What is most valuable?
The most valuable feature is the merge function, which is essentially the upsert function. It's become our ELT pattern. Previously, when we used the ETL tool to manage upserts, the load time was significantly longer. The merge function load time is pretty much flat relative to the volume of records processed.
How has it helped my organization?
HP Vertica has helped us democratize data, making it available to users across the organization.
What needs improvement?
We've had issues with query time taking longer than expected for our volume of data. However, this is due to not understanding the characteristics of the database and how to better tune its performance.
For how long have I used the solution?
We've been using HP Vertica for three years, but only in the last year have we really started to leverage it more. We're moving to a clustered environment to support the scale out of our data warehouse.
We use it as the database for the our data warehouse. In it's current configuration, we use it as a single node, but we're moving to a clustered environment, which is what the vendor recommends.
What was my experience with deployment of the solution?
We had no issues with the deployment.
What do I think about the stability of the solution?
We've had no issues with the stability.
What do I think about the scalability of the solution?
We've had no issues scaling it.
How are customer service and technical support?
I'd rate technical support as low to average. The tech support provides the usual canned response. We've had to learn most of how to harness the tool on our own.
Which solution did I use previously and why did I switch?
I haven't used anything similar.
How was the initial setup?
HP Vertica was in place when I joined the company, but it wasn't used as extensively as it is now.
What about the implementation team?
We implemented it in-house, I believe.
What other advice do I have?
Loading into HP Vertica is straightforward, similar to other data warehouse appliance databases such as Netezza. However, tuning it for querying requires a lot more thought. It uses projections that are similar to indexes. Knowing how to properly use projections does take time. One thing to be mindful of with columnar databases is that the fewer the columns in your query, the faster the performance. The number of rows impacts query time less.
My advice would be to try out the database connecting to your ETL tools and perform time studies on the load and query times. It's a good database. It works similar to Netezza from my experience but it is a lot cheaper. Pricing is based on the size of the database.