As an employee new to Actian I decided to dig into what makes Actian Vector a star performer. Three specific qualities, covered below, caught my eye as I reviewed the technical overview.
Vectorization: When I first heard this term, my memory went back to 30 years ago, when IBM offered my employer, Watson Calculating Services Limited, a free trial of the Vector Facility for our ES9000 mainframe. It was cool because we saw massive improvement in our FORTRAN applications without having to rewrite them. A simple compiler directive was all that it took to take advantage of vectorization. So, how does this extend to what Vector does? Actian have applied the techniques developed from the acceleration of floating-point operations and high performance computing using specialized hardware to accelerate database workloads. The result is 100x performance improvements without specialized hardware. Actian provides these performance improvements on industry-standard Intel x86 architecture server processors, transparently, without having to rewrite standard SQL queries.
Hybrid Column Store: Relational databases store data that is optimized for row at a time access. However, for fast analytics on a subset of columns, storing data in a compressed columnar format is the way to go because analytics workloads in traditional data warehouses tend to use de-normalized tables to optimize read performance, but rarely analyze whole rows. Vector goes a step further, by optimizing the in-memory block format to minimize cache misses. This boosts memory access speeds to maximize performance.
Positional Delta Trees: Allowing incremental changes, while maintaining transcational read consistency is a tough challenge for columnar databases. Actian Vector maintains full multi-version read consistency, so every new transaction will see all previously committed transactions, so you don’t have to rely on large bulk data loads alone for updates. Actian Vector’s Positional Delta Trees (PDTs) store small incremental changes, as well as updates and deletes, so queries run lightning fast and any calculations add up despite changes that occur while the query executes.
In my judgment, these are some of the many qualities that make Vector stand out from the crowd.
Want to know more? Then, visit us at the Actian’s Hybrid Data Conference at London’s Amba hotel on the 9th of November to discuss in person with Actian engineering, executives, customers. Check out the full agenda and register here. It’s free to attend.