Snowflake’s Data Cloud enables organizations to unite siloed data, discover and securely share data, and execute diverse analytic workloads.
The UK’s Francis Crick Institute says that the use of Snowflake Data Cloud is helping it build secure IT infrastructure needed to support international, multi-participant scientific research initiatives. Use of Snowflake can slash the build time from up to two years to approximately 30 minutes.
But research aside, some Snowflake users encounter difficulties with high prices as their usage grows. In particular, the pricing may be high for SaaS providers and less amenable to their business model.
The First Taste is “Free”
The potential high costs of Snowflake are particularly disconcerting, considering the platform was first launched as a low cost alternative to traditional solutions.
Most of the costs associated with traditional data warehouses such as Oracle or Teradata were upfront license fees of six or even seven figures. Snowflake, on the other hand, gets users in the door with zero infrastructure costs upfront.
The Surprising High Costs of Snowflake
With no barriers to entry, you can get started for nothing. Similar to public cloud infrastructure, your expenditure scales with your usage. It’s logical to increase price as value increases, but only if the increases are proportionate. Unfortunately, that doesn’t seem to always be the case with Snowflake.
In a blog on Medium, a Snowflake Senior Solutions Architect acknowledged the existence of “rumors around Snowflake being an expensive solution.”
Snowflake pricing is based on compute, storage, and cloud services usage. Warehouses are available in sizes x-small to 6X-large, with each tier doubling in cost and compute power. The cost model is both linear and exponential. If you retain the same compute size, with all jobs being relatively small and no heavy lifting, your costs will be linear. However, once you require more power, the jump is exponential, doubling at each tier. Too often, people don’t realize this until they receive the unpleasant surprise of a 5-digit bill for a single month.
Bills Can Quickly Spiral Out of Control
As self-service cloud cost observability platform Finout writes, “Still, if you don’t keep an eye on your setup, monthly bills can quickly spiral out of control.” One response to a query on PeerSpot states Snowflake, “can reduce costs in comparison to other options, such as Oracle.” Well, sure… a Rolls Royce is less expensive than a Bugatti Veyron, but it’s still a tad pricey for picking up groceries.
Great Rates in Some Use Cases
A traditional data warehouse model in which you load data every night might require one hour, equaling one credit. If you shut off Snowflake when that hour was complete, which is typically done, you would pay about $90 per month. Snowflake could be a great fit in this enterprise BI use case, where an analyst loads data from a single source of truth once daily and prepares additional executive reports quarterly.
Other low performance requirements such as testing scenarios are also typically well-served by this model.
And Potentially Not-So-Great Rates for SaaS Providers
In our first example, you could expect significant quantities of idle time, but that goes out the window when your app must be available almost instantly to everyone.
SaaS providers aim to increase usage of their apps. As we wrote in “Making the Cut,” empowering users to create their own dashboards, and making your app useful to more employees within each organization boosts the value of your app. With such changes, your app can become stickier, increasing customer retention, but also making utilization less predictable.
With the need for real time, interactive analytics that are always on and perform well, delivering responsive queries, you’ll likely need those larger environments. What were once reasonable Snowflake costs could potentially jump multiple thousands of dollars per month.
Redundant Queries are Free
Returning to the blog on Medium, a Snowflake Senior Solutions Architect describes how Snowflake uses “a global query results cache to eliminate compute costs for our customers for any redundant queries. This means once a query is executed by a compute, any subsequent identical query sent to Snowflake by any user will no longer need to use additional compute to access the results as long as the underlying data have not changed and the query syntax is identical.”
While this is an excellent feature,it’s much less useful for SaaS providers, as compared to enterprise BI users. Revisiting their goals, SaaS providers often strive to have their customers expand usage of the application, both to more organizations and more users within each organization. You’ll likely see less repetition of identical queries as you empower your users with the ability to build their own unique reports and dashboards.
How to Achieve Savings on Snowflake
Portion Data Based on Urgency
You can significantly reduce your Snowflake costs by using Qrvey and the data synchronization model. The recent release of Qrvey 8.0 includes Qrvey “Live Connect,” which enables real-time data support, making the optimization of data sources like Snowflake and Amazon Redshift much easier, improving performance and lowering costs.
This gives you the flexibility to optimize your usage by deciding what will sync live versus a predefined schedule. If your utilization patterns are predictable, you can optimize by pulling data only during a designated sync period.
Fine Grained Prioritizations
It’s best to allocate your highly skilled engineers to your most technically-complex, business-critical projects. Similarly, by using Qrvey, you can create a data optimization model that allows you to use tools like Snowflake and others to achieve the most cost-effective combination.
With Qrvey, you can achieve fine-grained distinctions, separating out data even down to the chart-level.
Use Snowflake less, shrink the type you want to use, run a small environment sync into Qrvey to run as an analytics layer?