NooBaa recognized that the “SDS hangover” of complex deployment and evaluation, vendor and configuration lock-in represent substantial costs to the enterprise. So NooBaa defined a series of challenging technical goals that would led to an entirely new architecture for storage:
1. The system should be able to utilize any capacity resource, including virtual or physical, large or small, fast or slow, and support this heterogeneity concurrently.
2. The system should accommodate these resources anywhere; in the datacenter, in the public cloud, or over the wide-area, without restrictions on network connectivity.
3. Public cloud should mean more than just on-premise caching or replicating for DR. Customers should be able to arbitrarily combine cloud resources with private resources in any mix.
4. Rather than imposing complex placement and tiering policies, the system should autonomously align data placement with resources, optimizing for capacity, performance, and economics.
5. Ease-of-use is critical. All this should be available in a software-only product that should be deployed by a customer with a 15-minute installation.
At VmWorld, on August 28, 2016, NooBaa announced general availability of the NooBaa product. In order to deliver these critical capabilities, NooBaa implemented a novel architecture based on orchestration concepts rather than traditional symmetric clustering. In the NooBaa system, the control-plane, cluster management, metadata run in a resilient VM environment, completely out of the data path. This central ‘brain’ (the NooBaa-Core) maintains a realtime view of cluster resources through a system of lightweight heartbeat telemetry, and issues instructions over the network to optimize data placement across naive and heterogeneous storage nodes. NooBaa assumes that storage resources could take a multitude of forms; fast or slow, large or small, virtual or physical. Not only can NooBaa accommodate heterogeneity, but it can actually take advantage of resource disparities to deliver a more powerful system.
The resulting product is the most agile scalable storage solution for unstructured data ever developed. Customers can quickly bootstrap hundred of Terabytes of storage capacity using Intel-based host resources that already exist, improving business responsiveness and datacenter utilization. Over the lifecycle of the product, customers can arbitrarily introduce the latest hardware, Intel® CPU technology, storage media, or networking technologies into the system, so customers are never at risk of being locked for potentially years into yesterday’s technology choices. Hybrid and multi-cloud capabilities are seamless and flexible, enabling unrestricted hybrid configurations in response to a customer’s dynamic application landscape.
At the heart of the solution lies the innovative NooBaa-Core logic system, which maintains a realtime view of application IO, data heat-map, as well as resource status. The NooBaa-Core uses machine-learning to solve a complex optimization challenge of maximizing application QOS, utilization, and reliability, while optimizing for storage economics. This includes resource-matching hot data to high-performing resources, as well as data localization; the ability to pre-emptively migrate data to minimize latency to application hosts, a useful capability for follow-the-sun content workflows or wide-area content distribution.
In addition to media archival and distribution, customers are finding NooBaa to be an ideal solution for Splunk data retention, enterprise archival under applications like Commvault and Veritas Enterprise Vault, large-scale retention of science data, and other applications that can utilize NooBaa’s Amazon S3-compatible API interface.
About the author:
Mike Davis has helped define and launch new storage products since 2001, specializing in distributed NAS and Object architectures. Mike runs the marketing effort for NooBaa and you can read more about NooBaa at www.noobaa.com/blog.