Introduction

If you are a user of DataStax Hyper-converged Database (HCD), then you should know that there are some exciting changes coming with the 2.0 release! In this short post, we will discuss some of these changes and explain how users will benefit from them.

What is changing?

These changes are coming about as a result of HCD’s code being rebased with Apache Cassandra® 5. This brings some new capabilities into HCD:

Let’s take a quick look at these features.

Benefits

These changes and new features have several benefits. Some of them may be easier to see than others, but they all work to make HCD 2.0 more scalable and performant.

Paxos v2

Paxos is the consensus protocol used by Cassandra’s lightweight transactions (LWTs), and Paxos v2 brings several improvements. First, it relies on fewer inter-node messages. By comparison to v1, Paxos v2 also significantly reduces the number of required network round-trips, and enables automatic repairs for LWT data. This improves both the performance and consistency of HCD applications relying on LWTs.

Advanced Indexing

HCD uses the latest storage attached index (SAI) from Cassandra 5, as well as an updated version of the JVector library. This gives HCD the ability to support more advanced text queries, as well as AI/vector workloads at a large scale.

Note: JVector has been “battle tested” at a large scale with DataStax Astra DB. More information on JVector can be found in its GitHub repository.

Java 17

Like Cassandra, HCD requires a Java Runtime Environment (JRE). Running on Java 17 allows for several performance improvements around garbage collection, specifically with the garbage-first collector (G1GC). Additionally, the Z Garbage Collector (ZGC) becomes available providing low latency with minimal configuration or “tuning.”

Binary Tarball Downloads available

HCD is now available for download through IBM Passport Advantage. This method of installation provides a file package which can be extracted and run in a few simple steps. This allows HCD to be run in the cloud, on bare metal, or on a virtual machine - without Kubernetes as a hard requirement.

mTLS authentication with Role Based Access Controls (RBAC)

This feature essentially allows connections to a Cassandra cluster to authenticate with a (mutual) TLS certificate, negating the use of the username/password credentialed approach. In this way, the cluster identifies and authenticates a connection through its certificate, and maps that certificate to an internal database role or a role defined in LDAP or OIDC. This is a great way for applications to securely connect to HCD without having to rely on the traditional credential approach.

Dynamic Data Masking (DDM)

Dynamic Data Masking allows users to hide sensitive data at the database level. DDM allows for the obfuscation of specific columns on a table, so that unprivileged roles will see a redacted version of that data. This is one way that HCD provides secure, at-rest data protection.

Conclusions

In this short post, we have given a high-level review of the changes coming to DataStax HCD 2.0, as a result of rebasing the code with Cassandra 5. These changes bring some exciting functionality into HCD, driving improvements in performance, scalability, and security.

If your enterprise Cassandra environment supports business critical AI applications at a large scale, then you should consider DataStax HCD. Have a look at the HCD product page to see how to get started.