Zebrium uses machine learning to automatically catch software incidents and show root cause. Their software learns to detect hotspots of abnormally correlated anomalies across logs and metrics without requiring any human setup.
As a quickly growing AI startup, they need to keep costs low while maintaining a fast pace of innovation and the reliability of their core AWS service stack. This is not an easy task, especially when AWS reservation options force hard tradeoffs between reliability, flexibility and savings.
To achieve their target savings along with the consistency & flexibility they needed, they had been using the new Compute Savings Plan, meaning they had to lock in to a certain AWS spend level for 3 years. Because of their pace of growth it was difficult to predict what fraction of their On-Demand usage to cover with the Compute Plan and even harder to forecast where they could take full advantage of the much higher savings rates of less flexible Standard EC2 Reserved Instances (RIs).
By shifting their strategy to leverage Reserved.ai Lease RIs they were able to double coverage of On-Demand usage, plus get the savings that come from a 3-year Standard RI while avoiding almost all of the lock-in. After an initial 30-day window, they can trade or sell the reservation back to Reserved.ai at any time, no strings attached. Lease RI gives you the best of both worlds: the highest savings and the most flexibility.
Starting in March of this year, Zebrium took advantage of Lease RIs for covering m5.large and r5.large EC2 instances. In contrast to the Compute Savings Plan, they’re only on the hook for the first 30 days of their reservation (AWS enforces a 30-day wait after you purchase an RI before you can resell it). After that, Reserved.ai can take the Lease RI back from them at any time.
This means they’re enjoying the full savings of a 3-year RI with only a 30 day commitment. In other words, their capital commitment is only 3% of the total cost of the contract.
So far they’ve achieved a 56% saving over the same instances On-Demand.
To get flexibility close to an RI Lease with native AWS options they could use a 1 Year Compute Savings Plan, which would have been a better deal than On-Demand, saving 28%, but the payment commitment would still be twelve times as large with only half the savings achieved compared to RI Leasing!
Zebrium has used RI Leasing to save 56% compared to On-Demand and save 40% compared to their existing 3 year Savings Plan while maintaining greater flexibility to resell their RIs at any time.
“Reserved.ai Lease RIs let us take advantage of the highest possible AWS saving rates while ensuring we take on none of the commitment or lock-in risk. It has truly allowed us to move even faster than with native AWS reservation options and at a materially lower cost.” said Rod Bagg, VP of Engineering at Zebrium.