Recoverable File System improves business continuity in upcoming release of UniVerse
For some time now, UniVerse customers have been asking for a high-availability Recoverable File System (RFS) option to complement Replication Disaster Recovery. Well, we’ve been paying attention. To help ensure that your business-critical applications survive unforeseen disasters, we’ve re-architected the RFS currently available in UniData for UniVerse.
New and improved UniVerse architecture and new and improved RFS
Our first step was to ensure that the vast amounts of data and transactions you’re dealing with today won’t create bottlenecks. The next step was to speed up transactions. To achieve those two important goals, we moved the processing to the backend where it wouldn’t impede performance. In UniVerse 12.1, tight integration of RFS with the database engine results in reduced latency. And with more of the application residing in shared memory, an application with multiple users should experience improved performance.
One of the guiding principles we used for rearchitecting was to take stand-alone features, like locking, and integrate them into the solution. The new architecture also helps meet modern compliance and security requirements as it no longer requires elevated privileges, such as root, to run. Instead, 12.1 uses DB user privilege to access the functions and data it needs to protect and preserve your data. In the future, businesses will be bringing large scale, data-intensive systems up and down in the cloud and this performant new architecture will scale to fit your needs.
Benefit of RFS
RFS protects your data with a log that will replay/recover data to a particular point in time to preserve transactional consistency. If a transaction is half done it’s rolled back. You’ll never recover corrupted data. And if RFS detects a problem, it will bring the system to a graceful shutdown, so the DevOps team can identify the point in time of the failure and recover to that point.
We’ve developed a two-process architecture that, in essence, brings requests in and then hands them off to backend processes which use a series of buffers to manage the data. These buffers help maintain before and after images, so we can guarantee consistency. If there’s a crash, RFS uses the logs to help the DevOps team recover the system without worrying about restoring corrupted data.
RFS has also been enhanced to work hand-in-hand with replication so the actual replication logs are brought under RFS protection, allowing you to control and secure both your publisher and subscriber environments.
The new architecture protects your UniVerse system from inadvertent kills, hangover or leftover files, pointers, or anything that can cause “garbage” files.
UniVerse 12.1 Beta Program—coming soon
For the last 12 months, as part of the development and testing phases for UniVerse 12.1, we worked with a large partner in Australia to prove that you can upgrade seamlessly from older versions of UniVerse to 12.1 and ensure acceptable performance. We used the partner’s application, their data, and their performance benchmarking tools. Using a real application, with real data and benchmarks, ensured that we were meeting real-world objectives. This experience proved that applications will run as fast or faster on 12.1. We are eager to share aspects of this hard data at our partner’s user conference in September in Australia, and with our UniVerse partners during our upcoming 12.1 Beta program.
This new architecture is truly hardened, scalable, and built for the future. It completes our HA/DR offering for business continuity.