Businesses running servers in the mobile ad exchange sector often face extreme loads that consist of billions of requests and process terabytes of raw data each day. When processing this much data each day, it is crucial for an organization to have a stable infrastructure that can deliver the best uptime possible. Organizations like Fyber looked to the cloud to try and reduce their cloud computing costs on Amazon Web Services (AWS) and quickly noticed that the infrastructure needed to meet their load and uptime demands is very costly.
Fyber uses Auto-Scaling Groups for EC2 and ECS to scale up their mobile ad exchange infrastructure to workload demands. During their journey to look for ways to reduce costs, they came across Elastigroup by Spot and were intrigued that existing customers were able to reduce their costs by up to 90% by running their workloads on excess compute capacity known as spot instances.
When Fyber began their evaluation of Elastigroup, they were pleased with how easy it was to get started. They were able to import their existing Auto-Scaling Groups so that their workloads will run on spot instances to reduce costs. For organizations that run an application or service behind AWS Elastic Load Balancer (ELB), Elastigroup provides the freedom to run a blended cluster from multiple instance types such as M3, M4, C3, and C4 with purchasing options for spot instances, on-demand, and reserved instances.
Fyber was also able to import their ECS clusters into Elastigroup and were happy with the features. With support for using mixed instance types and the ability to use a combination of on-demand and spot instances, they were able to reduce costs on ECS. The most important feature of using Elastigroup with ECS is the ability to autoscale the underlying infrastructure.
The Elastigroup ECS Autoscaler dynamically scales clusters up or down to ensure that there are sufficient resources to run all of the required tasks. The Elastigroup Autoscaler will also maximize resource efficiency at the same time by optimizing task placement across the cluster in a process called Tetris Scaling. For organizations like Fyber that want to manage Auto-Scaling operations themselves, Elastigroup has targeted and simple scaling policies.
Target scaling policies are autoscaling policies that track a specified metric relative to your desired target value and automatically adjust your Elastigroup to meet this target. Simple scaling policies provide the ability to scale ECS infrastructure up or down based on resource utilization metrics. Fyber is currently using simple scaling policies for their ECS Elastigroups to autoscale their infrastructure up or down based on their workload demands.
Here is how Elastigroup helped Fyber reduce costs and operate more efficiently:
- Seamless Integration – Elastigroup integrated with the existing AWS services used by Fyber: Elastic Load Balancer, Application Load Balancer, Auto-Scaling Groups, and Amazon ECS.
- Intelligent Auto-Scaling – Elastigroup provides tailor-made auto-scaling for your application. Metric-based or Event-driven, Elastigroup will always make sure that you have the optimal mix of Right-sized VMs running to meet your workloads needs.
- Spot Market – Elastigroup reliably leverages the spot instance market to optimize your underlying infrastructure for cost without compromising availability.
- Always Up – Elastigroup uses predictive algorithms to identify and drain VMs that are about to be preempted. Prior to VM termination, Elastigroup seamlessly launches a new VM and replace it seamlessly. Elastigroup will also make sure to distribute your VMs across different sizes and types to optimize cost reduction and VM longevity.
It took the DevOps team at Fyber just two days to configure its infrastructure to run on Spot. “Our setup is complicated because we have a lot of different Auto-Scaling Groups,” says Fyber’s SVP Technology, Gal Aviv, “but the Spot dashboard is very easy to work with and we get a good overview of stats.”
Using Elastigroup, Fyber was able to seamlessly import their existing Auto-Scaling Groups and ECS clusters without making drastic changes. By partnering with Spot and using spot instances, Fyber has been saving money while maintaining the performance of its platform. “The company saved tens of thousands of dollars – that’s between 20% – 30% of our total monthly AWS bill, and 70% of EC2 costs. We were easily able to maintain good performance while further reducing infrastructure costs – which made our board and management happy.” – Gal Aviv.
With offices in Israel, San Francisco, New York, London, Berlin, Beijing and Seoul, Fyber is a global technology company, developing a next-generation monetization platform for mobile app developers. Fyber combines proprietary technologies and expertise in mediation, RTB, video, and audience segmentation to create holistic supply-side solutions that shape the future of the mobile app economy. Their vision is to create a world in which people have affordable access to quality content while enjoying a pleasant, relevant user experience.https://www.fyber.com/