Actian

Figure 2 illustrates the Actian Avalanche architecture. However, this will take some explanation. For example, DataFlow is a high-performance integration engine that was originally developed by Pervasive, while Zen (which was previously PSQL and before that Btrieve) is used in Internet of Things environments as an edge or mobile database, hence the gateway provided. In this context, it is worth commenting that time-series data is not currently supported by Avalanche, though it is supported in Zen. Similarly, geospatial data has not yet been implemented in Avalanche though it is supported in the Actian X database (which is targeted at hybrid transactional and analytic environments).

Vector is the database engine, and it has some notable characteristics. For instance, it exploits Intel’s vector instruction set (hence the Vector name) to process more data elements per instruction (SIMD: single instruction, multiple data) and optimises for L1 and L2 cache. Indeed, there has been considerable academic work involved in the development of Vector, which was created initially at CWI, the Dutch National Research Institute for Mathematics and Computer Science, and the technology incorporates patented algorithms for optimising the use of memory. A further patented technology known as Positional Delta Trees enables real-time updates without impact to query performance. For what it is worth benchmarks suggest that Avalanche has significant performance benefits compared to other cloud data warehouses.

The warehouse understands Spark datatypes and the optimiser treats third-party storage such as Amazon S3 and Hadoop as external tables with the ability to push-down predicates so that Avalanche supports full data virtualisation across these sources rather than mere query federation. User defined functions are available to support machine learning and there is support for R, Scala, Python, and all the usual business intelligence and analytics tools, and TensorFlow will be supported later this year. The ingestion of JSON documents is supported and the company plans to introduce features to optimise performance for the analysis of these.

Avalanche is a relatively new offering, having only been released in March 2019. As such it lacks some features that we would like to see, notably geospatial and time-series capabilities, which will be important for some IoT applications. That said, this sort of functionality are on the roadmap and expected to be released in the second half of 2020. Leaving that aside, the product shows significant promise. The company has a significant history in the database market and knows what it is talking about when it comes to performance, database optimisation, administration and management (Actian FlexPath cloud resource management) disaster recovery, high availability, security (multi-factor authentication) and so forth.

We especially like the approach Actian has taken towards migration from legacy data warehousing implementations, both in the tools that the company makes available to ease this process, and in the way that the hybrid nature of the environment will allow you to migrate to the cloud at your own pace.

The Bottom Line

Actian has always had excellent technology. It has, however, struggled with marketing and visibility. With HCL fully committed to Actian Avalanche, and the ability to leverage that company’s worldwide salesforce, there is every possibility of Avalanche being recognised as a market-leading offering within the cloud/hybrid data warehousing space.

One notable feature of Actian X is its support for a patented technology known as Positional Delta Trees. The following is an extract from the original academic research paper: “our goal is that read-only queries always see the latest database state yet are not (significantly) slowed down by the update processing. To this end, we propose the Positional Delta Tree (PDT), that is designed to minimize the overhead of on-the-fly merging of differential updates into (index) scans on stale disk-based data.” In other words, this is about improved query performance regardless of whatever updates are being made, while preserving consistency. It is also symptomatic of Actian’s approach, in that it leverages a number of patented technologies to maximise performance. In particular, there has been considerable academic work involved in the development of Vector, which was created initially at CWI, the Dutch National Research Institute for Mathematics and Computer Science. As another example, Actian X (the X100 engine) incorporates patented algorithms for optimising the use of memory.

Fig 02 The Actian X ecosystem

The broader ecosystem in which Actian X operates is illustrated in Figure 2. Not shown is support for Kafka and for PMML (predictive modelling mark-up language) though TensorFlow is not supported. Also not shown in this diagram is the fact that there is a managed cloud service available for storing and managing Active X backups; there are a number of geospatial capabilities, including 3D support for R-Tree indexes; and there are database health monitoring and capabilities. Also included with Actian X is DataConnect for Actian X, which provides a development engine for designing and testing integrations, plus a production engine for deployment purposes.

Ingres has a long-established pedigree for supporting mission critical transactional environments, while VectorWise has proven high-performance characteristics in supporting analytics, so we have no question marks over the performance and capabilities of the two acting in combination. And the integration tools and other functions provided by Actian are attractive. For existing Ingres users wanting to add analytic capabilities to their existing environment, Actian X is an obvious choice.

Non-Ingres users fall into two camps: existing users of other relational database systems and greenfield opportunities. We do not envisage that Actian is targeting the former. For greenfield environments, it is important to bear in mind that Actian X is an SMP-based system that scales up rather than scales out. It is not likely, therefore, to be cost-effective for the very largest hybrid environments where you may have hundreds of terabytes of data to process. Where that is not an issue then Actian X may well be worth consideration. If the company wants to address the extreme end of this market then it will need to extend the use of Vector within Actian X to Vector Cluster.

The Bottom Line

If you are an existing Ingres user then Active X should be a trivial decision to support hybrid analytic and transactional processing within the same environment. It is also a strong contender for mid-sized and smaller deployments, especially where SQL and traditional relational approaches are preferred.