Customer
For several years, the client has developed more than 200 java services that implement a complex electronic document management system integrated with ERP.
Problem
While there were few services, everything was fine, but at some point it became difficult to maintain and make change
How we solve
In the procedure of audit we found that 200 services, having as high as 90% code duplication. Lack of unit and integration tests made it hard to change. The manual management of 10 dev and 10 prod tomcats, along with the manual 10+ steps deployment process, made it a tedious and time-consuming process. The Git repo size was a staggering 3+ GB
After the Unison team's involvement, the system has been transformed. While the number of services remains the same, the code duplications were cut by 50% and steadily declining as we continue our efforts. The team has implemented unit tests for 10% and integration tests for 20% of the services, significantly improving the quality of the system with 0% test coverage in the beginning. All services are deployed to Kubernetes fabric, and the entire deployment process is automated and triggered by a single click in GitLab.
Success
The Unison team has successfully reengineered the client's infrastructure, resulting in significant improvements to their IT infrastructure. Startup time was reduced by 30%, deployment was 70% faster, and product SLAs were improved to the satisfaction of business users. Kubernetes manages all microservice lifecycles, ensuring smooth and efficient operations and scalability. Importantly, the reengineering was done gradually without any negative impact for the production environment. The business was not interrupted during the whole reengineering process.
Customer
A company working with big data and analytics tasks to improve the quality of support and reduce costs, decided to build an updated data processing platform based on Open Source Software.
Problem
Difficulties with updating vendor solutions. Lack of platform customization options. Lack of very narrow expertise.
How we solve
The Unison team engineered a custom Hadoop distribution with the customer's specific requirements in mind. The software distribution is based on the open source Apache Hadoop components with custom add-ons and backports, combined with Ambari cluster manager support and extra components like Apache Ozone. This allowed customers to get rid of vendor-lock and significantly reduce the cluster management and operating overheads. We have introduced Apache Ozone storage layer to a couple of the clients to optimize cost efficiency of the data storage vs Hadoop clusters. As every open-source project Apache Ozone had to be improved and sustained to match customers’ expectations.
Success
As a result of the work, the client received:
Customer
A company offering solutions in the field of replication and synchronization of distributed systems, which has a patent for its own distributed consensus algorithm.
Problem
Lack of expertise and experience with cloud storage and information processing platforms for quick launch of the service on the market.
How we solve
We were involved in both creating software products from scratch and enhancing existing solutions. The portfolio of our projects includes software that is able to perform active-active and active-passive replication of data on huge clusters with over a thousand nodes. We understand that every business has unique needs in data replication and our team is always ready to offer a comprehensive approach to meet their objectives. The team takes full ownership of the components and is working with product owners and stakeholders on the complete product cycle including the discovery phase of prospective product features, development, integration testing and the extensive support in the customer environments with optimizations for client’s specific needs.
Success
We built a solution for continuous data migration from Hadoop to cloud providers, as well as between data centers. The client successfully sells this solution and its support.
