Supply Chain Software Leader Uncovers Snowflake Cost Savings & Optimizations
Customer's Challenge
A celebrated supply chain software provider focused on aiding government agencies and corporations in navigating risk and compliance sought a partner to help uncover Snowflake cost savings and optimizations related to their end-user experience.
phData's Solution
phData proposed five suggestions for the customer’s Snowflake environment that would collectively produce cost-savings for the organization while also improving outcomes for their end users. The suggestions included modifications to Snowpark usage, a decrease in warehouse size, a switch to transient tables, the replacement of atomic inserts with batches, and the removal of ineffective clustering.
Results
phData closed the customer engagement by providing them with five clear suggestions on how to take advantage of additional cost-savings and optimizations in their Snowflake environment. Once implemented, the suggestions would set the customer up with improved processes that could positively impact the end-user experience and provide them with over $100,000 in annual savings.Â
The Full Story
What began as a customer’s quest to improve end-user reporting experiences resulted in a complete analysis that also revealed cost savings and optimizations for the customer.
The client began the project focused on improving its dashboard reporting speeds for its end users. phData brought in a team to identify use cases and test reporting times. For the most part, reporting speeds on the majority of the use cases were optimal. There were some reporting time issues noted that the customer was encouraged to address directly with their third-party dashboard provider.
In addition to providing advice on end-user expectation setting on reporting speeds, phData set out to deliver the customer impactful Snowflake cost-saving optimizations.
Why phData?
The customer is a long-time user of Snowflake and started by first reaching out to them to uncover optimizations for their end-user reporting experience. Snowflake referred the customer to phData for a deeper analysis of their entire environment, knowing phData’s extensive experience in scenarios like this.
phData Uncovers 5 Optimization Suggestions
Discontinue Using Snowpark-Optimized When Unnecessary
phData noted that some of the customer’s warehouses were switched to Snowpark-Optimized, in an effort to increase performance. This strategy can be helpful if the warehouse is experiencing memory pressure, which reveals itself through excessive spillage to a remote disk. However, none of the customer’s warehouses were experiencing memory pressure.
Determine an Appropriate Warehouse Size
phData did an analysis of reporting speed times correlated to warehouse size. The customer was presently utilizing a 2XL warehouse. Because there were no cases of excessive spillage to a remote disk, and processing speeds were nearly the same, phData recommended that the customer decrease their warehouse size and save the difference.
Switch to Transient Tables
The customer had implemented some good practices in reducing storage costs like only using time-travel for critical tables. Additionally, the customer could convert non-critical tables to transient tables to further increase the cost savings.
Remove Ineffective Clustering
The majority of the client’s tables with cluster keys did not meet the recommended size requirements of 1 TB or larger. Although the customer’s costs could not yet specifically be tied to this factor, clustering should be monitored closely over time. Improper clustering will increase serverless compute costs.
Replace Atomic Inserts with Batch Processing
phData uncovered instances where developers had written the customer’s code as atomic inserts rather than batches. Snowflake was designed for batch processing, and atomic inserts increase the customer’s overall cost and reporting speed.
Reference Diagram
Take the next step
with phData.
Learn how phData can help solve your most challenging data analytics and machine learning problems.
