How Yotpo Covers 95% of their EC2 with
Reserved Instances
“It’s completely hands-free and we save 40% each month on our On-Demand costs”
“We are able to focus our time on our clients and not deal with cloud optimization”
“It’s cloud savings on auto-pilot”
Customer
Brief
Yotpo is a leading E-commerce marketing platform that provides advanced solutions for customer reviews, visual marketing, loyalty, referrals and SMS marketing. They help thousands of e-commerce companies accelerate their brand loyalty and consumer growth. With a $1.4 billion valuation, Yotpo has thousands of global customers and 600 employees worldwide.
Yotpo launched into the Australian market at the end of 2021 following a multi-year platform collaboration with e-commerce giant, Shopify, and a USD$230 million capital raise to drive expansion.
Key
Challenges
The complexity of forecasting the company’s compute usage was causing engineers to spend a lot of time managing, monitoring, and adjusting cost-saving commitments, but still paying a lot for premium On-Demand Instances.
Key
Results
Over the course of the year since Yotpo deployed Zesty, they have been able to cover around 95% of their EC2 usage and dramatically reduce their costs by more than 40%.
The Challenge:
To ensure ongoing production across their user base, Yotpo maintains thousands of AWS EC2 machines on which they run their Kubernetes and Nomad-based applications. Their cloud environment is very dynamic and application needs constantly change as their customer base evolves. This makes the task of buying AWS Reserved Instances three years in advance extremely difficult, forcing them to either buy too many or too few commitments. According to Yair Leshem, DevOps Engineer at Yotpo, “it was impossible to accurately predict and purchase what exact usage is going to be three years out when it’s difficult to predict what it’s going to be 3 weeks out.”
In order to minimize this risk, a lot of time was spent managing Reserved Instances so that they could monitor usage and try to more accurately predict how much future coverage was needed. However, as changes are constant, this quickly became a time-consuming task. Engineers were finding it very difficult to keep up with the scale of managing thousands of computing instances that are each highly dynamic. As a result, Yotpo was using a lot of their instances On-Demand, causing their cloud costs to be very high.
Zesty’s Solution:
Zesty’s Commitment Manager provides a simple CloudFormation stack that buys and sells Reserved Instances on their customer’s behalf. For Yotpo, this has meant that as they scale down in capacity, Commitment Manager sells their RIs, and when they scale back up, they buy RIs on Yotpo’s behalf.
Commitment Manager achieves this by analyzing the client’s logs on the CUR, CloudWatch and CloudTrail. This real-time data is analyzed and used to train a prediction model based on the specifics of the account. At the same time, the algorithm is analyzing the sale prices of Reserved Instances in the AWS marketplace, so that the most lucrative Reserved Instance discounts can be found. As the client needs more coverage, Commitment Manager grabs very small increments of highly discounted Reserved Instances and as the client uses less compute, these small increments of coverage are shed.
The Result:
Currently, Yotpo is covering on average 95% of their compute with Reserved Instances. This has cut down their compute costs by just less than 40% equating to hundreds of thousands of dollars every year.
For Leshem, this enormous cost reduction has relieved him and his colleagues from a lot of manual labor that would have otherwise been required. Instead, they are able to focus their efforts on developing Yotpo’s products in line with their client’s wishes, and not get sidetracked with cloud management and optimization tasks. In his own words “We love it”.
Customer
Brief
Yotpo is a leading E-commerce marketing platform that provides advanced solutions for customer reviews, visual marketing, loyalty, referrals and SMS marketing. They help thousands of e-commerce companies accelerate their brand loyalty and consumer growth. With a $1.4 billion valuation, Yotpo has thousands of global customers and 600 employees worldwide.
Yotpo launched into the Australian market at the end of 2021 following a multi-year platform collaboration with e-commerce giant, Shopify, and a USD$230 million capital raise to drive expansion.
Key
Challenges
The complexity of forecasting the company’s compute usage was causing engineers to spend a lot of time managing, monitoring, and adjusting cost-saving commitments, but still paying a lot for premium On-Demand Instances.
Key
Results
Over the course of the year since Yotpo deployed Zesty, they have been able to cover around 95% of their EC2 usage and dramatically reduce their costs by more than 40%.
“It’s completely hands-free and we save 40% each month on our On-Demand costs”
To ensure ongoing production across their user base, Yotpo maintains thousands of AWS EC2 machines on which they run their Kubernetes and Nomad-based applications. Their cloud environment is very dynamic and application needs constantly change as their customer base evolves. This makes the task of buying AWS Reserved Instances three years in advance extremely difficult, forcing them to either buy too many or too few commitments. According to Yair Leshem, DevOps Engineer at Yotpo, “it was impossible to accurately predict and purchase what exact usage is going to be three years out when it’s difficult to predict what it’s going to be 3 weeks out.”
In order to minimize this risk, a lot of time was spent managing Reserved Instances so that they could monitor usage and try to more accurately predict how much future coverage was needed. However, as changes are constant, this quickly became a time-consuming task. Engineers were finding it very difficult to keep up with the scale of managing thousands of computing instances that are each highly dynamic. As a result, Yotpo was using a lot of their instances On-Demand, causing their cloud costs to be very high.
“We are able to focus our time on our clients and not deal with cloud optimization”
Zesty’s Commitment Manager provides a simple CloudFormation stack that buys and sells Reserved Instances on their customer’s behalf. For Yotpo, this has meant that as they scale down in capacity, Commitment Manager sells their RIs, and when they scale back up, they buy RIs on Yotpo’s behalf.
Commitment Manager achieves this by analyzing the client’s logs on the CUR, CloudWatch and CloudTrail. This real-time data is analyzed and used to train a prediction model based on the specifics of the account. At the same time, the algorithm is analyzing the sale prices of Reserved Instances in the AWS marketplace, so that the most lucrative Reserved Instance discounts can be found. As the client needs more coverage, Commitment Manager grabs very small increments of highly discounted Reserved Instances and as the client uses less compute, these small increments of coverage are shed.
Currently, Yotpo is covering on average 95% of their compute with Reserved Instances. This has cut down their compute costs by just less than 40% equating to hundreds of thousands of dollars every year.
For Leshem, this enormous cost reduction has relieved him and his colleagues from a lot of manual labor that would have otherwise been required. Instead, they are able to focus their efforts on developing Yotpo’s products in line with their client’s wishes, and not get sidetracked with cloud management and optimization tasks. In his own words “We love it”.
“It’s cloud savings on auto-pilot”
“Zesty was very easy to onboard, I only needed to create an IAM role and it works. I don’t need to do anything to get it to work, it just works in the background.”
Nadav Izraely
Senior Director Platform Delivery
“Zesty has allowed us to reduce our overhead. We now have a zero point deviation from cost KPIs”
Manoj Srikantaiah
Lead DevOps Engineer
“We chose Zesty because of the significant savings that we could achieve with absolutely minimal effort involved.”
David Ting
Senior VP of Engineering
