts and getting more user interaction to define prefences instead of static list of preferences.

So above addition  of technologies can scale up the application to more number of concurrent user, can trigger back ground processing of data if needed and provide real time update to user regarding trending data. ,  which should help not  scaling  hardware  horizontally  incremental way as users/user clicks increase and increase cost of infrastructure. We can look in to more details .

Lot of samples and examples mentioned before are mostly used in my web sites. ( talash.azurewebsites.net  or TalashPDFDrive or other)  This time I am also planning to create separate repo which can be used by others and also for my web sites. Should be good attempt to create some thing generic and open source. Again Talash will be a platform byitself with quite generic components to process all types of consent and advanced functionality like content promotion and spam detection.

Technical Challenges:

1. Getting near real time trends will be processing data real time using data bricks. Demo app need not show near real time but design will explore details required to achieve.

2. Creating prompts for collecting user preferences where it can query users to get preferences and related information e.g. if     some one likes jokes and further checking if they like marriage jokes or pub jokes.

 This is one area where IT exploding with ideas now a days. I went to ecommerce web site and want to change my delivery address as I have given wrong suburb. So no need to call some one and no need to write mails or edit order which probably no ecommerce site can provide edit feature without a bug. So what is happening is just type i want to change my address then next it asks order id, then it authenticates and send email to customer care. More or less above prompt should have saved at least 2 years of development needed for ecommerce site to achieve same result.  So another hard R&D areas and will be technically challenging to design some thing efficient. there will be (Blood .. my famous quote) lot of learning by achieving tis one.

3. Using signal R effectively to update real time data and getting more engagement fro Users. It is more 

    UI / Ux challenge. I saw mixing real time updates with cache design has its own challenges. So smart invalidation of cache required.

4. Getting microservices design properly so that each service are modular and scalable. It is another challenge when code base and number of services increase and it involves addressing common code for all services vs  communication with in microservices.  As micro services stands they should be able

to work independently But it is having different patterns to communicate properly and avoiding  duplicating functionality.

5. Making sure no scalability issues in processing. Using async as much as possible and other aspects.

6/ Biggest design challenge I will explain where  building widgets directly out of data bricks.  This one interesting area where lot of engineering evolved in decades. Just by showing static report to dynamic dashboards are two extreme use cases where technology and design decisions can be more complex.

At the end of the day data tools (ETL and Reporting) always show  better way in delivering smart widgets compared to any smart coding to show trends and reports.  Now combining AI features in data tools  taking them much ahead in game.  so 80% of the widget logic will be build at the place of data rather than fetching raw data and building widget. I will deal with simple usecase whre trends are processed and combined with user preferences. So I can get widgets for above to plug in the UI like trends category wise or  charting  processed ETL data and showing them as widget on UI.

Above list will grow to more than 25 as it was developed , even small function app deployment took some time and the reason is pretty simple that .azurefucntion could not be found error. Finally it was given that files with . cannot be part of compressed deployment artifact in Linux and then changed function app hosted on windows. Still it is technically not clear but I could make it work. So there will be challenges to address on the way.

Week1 preparation:(Inception and Visualization draw.io ) 

In above design diagram User action component already designed and build with service bus and function app and saved data to mongo db. The next block to design is more analysing data and building trends and UI to enable user to subscribe and add to their preferences. As discussed before it is not going to be fetching raw data and processing in an additional component , it will be processed by Data tools and prepare response good enough to bind data to  UI quickly. Explored features of Databricks quickly and it is enhanced versions of ssis or adf , able to evaluate few data processing from mongo DB and created notebooks required to get materialized views to show some trends. 

First point to note here is using Medallion Architecture which probably very good for preparing data and reporting now a days. This one also more organized way to build additional processing units to enhance the preparation of trending data. It helps to add more checks at silver layer or get more aggregated results at the gold layer. while doing above exercise I come across two videos from data bricks demos which also are quite interesting. 

1. Ghodsi's Microsoft Ignite Demo: Databricks | Databricks

2. LakeFlow Designer: Production-Quality ETL with No Code Required | Databricks

Even though my design basically need some trending data above two videos showing much better capability of data bricks inters of building advanced widgets out of data bricks quickly. First impression  is can build prompts using vector db and RAG in 10 min and some copilot kind of features in Lake flow Designer. Previously i gone through AI search training and spend hours to build vector data and then prompts and getting end point to show that one in UI. So things are now simplified but again we need the capability to evaluate the simplicity vs efficiency. That is why we are also getting companies like Perflexity which can evaluate simplify vs efficiency of these LLMs and tools. May be few of these features will be embeded in above design. For now able to build required note books.

