Case Study of Convertale

Introduction

In the highly competitive world of e-commerce, success can be elusive. Cost cutting is becoming increasingly important for retailers to maintain competitiveness in today's market. One Turkish startup, Convertale, has implemented a real-time recommendation engine to aid its e-commerce clients in providing more value to their customers. The engine analyses users' actions in online stores to recommend items that customers are more likely to buy based on their previous selections and preferences. This allows businesses to tailor their online offerings to individual customers and increase sales in a cutthroat industry. It has been around since 2008, making it a veteran among Turkish recommendation engines (Somasundaram, 2021).

One Turkish startup, Convertale, has implemented a real-time recommendation engine to aid its e-commerce clients in providing more value to their customers. Companies like Convertale, use Elastic Load Balancing to make their company’s product reachable. With Elastic Load Balancing, your incoming traffic is automatically distributed among various targets, including EC2 instances, containers, & IP addresses, across one or even more Availability Zones (Jairam, 2020). It keeps tabs on the status of all registered targets and only sends traffic to those that are operational.

In contrast to ALB, which allows distribution via multiple ports as well as lambda functions, ELB only means allowing routing through a single port. Serverless coding, website development, and the generation of unique ALB targets are all made possible by Lambda functions, which can be managed and run by the user.

Application Load Load Balancers (ALB), Network Load Balancers (NLB), & Traditional Load Balancers are all supported by Elastic Load Balancing. These load balancers are compatible with Amazon Elastic Compute Cloud services. Load balancing appliances for applications are used to direct Layer 7 (Hypertext Transfer Protocol/Secure) traffic (Kumar et al., 2018).

When deploying on Amazon Web Services (AWS), use the load-balancing service known as Elastic Load Balancing (ELB). Incoming application traffic is automatically distributed and scaled by ELBs resources. To better accommodate incoming applications as well as network traffic, ELB allows IT departments to better tailor their resources.

The challenge

Convertale's recommendation engine was initially hosted on specialised in-house servers, but the company quickly outgrew them. Ismail Arslan, Convertale's Product Manager, explains, "The users build code inside the clients' websites in order to obtain data from the source." To put it another way: "When a client's website is popular, the servers also get a lot of hits from prospective buyers. So, the user needs to expand quickly. To put it simply, users couldn't handle that volume of traffic with dedicated infrastructure.

The on-premises hardware proved unresponsive throughout testing and experimentation and could not handle the volume of traffic. Moreover, Arslan says, "The data scientists have to experiment with massive data, which takes tremendous compute capacity, in order to uncover patterns in purchasing behaviour (Kumar and Sharma, 2018). Its on-premises architecture not only made it hard to foresee how many servers would be required for a specific experiment but also prevented us from reaching the full potential because users lacked a limitless supply of servers.

Convertible required a low-priced, upfront-payment-free platform to host its recommendation engine. Integration with other systems, especially open-source ones, was also a must for the chosen system.

Why use Amazon Based Services?

Amazon Web Services storage capacity was an excellent fit for the massive data sets needed to power a recommendation engine (Mishra et al., 2020). After doing some tests, "AWS was the obvious decision," as Arslan puts it. "It provided us with the scalability user required at the correct pricing."

Convertible employs Elastic Load Balancing in conjunction with Amazon EC2 instances again for the data collection phase of their operation (Backes et al., 2019). Scaling resources across Amazon EC2 instances are handled automatically by Auto Scaling, and real-time data processing is handled by Amazon Kinesis before being stored in Amazon DynamoDB & Amazon Simple Storage Service (i.e. Amazon S3).

Following this, Amazon Elastic MapReduce is used to process the data and extract useful insights (Amazon EMR). The suggestions are sent out using Elastic Load Balancing once more. For relational data management, Convertible uses Amazon RDS, and for service deployment, the company uses AWS Elastic Beanstalk (Mostamand Kashi, 2022).

In order to function, the recommendation engine uses several different technologies that are integrated with the company's AWS cloud architecture. These technologies include the Jenkins integration tools, RabbitMQ messaging, & various open-source tools.

"The user went fully operational on AWS in about two weeks," Arslan says. Also, every month, the firm processes over 500 million user actions for its e-commerce customers. Arslan notes, "The answer has always been knowledgeable" from AWS Support when his company has needed help. To put it simply, AWS's customer service staff is quite knowledgeable.

