A Conversation with Duda’s DevOps team
As we promised in yesterday’s blog post “We want to hear from you”, one of our focuses this year is to share more customer stories with you. These customers come from different cultural and industrial backgrounds and are in various stages of their IT Financial Management journey. But they shared one common goal that is to achieve their business outcomes and organizational goals by leveraging the appropriate financial management strategies.
We recently sat down with one of our customers and a valued partner – Duda to reflect upon their Cloud Financial Management (CFM) journey. In this blog post, we will give you a brief overview on how they tackled this important task.
Cloud Native since Day 1
Duda is a professional web design platform. The company serves all types of customers, from freelance web professionals and digital agencies, to the largest hosting companies, SaaS platforms, and online publishers in the world. Duda started using AWS services in March 2009, when the company was founded. They currently run all of their workloads on AWS. Among others, they are using Amazon S3, Amazon RDS, and Amazon EC2. Duda is a rapidly growing company and although the traffic to their sites went up more than 234%, their AWS bill grew by less than 193%. This resulted in the decrease of cost per business outcome unit (here: website visit) by 18% over the 3 years. However, even though their infrastructure is entirely cloud native, they are still adapting internally to meet the CFM needs.
Duda’s Cloud Financial Management Journey
It all started when Duda’s DevOps team (who also manages Duda’s FinOps tasks) noticed a raise in their cloud bill. With the help of AWS team, Duda’s team decided to look closer into this matter. Thanks to their continuous effort, now only 8% of their Compute workloads are On Demand, which led to approximately 37.5% in saving. Danny Mann and Itay Sued, two main heroes behind the effort, shared some good tips with us.
1. Make it an executive-level matter
By including the CFO and the executive team into the cloud spend dialogue – Duda’s DevOps team got clear executive level sponsorship, while earning trust from the finance leaders. Cloud spend optimization wasn’t solely an IT problem to solve.
2. Ask for help and leverage AWS resources
The AWS Cloud Financial Management team worked with Duda’s DevOps team to prepare a detailed analysis. They identified some EC2 instances with low utilization and made recommendations to replace them with more appropriate instance families for better CPU / memory mix.
“We have attended re:Invent for a few years, finance-related workshops and we use Account Manager and Solutions Architect, who we consult often around the bill, but not only. That’s where the majority of our knowledge comes from. It’s a lot of learning, a constant journey, we like to be informed about new features as well.” – Danny Mann, System Architect, Duda
Make sure a dedicated team is monitoring spending and analyzes the results. At Duda, a devoted team was tasked to monitor cost anomalies per service, per hour and daily. The mentioned team also screens service specific situation, such as Amazon s3 buckets sizes, number of EBS snapshots, orphaned snapshots and AWS Lambda invocations to mention just a few. The clear ownership and empowerment led to costs being streamlined.
4. Establish a clear process for monitoring and tagging
Duda’s DevOps teams are monitoring the bill by hour, making sure their users have access to self-service reports and stay on top of this – they have established a monthly process, where they compare the cost and usage growth with the established baseline. The mentioned team uses AWS Cost Explorer CLI for data extraction and injects it into Grafana (learn more about the integration between Grafana and CloudWatch here) to analyze the data and identify non-standard behaviors. Additionally alerts were added while the budget threshold is almost reached.
Duda’s team also uses tagging to identify environments and partners and in some cases specific resources (AWS Lambda, Amazon s3 bucket, Amazon Dynamo table etc). They also make sure to monitor that the required tags exist for every resource. In some cases, an automatic tag is being applied. Duda’s team has admitted that the work with the tagging is not finished yet and this, along with implementing some other strategies suggested by the AWS Cloud Financial Management team will be the focus of their future FinOps efforts. This is an ongoing effort, there are always additional services to improve and to monitor, from cleaning S3 buckets and changing S3 storage tiers, to deleting unused EBS volumes, to monitoring AWS Lambda invocations. Duda’s team agreed that there is always more to learn; new tools, new strategies and this curiosity is what fuels their work.
5. Leverage programmatic purchase options
Duda has seen the improvement, cost drop and coverage increase just by replacing Reserved Instances with Savings Plans as broader AWS workloads can be covered by Savings Plans. Duda uses a combination of Savings Plans for Compute instances and Reserved Instances for RDS/Elasticsearch/Redshift/DynamoDB. This strategy achieves higher reservation coverage while keeping some flexibility of infrastructure changes. As mentioned earlier, only 8% of their Compute workloads are using the On-demand pricing, which resulted in 37.5% savings. And thanks to the use of Reserved Instances, they’ve also achieved almost 40% in savings for RDS.
Whether you are cloud native, building your infrastructure on the cloud since Day 1, or enterprises, who are migrating your workloads to AWS, effective IT Financial Management strategies can help you proactively identify ways to save and optimize. There are several essential tips and tricks in the toolbox that you will find useful for your near-term and long-term needs.
We’d like to extend our thanks to the Duda team, and our colleagues, Sandra Topic (Cloud Economics Senior PM), Cyril Deblois (Cloud Financial Management Business Development Manager), Tali Ellenbogen (Account Manager), and Shani Turiski (Customer Enablement Specialist) for working with Duda.