The topic of cloud cost optimization is increasing with an economy that is flirting with a recession. Organizations that made decisions to leverage AWS cloud computing for lower total cost of ownership and elimination of large CAPEX spend are scrutinizing their utility-based compute and storage costs. While examining resource utilization and identifying idle resources, unattached EBS volumes and recommendations for reserved instances are imperative. Cloud consumers should not forget that leveraging serverless with Amazon Lambda can have the greatest impact on costs where it is a viable solution.
Challenges
The successful adoption of cloud computing requires a shift in thinking from traditional datacenter architectures, as well as the organizational process and oversight to drive the initiative. In the past PTP has discussed how the Cloud Center of Excellence (CCoE) plays a key role in the development, operations and governance of the utility-based service. From an architecture perspective, building compute/storage the same way in the cloud as in a virtualized datacenter environment does not maximize the available services, performance, scalability and cost. Rearchitecture of applications and their delivery can pave the way for huge improvements, though if this is not accomplished the only benefit reaped from the cloud is moving IT spending from CAPEX to OPEX, but with no improvement to TCO. This case study, however, is not about a PTP customer, instead it is about us and our journey to convert our PeakPlus Managed Services customer portal to be 100% serverless by leveraging Amazon Lambda.
The platform of tools leveraged by PTP’s PeakPlus Managed Services team is robust and required to deliver a full set of services to our life sciences customers in the areas of CloudOps, DevOps, SecOps, FinOps and NetOps. Delivering relevant information from each of the platforms to customers who subscribe to specific services requires gathering numerous service metrics across disparate tools to present to our customers. In the virtualization days of the past, this was commonly achieved by creating a separate file-based data store for making static reports available to customers to view or download on demand. In addition to the OS and database licensing, this proved challenging when “breaks” occurred between the tools and the file store and also created constantly growing data storage needs which ultimately affected our customers’ portal experience. This had to change.
Solution: Serverless with Amazon Lambda
As a “cloud-first” company born in 2017, PTP did not want to operate a datacenter, instead choosing to utilize applications in a SaaS model along with leveraging AWS. When building our PeakPlus View portal, Amazon Lambda was a forward-thinking approach by our architects to make calls back to our platform of applications to present the data that our customers needed. Besides Lambda, PTP leveraged CodeCommit, API Gateway, RDS and DynamoDB.
Our DevOps team can make changes to PeakPlus View by committing changes to the master branch. This will trigger a CodePipeline that will rebuild the application and deploy it. This uses a CodeBuild template that is located in the CodeCommit repository to control this process. The CodeDeploy is controlled by the template.yaml file which is also located in the code repository. All the data from PeakPlus View is accessed and delivered via Lambda functions that are defined in the template.yaml file. There are no permanent servers for the portal except for the database and the Lambda functions are front-ended by the API Gateway to provide secure web access. Lambda functions inside VPC are utilized to provide additional security and the ability to control our outbound IPs via a NAT gateway to access IP controlled APIs from partners (more detail on this in a future blog). Most data for the service is stored in an RDS database except for session data, which is stored in DynamoDB, and access to DynamoDB is controlled via roles and user permissions.
Results
PTP has achieved over 25% reduction in platform operational costs, 50% reduction in SysAdmin resource time and an improved experience from earlier datacenter-based portal platforms used. For most of our customers, compute, mainly in the form of EC2 On-Demand, makes up 60-80% of their monthly spend. In many cases the compute instances are required, and we can assist with Reserved Instance programs that combine flexibility with savings. However, there are often opportunities to architect differently and leverage more serverless functions at a drastically lower cost. AWS gives you the power of access to resources and hundreds of services – it’s up to you to use it wisely. PTP has hands on experience to assist.