Benefits of AWS

Convertible can provide a consistent and informative service to its clients because AWS provides the necessary scalability (Toosi et al., 2019). The company's system availability was 98 per cent when it was deployed on-premises, but it has increased to 99.965 per cent when hosted in the cloud. Working inside the cloud also eliminates the need for the business to schedule downtime and risk disruptions in service delivery if new features, as well as services, are introduced (Cornetta et al., 2019). Previously, it had to notify customers and temporarily disable all services for up to 10 minutes whenever the user wished to publish a new feature. This allows us to roll out the new service with zero interruptions. Because of this, service to customers will run more smoothly and they will experience fewer interruptions. The days of unscheduled maintenance are also over. There were times when problems with the on-premises system caused the unscheduled outage of up to 20 minutes. The AWS platform eliminates such occurrences.

Most importantly, the startup's existing infrastructure has been upgraded to better handle massive amounts of data (Cook, 2018). At the time of the initial move to AWS, they were deep in negotiations with a major new client (Al-Dulaimy et al., 2020). It couldn't afford to expand our own data centre to meet the demands of this client (Afzal and Kavitha, 2019). They were able to confidently bring on that client thanks to the scalability of the AWS cloud, which allowed us to employ Amazon Web Services (AWS) to do so (Dineva and Atanasova, 2021). Also, unlike some other cloud providers, AWS requires no initial investment. For us, it's a fantastic example," Arslan says.

He went on to say, "Users routinely execute Amazon Elastic MapReduce projects that may require a large number of servers simultaneously, depending on the client." Compared to processing this amount of data on physical infrastructure, users can accomplish it in a matter of minutes at a fraction of the cost. Users simply deploy them, run the necessary processing, and then turn them off. Likewise, this helps the data scientists at Convertale who are probing interconnections in massive datasets by giving them a testing and experimentation environment. It's simple to start and stop servers without having to invest the time and money required to provision hardware. According to Arslan, "when developers make changes to the recommender system, they should do so in a setting that's as close as possible to the engine's final, production-ready form (Houmani, 2021). The use of dedicated hardware prevents you from doing so. The use of AWS allows us to set up several distinct settings simultaneously. This method streamlines the testing process so that it takes less time and costs less money. Once they’re satisfied with the new functionality, they’ll cancel the AWS service & quit paying for it.

A further perk of the AWS price structure is that it enables Convertible to provide more competitive rates for its offerings. They're able to keep user prices down thanks to AWS," Arslan says. As a result, it has a leg up on the competition and can hold the user base against the world's biggest companies. Convertible has also saved money on operational expenses since implementing AWS technologies; it has also completely removed its initial investment (Xu et al., 2019). Since implementing Amazon Web Services (AWS) technologies like Amazon DynamoDB, Amazon Elastic MapReduce, and Amazon Kinesis, "It has witnessed a reduction in the cost of roughly 50 percent," adds Arslan.

Arslan sums up his time with AWS by saying, "AWS is wonderful." Wheels made of rubber instead of stone. It travels farther, faster, and with greater comfort and less wear and tears on the vehicle.

Conclusion

Thus, this study concludes the benefits of ELB’s based services, challenges and benefits for an organisation that chooses to grow by using the ELB. In order to help its e-commerce clients provide more value to their customers, the Turkish startup Convertale has created a real-time recommendation engine (Koritsas, 2022). Online shoppers' shopping behaviour is analysed by the engine to determine what things will most likely pique their interest and result in a purchase (Gupta et al., 2021). In spite of extensive testing and experimenting, the local hardware consistently failed to respond. The huge amounts of data needed to fuel a recommendation engine were a perfect fit for the storage capacity offered by Amazon Web Services. Amazon Kinesis processes data in real-time, and then stores the results in Amazon DynamoDB and Amazon Simple Storage Service (i.e. Amazon S3).

When hosted inside the Amazon Web Services (i.e. AWS) cloud, Convertale's availability increases from 98% to 99.965%. The existing infrastructure of the firm has been enhanced to better process enormous data sets. When compared to other cloud service providers, AWS doesn't necessitate any sort of up-front payment. Convertible leverages Amazon Web Services (i.e. AWS) to speedily and cheaply analyse enormous amounts of data. Since deploying AWS technologies including Amazon DynamoDB, Amazon Kinesis, & Amazon Elastic MapReduce the organisation have seen a cost savings of roughly 50%. In the future, most of the companies will plan to implement the AWS based services to grow their business model and services they offer.