Costco Travel Cuts Snowflake Costs in Half with Optimizations from Keebo
Challenges
- Increased processing and compute costs due to frequent data changes and fixes
- More advanced reporting required due to growing workload and tightening SLAs
- Limited data engineering resources due to small team size and ongoing industry-wide, pandemic recovery
Results with Keebo
Costco Travel offers its members exclusive savings on top-quality vacations, hotels, cruises and rental cars. Amidst ongoing, industry-wide pandemic recovery, Costco Travel faced tightening SLAs and minimal data engineering resources with which to meet them. By implementing Keebo’s AI-powered Snowflake, Costco Travel reduced Snowflake costs by 50%, freeing up engineers for more user-centric tasks.
Before using Keebo, we were constantly struggling to find the right balance between cost and performance. Traditional optimization methods required constant monitoring and manual adjustments, which took away from our core focus. With Keebo, we’ve automated this process, and the results have been significant. We’ve seen cost savings while maintaining peak performance without having to continuously tweak settings. It’s been a game-changer for our small team in a demanding environment.
Bhaskar Maitra
Data Platform Manager
Optimizing with Snowflake’s advice
Initially, Costco Travel began implementing Snowflake’s boilerplate cost optimization advice: write better queries, monitor performance parameters—bytes scanned, execution time, query load percent, spillage, and queuing time—and adjust warehouse sizes manually.
The problem: Costco Travel’s data engineers had no reliable way to predict optimum warehouse sizes without constant monitoring. Not only was this a waste of precious data resources, but it detracted from more user-focused activities like data pipelines and reports.
Machine learning-powered optimizations
To solve this problem, Costco Travel implemented Keebo’s artificial intelligence and machine learning optimization tool to dynamically adjust warehouse size and auto-suspend. If Keebo determined a smaller warehouse could handle incoming requests with negligible impact on performance, it would scale down the warehouse until query and user volume ramped back up. Likewise, if Keebo determined a low likelihood of warehouse usage before the default auto-suspend period, it would preemptively suspend the warehouse, saving additional compute costs.
Reducing Snowflake costs
By implementing automated optimizations across Costco Travel’s Snowflake footprint, Keebo reduced costs, enabling Costco Travel to do more at higher quality. These optimizations added up to a 50% reduction in overall Snowflake costs. Given the lean nature of their data team, these cost reductions enabled increased capabilities, not only in terms of data processing, but freeing up human resources for higher impact, user-centric activities.
Keebo’s automated optimizations have been a huge win for us. We saw a 50% reduction while empowering our team to focus on delivering value to our customers. The ability to maintain performance without manual intervention has completely transformed how we manage our Snowflake environment.
Matt Browning
AVP of International & Business Services
Author
