How Cloudamize and SoftwareOne Ukraine helped Ukraine’s largest retailer build a data-driven cost model and optimal AWS target architecture using DISCOVER, ANALYZE, and PLAN.
About the Customer
Epicentr K is one of Ukraine’s largest multi-category retail groups, operating over 72 shopping centers nationwide. To support its scale, the company runs a large on-premises IT estate spanning a broad range of mission-critical workloads, from search and databases to containerized applications and messaging systems.
About the Partner
SoftwareOne Ukraine (Crayon Ukraine) is a global software and cloud solutions provider and distributor. With a presence in over 70 countries and a team of about 12,000 professionals, they combine global scale and local expertise to help partners and customers optimize costs, source and procure, accelerate growth, and navigate complex IT environments with confidence. Leveraging deep capabilities in cloud, software, and data and AI, the company empowers organizations to modernize, innovate, and unlock the full value of their technology investments.
The Challenge
With a public cloud migration on the horizon, Epicentr K needed a clear, evidence-based picture of their existing infrastructure before committing to a target architecture or cost model. The scope was significant — 335 servers supporting a diverse mix of workloads including search engine clusters, relational databases, container orchestration platforms, web applications, and message broker systems.
Without reliable data on actual workload performance and resource utilization, decisions about right-sizing and cloud spend would have been guesswork. Epicentr K needed confidence in their numbers before taking the next step.
The Solution
Working with SoftwareOne Ukraine, Epicentr K deployed Cloudamize DISCOVER using an agent-based data collection method across all 335 servers over a one-month observation period. Lightweight real-time monitoring agents captured accurate workload performance metrics continuously, providing a far richer dataset than point-in-time snapshots or flat-file imports.
Key highlights of the engagement included:
- Real-time data collection: All 335 servers were monitored continuously, capturing CPU utilization, memory consumption, disk I/O, and network throughput with industry-leading collection resolution.
- Workload-based right-sizing: Cloudamize ANALYZE processed the collected data to recommend optimal AWS instance types for each workload based on actual consumption, not provisioned capacity.
- Constraint-based optimization: Analysis revealed that 74% of servers are memory-constrained, 13% are CPU-constrained, and 8% require both CPU and memory optimization, enabling precise instance family selection tailored to each workload profile.
- Cost modeling: One-year reserved capacity pricing (no upfront payment) was applied across all recommendations, maximizing cost savings while maintaining operational flexibility.
Cloudamize PLAN then mapped each workload category to the appropriate AWS instance family:
- Elasticsearch cluster (56 nodes): c7i.8xlarge and m7i-flex.4xlarge for search and analytics workloads
- Kubernetes workers (63 nodes): r7i.large for container orchestration
- PostgreSQL databases (31 nodes): r7i/r7a instance families for memory-intensive database operations
- Web and application servers (17 nodes): t3/c7a instances for web traffic
- RabbitMQ messaging (14 nodes): appropriately right-sized for message queuing workloads
The Result
Epicentr K completed a comprehensive, data-driven cloud readiness assessment across all 335 servers, covering 309 Ubuntu, 22 CentOS, and 4 Oracle Linux systems with a combined 2,204 vCPUs. With right-sizing recommendations grounded in real performance data, the team was able to move forward with a clearly defined migration plan and a cost model they could trust.
Cloudamize identified that 58% of servers (194 out of 335) had average CPU utilization below 40%, highlighting significant right-sizing opportunities that would have been invisible without high-resolution workload data. The result was a migration plan optimized for both performance and cost from day one.
|
25% Projected Reduction |
|
Author
