Comments on: On Column Stores. Interview with Shilpa Lawande http://www.odbms.org/blog/2014/07/interview-shilpa-lawande-2/ Trends and Information on AI, Big Data, New Data Management Technologies, Data Science and Innovation. Tue, 18 Jun 2019 08:07:11 +0000 hourly 1 https://wordpress.org/?v=4.2.23 By: Shilpa Lawande http://www.odbms.org/blog/2014/07/interview-shilpa-lawande-2/#comment-263 Wed, 23 Jul 2014 15:28:36 +0000 http://www.odbms.org/blog/?p=3357#comment-263 Sorry for the delay Stefan and thank you for the questions.

We do not spend a lot of time on benchmarks because, as you’d probably agree, it is possible for any system to be made to look better than any other system in a canned benchmark. Our value is demonstrated by the customer successes we have on their real-life business scenarios. Specifically we’ve never considered comparing ourselves to PostgreSQL per se because we don’t think it is solving the same kind of problem as Vertica is. Postgres is a great open-source database for OLTP workloads but comparing to Vertica would be apples v/s oranges. There is a key principle we take to heart – One size doesn’t fit all, use the right tool for the job. We do not claim to be a general purpose RDBMS and we do not aim to solve the OLTP problem at all.

When Mike Stonebraker founded Vertica in 2005, you might have expected him to base his new analytics database by sharding Postgres which he invented over 20 years earlier. Instead Vertica was built mostly from scratch. We did borrow the SQL parser and semantic analysis, and some client code from Postgres (why reinvent the wheel?) but much of the database engine, anything that impacts performance for analytic workloads is written from scratch – the catalog, optimizer, engine, transaction, recovery, etc. Also, we’ve also deviated quite a ways from Postgres to make the behavior compliant with SQL standards.

Our documentation is public and details all of our functionality including data types, transaction isolation level and SQL functionality. Note that http://www.vertica.com/documentation/hp-vertica-analytics-platform-7-0-x-product-documentation/

Regarding giving back to the community, we offer a Community Edition of Vertica, a completely free version of Vertica, usable up to 1TB on 3 nodes. Anyone can download this version and use it forever. With this version, you get all of the great new featured we’ve developed over the years. Many users take advantage of this offer and use Vertica’s fast analytics without paying us a dime. It’s available at https://my.vertica.com/community/ We also have a number of extensions to Vertica that we share with the community including components to connect to Vertica from Hadoop and Pig as open source.

The Vertica user community also supports us. Our new marketplace (http://vertica.com/marketplace) is full of add-ons to Vertica that will help with data ingestion, visualization and specialized analytics to name a few. These add-on greatly enhance the value of the community version of HP Vertica, as well at the Enterprise version, and many are also free to use.

]]>
By: Stefan http://www.odbms.org/blog/2014/07/interview-shilpa-lawande-2/#comment-262 Thu, 17 Jul 2014 00:39:50 +0000 http://www.odbms.org/blog/?p=3357#comment-262 Interesting interview. I have following questions to Shilpa:
1. Are there benchmarks between Vertica and PostgreSQL?
2. Vertica white papers report they have added float and varchar to c-store’s integer. What about the dozen other data types which exist in RDMS’ like PostgreSQL?
3. How does Vertica cover SQL syntax? What about window functions or CTE?
4. Which transaction isolation levels are supported by Vertica?
5. You (Vertica) took code from #PostgreSQL. What are you giving back to the open source community?

]]>