Case study

eBag

eBag

Our impression of Delta.BG is very positive. In the beginning their technical team listened carefully to our requirements and proposed architecture based on our suggestion, but with some additional, real improvements. They invested enough time to answer our concerns and helped us find where exactly the problem lies. During the whole working process, the team demonstrated excellent technical knowledge, sound communication and quick task delivery. They were open to share the source-code and their know-how. This gave us confidence that we can manage the solution in the long term. At the end, in the production phase, Delta.BG again reacted adequately and promptly to solving various unforeseen issues and events. After the successful completion of this project, we were convinced it will be useful to continue working with Delta.BG, because there is a lot we can learn from them.

CTO of eBag

Head of IT

  • Who are eBag.bg?

    eBag is a specialized online store, offering more than 10 000 high-quality grocery products from selected closed-loop farms. Their offerings include meat, milk, fruits and vegetables, as well as mass commercial products. All items are in stock with up to 3 hours delivery time. More than 35 000 households are actively shopping from eBag.

    eBag’s mission is through the power of innovation and technologies to give people access to quality food and better life.

  • The Issue

    During the COVID-19 lockdown in the spring of 2020, the orders through eBag’s online platform started rising exponentially. At the same time, driven by the need to find new markets for their merchandise, the big food retailers and other new players headed to join the online market. The company faced exceptional opportunity for growth, but at the same time every misstep could have had a devastating effect on eBag’s performance due to the increased competition.

    At this key moment, some of the components of the infrastructure started misbehaving. Initially, the developers managed to to keep the issue at bay, however the cases became more frequent and it seemed as if it was a matter of time for the first disappointed customer to complain. The eBag team suspected resource shortage in the face of rising traffic as the main “culprit”. Would simply throwing more hardware at the problem solve it and how exactly this had to be done, remained pending questions that required deeper investigation.

  • The solution

    In the beginning of April 2020, eBag turned to us for help in solving this issue that held the company back in such a crucial moment. As the infrastructure of the store is hosted by Delta.BG, we already had an overview of the current resources, hence we were able to make thorough analysis in very short terms. It turned out that a simple scaling up would patch up the problem in the first place. However, with the expectation of a further increase in traffic, the issue would inevitably come back again. Therefore, the optimal solution was a combination of horizontal scaling of server resources with full automation of the deployment process, and a change in the error-and-failure detection algorithm.

    Vertical or horizontal scaling?

    Vertical increase (or vertical scaling in more general sense) usually means adding hardware resources (CPU, RAM, storage) to the current system. Horizontal scaling means adding whole nodes to the already existing ones and balancing the load between them.

    If you would imagine a snack kiosk, servicing a queue of hungry customers, vertical scaling would be increasing the number of people handling orders in the kiosk. Horizontal scaling would be opening parallel kiosks, copies of the original, to meet the rising demand. When resources become scarce, the obvious question is in which direction to scale. In case the server in question is on-premise, we have just one option – vertically (scaling in our own hypervisor is in fact still vertical). Placing their infrastructure in the cloud, eBag had the opportunity to scale horizontally, obtaining the following advantages:

    • Possibility for unlimited hardware upgrade;

    • Close binding between operational expenditures and revenues;

    • More system stability;

    • Dropping the necessity to pay constantly at the rate of the highest used capacity.

    Automation of the deployment process

    For the implementation of the flexible horizontal infrastructure scaling, according to the established best practices, we suggested moving to Ansible as Configuration Management solution. The advantages of this method over shell scripting are many, but few of the more essential ones are: parametrization, config files templates, and full automation of the deployment process. The last one is crucial for optimizing development because it allows testing and going live with just the press of a single button – without interruptions and without customers noticing that work is being done on the store. Yet, in case of a failure, the automation of the deployment allows for a full rollback to the last stable version, so that an additional error-testing can be done. The mix of technologies we offered were open-source, so they can be audited or managed by a third-party without locking the customer to a certain service or developer.

    Employing best practices for error and failure detection

    Even with the most flexible and sturdy infrastructure, even with fully-automated deployment process, issues in the systems are possible and should be detected and removed as fast as possible. Accessing every single server to check its logs could quickly change from a minor challenge to a nightmare, especially with the increase of the number of servers to meet the increase in traffic. The eBag team knew that centralizing logs would avoid such a scenario, but they needed a more specific solution. We suggested installing small shipper from ELK Stack on every server. This would allow for searching, archiving and processing of the important information. Except for optimizing the work process, this would also free up server resources. Again the solution was open-source, which saved the customer additional licensing and fees.

    Following the above events, the plan for a renewal of the systems was developed and presented to eBag and after their approval Delta.BG was able to test it and deploy it in three weeks. Virtual testing platform was built for the project and a copy of the new store was treated with peak traffic simulations. The results were excellent and we went into production. Going live was done without any interruptions, so not a single customer can be disturbed in any way.

  • The Result

    After the changes, the eBag team can be assured that the website will work without outages, irrespective of the traffic that it awaits. The store can scale up and down, to face all of its customers and to service them without any delay. Crashing of services in peak hours is now a thing in the past. The deployment process is now automated and facilitates the developers’ job. All changes can be made quickly, so to respond timely to all marketing needs. In case of errors in the system, they are handled much easier, and importantly, much faster. The website is scalable, sturdy and flexible. Because of the changes done, eBag can now not only process more traffic, but can confidently grow their business to other cities, facing new increases of their customer and vendor base. Dela.BG team ensured that their infrastructure will stand any challenges that might come in future.

Other testimonials

The success of our clients is a success for Delta! See all

HiComm

The sites of HiComm and GoGuide are our main business tools and any disturbancies in the work of our editors on the backend, as well as any end user access interruptions are critical problems for us. After moving to the Delta Cloud solution any of these issues were gone.

Nikola Papukchiev

Chief Editor and Managing Partner, HiComm

Prodesgin.

When you need personal approach from your cloud provider, you will certainly find it with Delta.

Marian Naydenov

CEO, Prodesgin.

Sanita.bg

Our website was down often, mainly because of reaching the limit of our resources. In such moments we lose done deals just in the months busiest in marketing activities. After we moved to Delta Cloud the site is loading several times faster on a normal day and the cloud is giving us the possibility to scale easily our resources, whenever we need it.

Tatyana Atanasova

Website Manager, Sanita.bg