As a vector database, Pinecone fields hundreds of millions of compute-intensive searches daily. In fact, compute resource needs account for over 90% of its cloud usage across both Amazon Web Services (AWS) and Google Cloud environments.
Additionally, as an early-stage startup that is still growing rapidly, it must be especially mindful of its cloud costs to ensure that its financials are sound and optimized. However, its primary method of managing costs has been to secure discounts through long-term commitments – commitments that are nearly impossible to forecast due to the scale of monthly usage growth.
The company’s compute usage depends almost entirely on the number of customers it has and the volume of searches those customers run. This dynamic usage makes it very difficult to predict the amount of compute they will need from month to month; in 2022 alone, Pinecone’s compute needs have increased roughly tenfold. Given those challenges, it needed an automated management solution that could ensure discounted compute spend without demanding too much effort from the Pinecone operations team.