Joe McKendrick captures a nice series of thoughts about cloud adoption and maturity of cloud-focused data management.
Cloud database technology may be ready for the enterprise, but enterprises are not quite ready for cloud databases. Even leading cloud database proponents agree that cloud databases are a relatively new-and untested phenomenon.
We are seeing the emergence of data-as-a-service via an implementation that is largely the integration of new, cloud-friendly technologies growing from the big-data-store (HDFS) upwards to high-value real time processing.
If your application is collecting data rapidly, you know the importance and challenge of ingesting, analyzing and making real time decisions as a per-event process. That data is going to end up in HDFS (or another long term store). That storage capability will live in private and public clouds. And the gravity of that destination pulls the technologies that deliver real time value to the cloud as well.
This is why it is critical that your real time transaction platform is designed for virtualize-able deployment.
It all boils down to architecture. An architecture designed first around clustering, replication, and ease of virtualization is critical to cloud deployments, said Ryan Betts, CTO of VoltDB. “This means eliminating expensive shared storage with modern shared-nothing cluster architectures.” In the last several years, there has been an explosion of data management solutions—and with good cause, he said. “The legacy architectures of incumbent products do not meet the elastic, shared-nothing, virtualization, and horizontal scaling requirement for cloud deployments.”