Org Impact Analysis
Spark based SaaS vs. Snowflake
A Case Study in Managed Service vs. Software as a Service
In the majority of organizations:
HSC – hardware, software, cloud accounts for 30% of the organizations budgetIL/VL – internal labor, vendor labor accounts for 70% of the organizations budget
Yet, when people compare Spark based SaaS vs. Snowflake, they tend to only look at price-performance and fail to recognize the impact to their organization. The decision on the software vendor will provide a savings on impact to your org that is larger than your total spend to either one of the software vendors.
As it relates to performance, out of the box, Snowflake beats Spark based SaaS’s substantially. Snowflake is self-optimized. With a bit of tuning, you can get the Spark based SaaS pretty close to Snowflake. Stay tuned, we will devote an entire case study to price performance in late 1Q24, but our experience and the benchmarks we and others have done so far show a roughly 30% price performance improvement in Snowflake over Spark based SaaS’s platforms.
Functional
While price performance is better with Snowflake, from an impact perspective this functional stuff is the stuff that really matters. Here at Squadron Data, we spent our first 7 years as a Hadoop SI (Cloudera PS #1 Partner in North America) and the last year as a Snowflake Select SI Partner.
We continually heard “but Snowflake is so much easier”.
But how is it so much easier? Where is it so much easier? We took a look at a typical 40 person data and analytics org and looked at the different personas and the functions they performed. Here are the results:
Mid-sized Data Analytics organization (40 people) comparative impact analysis results:
Please note: There will be variance in your organizational savings by moving to Snowflake instead of a Spark based SaaS. Not all companies will have the same savings, but all companies will have savings – and it will be significant. Your savings on impact to your org is larger than your spend to either one of them.
The total difference in the above mid-sized organization is 33 heads vs. 43 heads.
The obvious difference is in the platform administrator and the job optimization. Snowflake there is no platform administrator and no job optimization. But there are large savings beyond these roles.
In all estimates, we use $80/hr (50% onshore, 50% offshore), except the data science we use $120/hr. The numbers shown are annual costs.
Resource Allocation:
We do not recommend removing the 10 heads – good people are hard to find. We recommend placing those folks into the higher value work (delivering answers to the business). Here is how the work delivery of low (platform admin roles), medium (data engineer roles) and high (citizen roles) value roles differ between Spark based SaaS and Snowflake for a mid-sized organization:
