Vertica Review

Easy to implement, by tuning the model (projection design) you get great performance


How has it helped my organization?

It enabled delivery of a new Agile Data Warehousing Service.

It enabled us to close large deals. Customers with large data sets had to be migrated from PostgreSQL to Vertica due to performance.

What is most valuable?

  • Clustered database
  • Horizontal scaling
  • Disaster recovery
  • Columnar Storage
  • Compression (you read only columns you need)
  • Immutable storage
  • Fast ingesting

What needs improvement?

Performance of management of metadata layer (database catalog) needs improvement. We still have to have smaller customers on PostgreSQL; Vertica cannot manage thousands of schemata.

Query performance: Improve either Database Designer (automation of projection design) or performance of queries using suboptimal projection design.

Scaling of execution independently on storage: Upcoming Eon Mode (now Beta in Amazon) will hopefully solves this.

For how long have I used the solution?

One to three years.

What do I think about the stability of the solution?

Encountered stability issues three times during last three years.

What do I think about the scalability of the solution?

Suboptimal projection design causes queries to not scale linearly.

The metadata layer does not scale linearly.

Metadata for database files scale okay, but metadata related to tables/columns/sequences must be stored on all nodes.

How is customer service and technical support?

I have experience with legacy vendors of enterprise RDBMS solutions, and I rate Vertica support to be much better.

Which solutions did we use previously?

In my current company I was not responsible for the switch. As far as I know, they switched from PostgreSQL, especially because of performance of analytical queries processing large data.

How was the initial setup?

Just getting Vertica running is straightforward. However, with an increasing number of customers, we had to develop our own tooling. For example:

  • Automated deployment
  • Monitoring, alerting
  • Backup/restore.

What's my experience with pricing, setup cost, and licensing?

Start with license per 1TB. Starting from hundreds of TB there is unlimited licensing to be considered.

Move historical data to HDFS/S3 which are significantly cheaper or even free.

Vertica is delivering more and more features to support load/unload for external storages.

Which other solutions did I evaluate?

2012 - Detailed evaluation including benchmarks of: Greenplum, Vectorwise.

2017 - Evaluation of features and initial communication with vendors, if needed, for: Greenplum, EXASOL, Amazon Redshift, Spark, SAP HANA, IBM dashDB, Snowflake, Azure SQL.

What other advice do I have?

It is easy to implement this solution for one customer. By tuning the model (projection design) you get incredible performance. You won’t face issues with metadata (catalog) layer up to tens of thousands of tables.

It can be a challenge to operate clusters for many customers with varied data pipelines. Consider using Database Designer.

Don't hesitate to push Vertica (through support/product management) to improve it.

Consider implementing your own tools to automate performance tuning tasks.

Disclosure: My company has a business relationship with this vendor other than being a customer: Partner.
Add a Comment
Guest
Sign Up with Email