Week2 Preparation:(Design blueprint)

critical decision making week as it is not just embedding new technologies like data bricks and creating prompts it is just fine tuning exact technical objectives. So given this feature a name first as "WishperMyWishlist( wmw).  Idea is making sure content is close to the User want to see. As discussed prepare trending data based on topic and then prompt user for any interesting category  as user preference and again popping up some static list which can be dynamically changed based on the top user list or based on some User data.  The whole thing will be designed as an widget. Here the only challenge is data will be pulled by data bricks and process and push it back which will be used by prompts and static list to reflect and ask User to select more relevant categories. Efficiently of the whole widget is User can see real time the category required and with good content. Content agility is one concept still not sure how much being used with same name is to push data to user quickest possible time Hadoop being adopted to get better content agility and "Buzz"" is the word to to pop data up the ranks based on user clicks. at the end of day  showing most relevant data User want. Sentiment analysis is far challenging and beyond the scope of this feature implementation.

Design approach #1:

1. Get preference save to DB. 2. Generate meta data for existing content and append to existing meta data.(AI tools can be handy to get synonyms or relevant tags and even now a days generate tags for images). 3. Update trends and push trends data back to DB,

Design Approach #2:

In above design no content will be crawled. But tags are managed better way. In approach 2 content crawling done then instead of re tagging crawled content and tags will be used and generate trends

Design Approach #3:

My favourite approach and robust one is Bring crawled data and then re tag content with filters and spam detection and then present back to user.

Next week onwards it is more implementation to meet above Objectives and let us discuss more deeper and feasible / posible solution.

Week 3 Preparation: (POC)

It is all about code. First widget is all about user activity. I want to see which content being viewed, liked or which content downloaded.  There are two journeys implemented 1. User can use PDF app to upload data at htps://talashlogs.z26.web.core.windows.net as mentioned above adding meta data related to creation time, keywords and summary of pdf all avialble with azure Text analyser API and it is already being used in above App.  Now to integrate data bricks for providing few analytics widgets  keywords an summary saved to DB. So this widget can work real time with following integrations

1. User uploads pdf and that will trigger function app using blob trigger.

2. The function app get all keywords and summary form AI Text Analyser and saves data to DB.

3. Data bricks pull above data it got connector Estuary Mongo db connector and which works perfectly. Now I got note books which can show how many files uploaded, how many key words generated and how many files not able to get summary form AI tools and exceptions and based on keywords a preview available to see if any interesting content uploaded which can be used for further promoting and bringing to the relevant users. 

As usual publishing function app worked perfectly form VScode and spend some time in performance, if 500 files uploaded can all data processed by function app and generate keywords or it needs Event grid to queue requests and process them. I like this connectors used by data bricks which are bringing quite real time  don't need to add any sync interval so far. The Note books process MongoDB data , data bricks push widget with new data  to web site in a scheduled way and it is quite working end to end.

When it is required to further fine tune the above widget there will be more challenges and I am happy to see some thing done this week. to summarise few technical challenges left 

1. Text Analyser has some limits related to the Free tier I am using.

2. Still need to find how many concurrent file uploads can be done.

3.  Bringing some analytics about download of content .

4.  Adding more intelligence for relevance of the keywords and summary. As I use two types of api to summaries I got better results and I am thinking I can use multiple AI tool like chat GPT and other to compare results and make some decisions.

Week 4 Preparation:(Alternative design)

1. Looking at alternative designs and do some analysis.  Started looking at cosmos db. Initial impression is expensive for small data sets compared to Mongo db. Looking at other features like replication and  data quality point of view. 

2. Achieving real time computations and showing analytics. It is quite possible and data sync to data brics happening real time. Dashboard updating is possible quite close to real time right now scheduled to 15 min. 

3. RBAC when shared data to users especially analytics.  the public data is quite easy right now service priciple created in data bricks which enables embeded widgets and sharing to external users. Now looking same feature to work with registered users and their analytics.

4. Eshop container has very good module to view visually different dependant services and thier status and integration of logs. This is a enhancement for the code quality rather than feature, but would like to integrate to quickly understand why numbers changing on dashboards.

5. Content promotion "BUZZ"  based on tags and improving the quality of summary of pdfs possible by using different AI engines. 

Above 5 basically will tighten nuts and bolts on the processing side and then UI can be updated quickly to provide the design on Apps.

