Snowflake Computing raises $263M for its data warehouse
venturebeat.com
venturebeat.com
Absurd. When I see deals this large, especially in enterprise software, it raises alarm bells. It seems likely there are at least a few “quid pro quos” involved here. When your customers and investors overlap and run in the same circles, it’s hard to avoid the nepotism inherent in that world. The board members, investors, and customers are probably all executives at the same 50-100 companies. They have the power to purchase the product (demonstrating growth), invest in the product (demonstrating increasing valuation), and then acquire the product at a multiple of what they invested in it for.
I also wonder how much the founders are controlling, or if this is a case like Box where the founder owns <10% and the rest of the huge valuation is spread across dozens of different investors.
Now Snowflake has grown 3x over consecutive years, and with all these big enterprises coming in, there has gotta be a true advantage.
Now the question is if this current advantage can survive an upgrade of say e.g. the capabilities of Amazon Redshift, that moves the latter to almost feature parity.
IMO, every company in every industry will need data warehousing at some point in their growth, if they are to survive in the modern world. It's difficult to get right, and Redshift is far from a white-glove experience for those attempting to navigate the space. Compared to B2C deals of this magnitude, this is significantly more sane.
Snowflake is built on AWS, you can be sure Amazon has been watching them for years.
I’ll put my money that by 2020 AWS will have a strong Snowflake competitor. Honestly, Athena/Redshift Spectrum are getting somewhat close.
Why?
If it was a smaller company then spreading resources out across more and more products could result in a drop in quality, but this is Amazon. They have practically unlimited resources. They can afford to put far more in to products than other companies, and they've already got a team who can build some really good stuff. If they can capture what works and use it to understand how to build their products that should result in better products rather than worse.
Scale means you can throw lots of people at something, it doesn't mean you can make it good.
Bureaucracy, internal politics, budget/cost cutting, unpleasant work environment, short-term stock price targets instead of long-term goal based targets... There are a million and one reasons why big companies start to suck.
One of the standout features when I used them a few years ago was their ability to do analytic queries on JSON columns in huge databases. Maybe Postgres is up to the task on it's own now, but MongoDB at the time was not.
https://www.youtube.com/watch?v=dUL8GO4ZK9s&feature=youtu.be
Essentially, their product is (at least) on par with major competitors like Google and AWS, no small feat for a young company.
When a Redshift cluster gets too big: 1. Takes forever to take backups, create read-replicas, modify anything 2. Redshift clusters can still only handle X concurrent queries, no matter how big the cluster is 3. Also Redshift is dragging their feet on adding new features i.e. Active Directory support.
Snowflake clusters allow a theoretically unlimited amount of users to query the same tables in a database - so it scales up forever.
And each team can create their own "compute cluster" which allows teams to track and divide costs much more easily then trying to associate this-batch-user on Redshift with this-team/line-of-business.
Snowflake is well-designed for an enterprise environment where line-of-businesses need to share/access each others data but still have separate infrastructure. Redshift does not make it easy go between Redshift clusters.
Also Snowflake as a company is still hungry and is willing to add features if you pay enough
EDIT: However Snowflake is more expensive than Redshift
EDIT: Splitting data between two Redshift clusters was not efficient for us and trying to have the same table in multiple Redshift clusters requires you creating your own tooling and the data might not stay in sync so the only real option was keeping everything in one massive cluster
Snowflake would just run an EC2 instance in the background for compute, and the very largest EC2 instances were more expensive to run than a Redshift cluster of similar power.
[1] https://blog.fivetran.com/warehouse-benchmark-dce9f4c529c1
Meanwhile "Redshift enforces a query concurrency limit of 15 on a cluster." http://blog.blazeclan.com/what-is-amazon-redshift-11-key-poi...
I'll mark your comment editable for a while if you want to correct this, and can delete this comment if you do.
Even though this is less relevant in a B2B context, because research is usually more thorough, it's still a warning sign in terms of branding when you have a different brand that could be confused with yours: https://snowflakesoftware.com/about/ (as posted by another user, strangely downvoted)
Also, snowflake data models are actually not very common (or appropriate/useful) in data warehouse/mart environments, as they tend to be harder to understand and use on a self-service approach, and require more joins between datasets/tables, unlike dimensional models.