I am stupid so...
What do they actually do?
Cloud data warehouser Snowflake has raised nearly half a billion dollars in its latest tranche, taking total funding to $923m. Its previous biggest round was $263.5m in January, a mere eight months after raising $105m. Its total cash raised in January was $473m and now it has raised almost as much again – at $450m – in a …
It's a software as a service database - but not just running SQL Server or MySQL in the cloud it's actually a whole new product.
It's actually vety impressive speed and scability wise - especially the concept on decoupling the storage from the compute.
It won't solve every problem but they've done a genuinely good job of imagining what you'd do if you were writing a columnar database from scratch for cloud computing.
"Teradata cloud and the AWS and Azure public clouds but is some years behind Snowflake as a public cloud product." This is debatable.
What isn't though is that Teradata's SQL engine is many years ahead of Snowflake's. Snowflake claims that they have a no tuning needed optimizer. "It just works!" Claims that we have heard from previous defunct vendors (netezza, ParaAccel now Redshift).
Besides that, Teradata now has advanced analytics built into their SQL engine, such as Sessionize, Pathing, Attribution, Time Series and Model scoring. With one Sql, you can join data from S3, Hive and Teradata (many more) data sources, and produce insights such as "what are the most common paths to a client buying a product compared to paths where the cart was abandoned?". No ingesting of data needed, no complicated multi step queries, no landing data into persistent storage. Just one sql!
Snowflake and Redshift are trying to be the best data warehouse when that title has been Teradata's for decades. Teradata is now moving on to what is really needed, an advanced Analytics platform (Vantage) in a new category called Pervasive Data Intelligence.
Yes.. I work for Teradata :)
Snowflake claims that they have a no tuning needed optimizer. "It just works!"
Ha Ha Ha Ha Ha. As the previous respondent said I've heard this all before. Then it turns out it's only scaleable for 'some' workloads. I've lost count of the number of products where I have had to open the box and get dev's to rewrite i/o routines for particular transactions because, heh ho, it turns out that when you are dealing with complex data sets you do need to optimise the queries or invest in a bigger data center to house the additional kit required. in a real world example the vendor wanted an additional £50 million to provide the production platform. Bringing a tuning expert in reduced the cost by 10's of millions of pounds. Unfortunately these skills are getting rarer as the best database specialists are retiring to live in caves and sit on the piles of gold they have accrued in the past 20 years.
Biting the hand that feeds IT © 1998–2020