I’m terrible at keeping secrets, but I’ve managed to keep quiet the fact that Actian Vector 3.5, the new branding for the former Vectorwise product, has been available for download from http://esd.actian.com for almost two weeks. There are a lot of cool new features in this release and as the product has matured over the past four years we’ve made major strides in terms of product stability and availability. Concurrent with Actian Vector 3.5 we also launched Actian Director 1.2.
In a world where people expect instant gratification, the fact that it could take several hours to load data into an Actian Vector instance before they could start playing with their new toy was something we needed to address. To reduce the time to load data we’ve added data partitioning and parallel load capabilities. In the engineering lab we’ve seen the data load times drop to 20% of what they were in our prior release.
In the case where we need to spill to disk there has been a noticeable degradation in performance. In Actian Vector 3.5 we’ve focused on being smarter about how we tackle this to minimize the performance differential. Some of the changes include Partial Hash Join spilling to decrease step-like threshold when starting spilling; improving disk spill storage to use compression and improving disk spill block management in the buffer-pool to reduce trashing for concurrent queries. We’ve also implemented spill to disk profiling that will tell us things like how many partitions were spilled to disk; how many rows were spilled to disk; how much time it took reading / writing / distributing etc.
Another new feature that may help with performance is the introduction of secondary indexes in Vector 3.5 to speed up certain point queries and joins by leveraging improved memory-resident secondary indexes.
High Availability, Extended Functionality, and Improved Robustness
As part of the maturation of the product we’ve seen an increased demand from our customers, who are running Actian Vector in 24×7 production environment, for high availability features to improve overall database management and availability.
Actian Vector 3.5 introduces incremental backup to enable users to perform incremental backups while the database is in use. And, unlike traditional incremental backup techniques, Actian Vector uses block level intelligence to reduce recovery time to an absolute minimum.
We’ve also included a new utility called movedb to facilitate database cloning. This utility was developed at the request of a customer who wanted to be able to clone a production database to a development or test environment, but it has broader applicability.
As with all new releases we’ve added some new SQL functionality. One of our key customers asked us to implement an “UPSERT ” function i.e. update a value if it exists, insert it if it doesn’t. When we referenced the SQL standard we found that the standard incorporates a MERGE function which provides broader functionality than just this “UPSERT” requirement, and I suspect this will become a new favorite for simplifying SQL code. The new FIRST_VALUE and LAST_VALUE functions return the first or last position of each result row within a partition or group. These functions can be used as window functions or in a grouping specification We’ve been banging our heads off the range variable limit for a while now and have bumped this up to 382 to allow for the page long queries that BI tools generate.
Actian Director 1.2
For those not familiar with Actian Director it is an intuitive Database Administration Tool built for Actian Vector and Ingres. Actian Director 1.2 introduces a new data import wizard which is the simplest and fastest way to import data from a flat file into Actian Vector.
To download Actian Vector 3.5 please go to http://esd.actian.com
Don’t forget to use the Ideas Center to provide input into the planning for future Actian Vector and Actian Director releases. For other questions, email us anytime at email@example.com
I look forward to hearing your feedback!