  

 

Design Uplift: VII - Using Design patterns Part 1

Using Design patterns is one of the core element in making any application design robust and scalable.

I will give in this article few examples which are quite generic and useful for most of web development,  These are my favorites and this one is quite  challenging skill for any architect/developer to bring best design pattern in the application development. But few patterns are not very complex which can be discussed and may be few references I can mention to understand more about it.

One of my favorite pattern is strategy pattern. As shown below just there many features  which can be selected based on the Execute strategy inputs.  There are many other patterns are very useful in  apps like singleton or facade. or factory.. and others. I need to start with some thing. One  interesting example when I am learning this design pattern is one of Ecommerce product has a folder. If we move any code in that folder ex show recommendations function to that folder feature will be enabled  in that ecommerce site . 

As it is open source product I could go though implementation and finally saw the following code implementation but done exceptionally well.

There are many interesting facts about design pattens. In one product the return object / structure  from Facade layer is BOject and company name is Business Objects. After couple of days of working with that code i realize that it is best way to implement Facade. So we can go though lot of these implementations one by one.

In my project talash.azurewebsites.net User action processing implemented using strategy.  Any user action from different web clients will post data to service bus. till that point it is common code.  When processing different user actions like "Like", "delete" "comment" etc. There are many common tasks like logging , calculate rank and other things which are common to assets ,  so they are part of base strategy class and the only code functionality where  other than updating db will be  passed to the stagy derived class and executed which is very specific to User action. 

Even though it looks like just switch statement on the user action but when code organized in to above class diagram it is much cleaner and modular. User action processing will be dynamic addition just by adding the processing logic required for new action.

Let us look at Builder pattern which is creational pattern and mostly used as well.

 A strategy pattern is use to perform an operation (or set of operations) in a particular manner. In the classic example, a factory might create different types of Animals: Dog, Cat, Tiger, while a strategy pattern would perform particular actions, for example, Move; using Run, Walk, or Lope strategies.

In fact the two can be used together. For example, you may have a factory that creates your business objects. It may use different strategies based on the persistence medium. If your data is stored locally in XML it would use one strategy. If the data were remote in a different database, it would use another.

same way many operational patterns depends on deletgate

Several design patterns use delegation. The State (338), Strategy (349), and Visitor (366) patterns depend on it. In the State pattern, an object delegates requests to a State object that represents its current state. In the Strategy pattern, an object delegates a specific request to an object that represents a strategy for carrying out the request. An object will only have one state, but it can have many strategies for different requests. The purpose of both patterns

is to change the behavior of an object by changing the objects to which it delegates requests. In Visitor, the operation that gets performed on each element of an object structure is always delegated to the Visitor object.

So some times design can use multiple patterns for depends on use case.

Above explanation given just to differentiate builder pattern with strategy pattern. some use cases we may implement both together as well. So now we can focus on Builder pattern which is very common creational patterns used. 

Here is classic example of builder pattern

class MealBuilder:
    def __init__(self):
        self.builder = Builder()
    def prepare_veg_meal(self):
        self.builder.build_part1("Salad")
        self.builder.build_part2("Vegetable Curry")
        self.builder.build_part3("Rice")
        return self.builder.get_product()
    def prepare_non_veg_meal(self):
        self.builder.build_part1("Chicken Soup")
        self.builder.build_part2("Grilled Chicken")
        self.builder.build_part3("Chicken Tikka")
        return self.builder.get_product()

The Builder Pattern is a creational design pattern used to construct complex objects step by step. It will be complex to just doing every thing in constructor to construct an object. By following builder pattern it is possible to depend on few methods of the builder to do construct object. The flow and methods used to build object will differ. 

Above example is simple one. But complex objects can be build 

Pizza pizza = new Pizza.Builder(12)
                       .cheese(true)
                       .pepperoni(true)
                       .bacon(true)
                       .build();

Builder's setter methods return the Builder object they are able to be chained. Pizza object construction done using builder class.

There are many practical examples we come across for this pattern.

NET StringBuilder class is a great example of builder pattern. It is mostly used to create a string in a series of steps. The final result you get on doing ToString() is always a string but the creation of that string varies according to what functions in the StringBuilder class were used. To sum up, the basic idea is to build complex objects and hide the implementation details of how it is being built.
Sql Query Builder.  API like fluent also uses above pattern to build the fluent API and pass different values to construct .

I like the following summrization of creational patterns 

 

L Let us look in to  more complex use cases and some of important aspects we need to consider while using these creational patterns.