Micro-compute network
Features
Use & Earn
▾
Hardware
Blog
About
FAQ
Contact
Talk to Sales
Get Started
KOVA
/
FAQ
Frequently Asked Questions
If you’re evaluating Kova, these answers should reduce ambiguity fast.
How is Kova cheaper than renting whole instances?
Kova bills per second and packs fractional slices. You pay for delivered work, not for idle headroom you were forced to rent.
What if my job gets interrupted?
Schedulers can retry and re-pack. Settlement is based on verified units of work, and you can set caps and policies per job.
Can I bring my own cloud or bare-metal?
That depends on deployment mode. Architecturally, Kova is interop-first: storage, CI/CD, and observability integrations are expected.
Do you support fractional GPUs?
Yes. MIG/MPS-aware scheduling is central: fractional GPUs must be reliable, isolated, and measurable—or the economics break.
What’s the settlement model?
General model: users pay for verified work; providers earn from completed units. Exact rails and token mechanics can vary by rollout.
How does reputation work?
Reputation policies score reliability and output quality over time. It influences routing and reduces incentive for low-quality behavior.
Is Kova decentralized?
The network is distributed by design. Degree of decentralization depends on the policy layer and deployment—verification keeps it accountable either way.
