With all of the attention focused on the war raging between Oracle and Hewlett-Packard, a significant HP announcement in late June seemed to slip under the radar of the industry press.
On June 20, the company announced general availability of Vertica 5.0, the newest version of the Vertica Analytics Platform, along with some integrated appliance-like bundles combining Vertica with HP hardware.
HP bought the company earlier this year and it looks like Vertica is to be HP’s key play in the burgeoning ‘Big Data’ market.
The foundation of the Vertica platform is a columnar database which, as the name implies, handles data in columns. This column-centric design can yield huge advantages vs. traditional row-oriented databases in certain situations – primarily read-centric data warehouses.
Query performance can be hundreds of times faster with a columnar database, since you only read the columns that you actually need for the query. The Vertica guys see their performance advantage vs. other databases ranging between 50-1,000x - which is quite a wide range indeed.
The Vertica reps repeatedly make the point that their product is designed from the ground up with high performance and high availability in mind. The database was designed to be columnar from the start, not just a column façade grafted onto an existing row-based DB.
Vertica also took the MPP (massively parallel processing) route to provide granularity, scaling, and availability advantages. Each node is completely self-contained and shares nothing with any of the others.
Each node independently runs queries, but can enlist help from other nodes so that a query can use the power of the entire assembled system. It sounds a bit grid-like, but with every node a head node.
With this architecture, adding more nodes adds more capacity and performance in a near-linear manner. Plus a failure in one node won’t take the entire system down; other nodes can pick up the jobs on the failed node and complete them.
The Vertica DB uses standard SQL and supports all of the typical database-y stuff you’d expect. HP- Vertica has also added some advanced functions and a Hadoop connector to round out the big analytics story.
The best place to get the technical ins and outs is at the Vertica website. Clicking around yields a trove of info and a customer list that includes an interesting mix of the new (Groupon, Twitter, Zynga) and the old (Verizon, Comcast, Sprint, AT&T).
You can get Vertica as a software-only package or in an appliance-like bundle on quarter-rack (96-core), half-rack (192-core), or full-rack (384-core) HP blade systems.
Vertica didn’t talk prices in my briefing, but they say they aim to be half the cost of competitive systems while offering several times the performance. Looks like Oracle and its flagship Exadata systems is the main competition.
Vertica isn’t the right choice to replace the OLTP functions in Oracle’s Exadata box but , it could potentially run rings around Exadata when it comes to data warehousing and hardcore analysis. Vertica’s software-only option automatically makes it more granular and flexible than Exadata – and probably significantly cheaper to acquire and maintain.
HP is mounting a multi-front sales push with Vertica, using the sales forces from the Business Critical and Industry Standard server/storage units along with Vertica’s existing sales personnel.
Just as Oracle used a full-court sales press when it introduced Exadata2, I’d expect to see HP doing the same with Vertica. As an independent company, Vertica landed 300 paying customers ranging from small to damned big. Combined with HP’s reach and range, Vertica could end up in thousands more data centers.