Value Proposition Of The Object Storage
Store petabyte-scale datasets with a minimal overhead. Gunkan is deliberatly minimal, 100% software defined, easy to operate. Gunkan Object Storage is designed to solve the storage problem for simple-yet-frequent use cases, even the smallest ones.
Gunkan fosters the best traits of OpenIO (load-balancing and indirections) and explores others trade-offs. One goal: Gunkan pushes the simplicity to its maximum.
It's About TCO
Why would you choose a solution that costs more, all things equal otherwise?
The key point is the hardware is so expensive that the software that unifies it shouldn't induce significant additional costs. Especially on the aspect of the human resources required to operate the system. Gunkan achieves a low TCO through its inner simplicity, that induces the bare minimal cost of operations.
Why would you choose a solution that is harder to scale, all things equal otherwise?
Scaling Up should be a no-brainer and involve no mandatory data movement, no mandatory rebalancing or restriping. This is the path followed by Gunkan as the experience on OpenIO tells us: avoid any static placement algorithm if possible and prefer a pair of placement/lookup algorithms.
Scaling down should be as easy as possible, too. As simple as moving the chunks of data, and the book-keeping counterpart reduced to the minimum.
Why would you choose a solution that is slower, all things equal otherwise?
When serving stored data, the perception of the performance from the PoV of the application resides in the delta between TTFB and the TTLB. In other words, the latency and the throughput. Not all applications require the smallest latency and the highest throughput, as pleasant as it can be. The idea behind Gunkan is to provide a configurable trade-off between storage vs. compute, index lookup vs. metadata searching. Just deploy useful parts of the system.
Key Characteristics Of the Object Storage
Correct > Simple > Efficient
Because Gunkan puts a maximal focus on correctness and simplicity, it will be the trade-off even when opposed to performance and scalability. This implies that the solutions will probably fulfill mid-sized use cases in terms of dataset size, depending on the expected workload.
Gunkan is a collection of simple tools, where each is dedicated to a single task. If sophistication is present in the implementation, it must and shall remain as simple as possible. The focus is then on a lean design and its correctness, more than any other topic.
Relationship with OpenIO
Gunkan is a lab of OpenIO around the topic of simplifying parts of OpenIO SDS. An initiative led by OpenIO members to incubate tools that will actually benefit OpenIO SDS. A R&D effort to abound, to experiment new ideas that benefit tens of accumulated years of experience but without any inertia related to a technical debt. Any positive result is simply destined to be eventually merged is OpenIO SDS!
Gunkan Is The Solution, Not A Product
Joining a marketing effort to the technical roadmap, since the very beginning, is our way to ensure that the roadmap won't drift out of its goal: provide usable tools. The public exposure ensures the transparency, because we need a maximal feedback. There is nothing hidden.
Focus On Simplicity
When OpenIO focuses on products dedicated to the industry and its most demanding use cases, Gunkan targets simple use cases with an emphasis on the simplicity of operations but less on the performance at a very large scale. Gunkan is also widely open to the feedback of the community.
Community Support Only
Gunkan does not offer more support than the best-effort answering of GitHub issues. Despite this warning, if you still want to run it in production, buy professional services, support with SLA... you should call OpenIO first. So, please, consider visiting the page describing the subscriptions plans or contact OpenIO directly. Solutions exist for everyone!