Snowflake Datawarehouse vs Teradata

Data warehousing is evolving, there is a fierce battle now among the tech giants. Snowflake is a new entrant (4 years old) in this space. Teradata is quite rightly known as a powerhouse in the data warehousing and analytics arena. Teradata was the industry leader in Data warehousing solutions for over a decade.

Before you dive deep into this article, A little bit about my background. I have 10 years of experience in data warehousing working on Teradata. And I have only spend a couple of months in Snowflake.

Let's talk about the similarities first:

• Both Teradata and Snowflake are Massive Parallel processing (MPP) systems. Which is a fancy term that implies the load is split evenly across the available compute nodes and parallelly executed
  
• Both Teradata and Snowflake run on ANSI SQL
• Both Teradata and Snowflake use their own patented proprietary technology (Secret sauce) that determines how the data is distributed and retried for the MPP
• Both Teradata and Snowflake has proven their ability to handle massive amounts of data

So, What's the difference between Teradata and Snowflake?

• When you purchase Teradata, They literally ship the hardware and software in a truck that is mounted on-premises (Teradata does have a cloud solution, but doesn't seem to be quite popular). Whereas Snowflake is cloud native, Everything resides in the cloud, the data, the software and even the SQL client that we use to access the snowflake warehouse. It literally takes few clicks to create a new data warehouse and spin up additional instances 
  
• Teradata uses proprietary hardware and software. Snowflake uses AWS / Azure hardware and its own proprietary layer to manage the resources and users
• Teradata ships with fixed capacity, if you need to expand, you may want to deal with Teradata to purchase additional hardware, whereas snowflake comes with UNLIMITED Storage and COMPUTE
• Snowflake stores the data encrypted by default, whereas Teradata doesn't claim any encryption
• The Hardware appliance in Teradata has to be shared between the ETL and the Reporting teams. Snowflake uses virtual warehouses, where each team can create its own virtual warehouse for their needs (The underlying data is shared)
• Teradata is Always ON. Snowflake goes into sleep mode when idle and thereby saving $$$

Show me the money $$$

Snowflake claims they could save you thousands of dollars when you migrate from Teradata. Snowflake claims that could get the performance of a 10M$ Teradata appliance with 100$ in Snowflake.  Snowflake charges an on-demand price of just $40/TB per Month or a pre-paid price of only $23/TB per Month for storage. (Compute is charged per usage). Whereas Teradata’s storage pricing starts at 1000s of $ per TB (https://www.teradata.com/Resources/Brochures/Teradata-Workload-Specific-Platform-Pricing/)

Snowflake simply charms with its simplicity

• Compression is enabled by default
  
• All the data is encrypted by default (Even Amazon/Snowflake itself can't read your data)
  
• No more bothering with Indexes, Collect stats, Primary keys. They aren't even available in Snowflake. You simply load the data and Snowflake just performs without any of that geeky stuff
  
• Snowflake is Columnar by default, so it only cares about that 5 columns in your report even if your table has a zillion columns
  
• Your ETL / Testing teams need a clone of Production data for testing? No problem its just a command away (and it does that without even duplicating the data)
  

Snowflake effortlessly equals/performs better than Teradata

In a research performed by GigaOM, Snowflake tops the chart in the data warehouse solutions race.

The databases are assessed in the following parameters and snowflake topped the charts with the highest score of 4.85 out of 5

• Robustness of SQL
  
• Built-in optimization
  
• On-the-fly elasticity
  
• Dynamic Environment Adaption
  
• Separation of computing from storage
  
• Support for diverse data
  

PS: I am not affiliated with snowflake /  Teradata

How is your experience with Snowflake /  Teradata? tell us in the comments

nVector

posted on 17 Oct 18