CloudAccel

OVERVIEW

CloudAccel helped one customer reduce production costs by over 65%, by transforming its complex infrastructure with multiple dependencies to scale up and down seamlessly on demand.

BACKGROUND

This customer has been working on the platform for over 4 years. As in most companies, the engineering team's focus was on features, functionality, scalability and less on associated cloud infrastructure costs. Over time the system became complex, the traffic grew, increasing the cloud costs. The customer was exploring ways to reduce the costs without affecting latencies and without significant redesign of the platform.

CHALLENGES

The system evolved over multiple years. Due to its unique functional requirements, complexity grew involving multiple services and dependencies. Engineering team also developed processes and scripts for build and deploy activities to satisfy system's unique requirements. These prevented the engineering team to take advantage of the cloud provider's native scaling solution with a simple switchover. As is expected for a business critical system, the customer required a solution migrating to which does not involve any downtime, does not have negative effect on latencies experienced by end users.

                "With dynamic scaling, you want to reduce costs. But equally critical, you want your system users to notice zero changes, both during and after the rollout.” 

ALTERNATIVES

Exploring the available solutions and solution providers, the customer found that most service providers offered expensive and complex solutions that required long timelines to deliver. In addition, they required significant changes to existing build and deploy procedures the engineering team had devised and stabilized over time.

CLOUDACCEL SOLUTION

CloudAccel assessed current deployment and release processes.
Modeled automation to align with current practices.
Addressed various complexities such as pulling correct versions of various artifacts for the current release, pulling them in right order and setting them up, using the right configuration for various services.
Built a simple deployment and release automation that bootstraps on instance startup.
Devised a rollout plan with multiple fallback options throughout the process, to address zero downtime requirement
Managed the rollout over a 5 week period, tracking and frequent reporting throughout the process to ensure zero downtime, on a multitude of metrics at various levels including cloud infrastructure, OS, processes etc.
Analyzed the traffic and scaling response, tweaking various scale out and scale in parameters to meet the customer's latency requirements while keeping costs low.

RESULTS AND BENEFITS

Reduced production server infrastructure cost by over 65%.
Rolled out the changes with zero downtime.
Production services scale up and down with the demand

For example, during nights and weekends, setup scales down infrastructure to 25% of peak capacity without any degradation of service response.

Engineering team spends zero time responding to capacity problems.
The customer appreciates the positive side effects of increased scalability and agility of the system