For the past few months, Snowflake has been establishing its print in the BI world with its unique architecture and cost-effective cloud technology. But many users might not be sure how Snowflake could be leveraged or should Snowflake replace some BI tool in their existing architecture.
In this blog, we are going to compare what would happen if a company has no other DWs other than Snowflake and check if it’s really going to be cost-effective. We are going to assume Company ABC Corp has 2000 users with 3 TB of data with roughly 1~1.5 Gigs of data growing daily, and they are going to utilize Snowflake as their only DW solution. We are going to assume that we have Snowflake Enterprise Edition with AWS server in the US East region. So essentially Snowflake is going to charge them $3/credit. Since they know beforehand their data size, they are going to pay the capacity storage cost upfront – $23/TB.
Cost Consumption in Snowflake:
Loading Part (365 days 3 hour/day)
For the daily load of data into Snowflake, we are assuming that they will consume 3 hours in their warehouse. This we assume happens at the end of the day (non-working hours). They have an independent warehouse setup for the daily loads with a ‘Small’ size. They have chosen 365 days because of the daily load.
Total Compute – 3 (hours per day) * 365 (days per year) * 2 (credits for S Size) = 2,190 Credits per year
|Warehouse||Active hours/day||Days/Year||Credits/Size (S)||Clusters||Total|
Development Part (250 days 10 hour/day)
For the Business Analyst and the developers who are involved in developing the required KPIs or insights derived from the business they are allocating a separate warehouse for development sized ‘Xtra Small’. Because multiple people will be working on it, they will set the number of clusters to 2, providing enough threads. Developers are not going to be there throughout the year, we are taking 250 days and 10 hours per day roughly for their workload.
Total Compute – 10 (hours per day) * 250 (days per year) * 1 (credits for XS size) * 2 (Clusters) = 5000 Credits per year
|Warehouse||Active hours/day||Days/Year||Credits/Size (M)||Clusters||Total|
Based on the business use, we are assuming the data warehouse is going to be running for close to 16 hours per working day.
We will have 3 warehouses one for the C level Executives, another for the Manager and one dedicated for the Data Science team. We are having a Warehouse with M size for Data Science team and a couple of XS for the C level and the Manager level.
XS size – 1 Credits/Hour 2 Warehouses
M size – 4 Credits/Hour
XS size Cluster Compute per year (Executives)= 8 (hours per day) * 260 (days per year) * 1 (credits for XS size) * 1 (Clusters) = 2080 Credits per year
XS size Cluster Compute per year (Managers/Analysts) = 16 (hours per day) * 300 (days per year) * 1 (credits for XS size) * 1 (Clusters) = 4800 Credits per year
M size Cluster Compute per year = 10 (hours per day) * 260 (days per year) * 4 (credits for M size) = 10,400 Credits per year. On combining, we have 17,280 Credits per year
|Warehouse||Active hours/day||Days/Year||Credits/Size (XS, M)||Clusters||Total|
Storage Capacity ($23 per TB per month)
3 TBs of Data is getting stored. 3 * 23 * 12 = $828 per year
Overall Credits/Cost Spent:
2,190+5,000+17,280 = 24,470 Credits Per year so in dollars, it’s 24,470 * 3 = $73,410+ 828 (for storage cost) coming up to a total of $74,238 per year
- Out of 2000 users, there are only 1200 active users. Out of which 800 would be using it as per need basis. We also assume there are only 50-75 concurrent users per time and designed our Warehouse accordingly.
- This architecture is designed in comparison with an existing On-prem Scenario vs Snowflake Warehouse for an imaginary Company ABC Corp so the costs may vary depending on the usage/use-cases.
- The Warehouses can be more optimally designed according to the front-end reporting tool or as per various domains depending on the architecture to leverage the advantages of Cache.
- We have not considered the costs spent for ingesting data into the Snowflake nor the cost of the reporting tool. We also assume that we don’t use Snow Pipe to Ingest as it involves an additional cost.
- The price mentioned is based on our assumption. The costs/credit may vary depending on the exact scenario
- All these depictions of hours are assumed in upper limits. We have data getting ingested into the system every day and users develop/consume the reports round the year.
- The price above is with no discounts, Snowflake may offer discounts which reduces cost further.