Joining us this week is Nick Alesandro, VP of Product at Overclock Labs; creators of The Akash Network.
About Overclock Labs
- We believe the Cloud should be distributed and decentralized so that no one provider can control the internet.
- We believe the cloud should be globally fault-tolerant to avoid any single points of failure.
- We believe the Cloud should be simple, automated, and accessible for all.
About Akash Network
Decentralized protocol for provisioning, scaling and securing cloud workloads: The world’s only on-chain auction marketplace for off-chain container deployments
Highlights
- 0 min 37 sec: Introduction of Guest
- 2 min 28 sec: High level description of the technology
- Cloud centralization is a problem; it needs to be decentralized
- Take over 100% or a partial amount of a machine added to the cloud network with special CoreOS based system
- 5 min 13 sec: Akash Network
- Donate your idle servers into an available system for usage in a cloud managed by Overclock Labs
- 2 Components – Blockchain for marketplace / Deployment platform
- 6 min 52 sec: How does marketplace work; who wants to use this network?
- Focus is on developers who want to use these systems
- #1 Reason – Cheaper than standard public clouds ; #2 Reason – Its distributed globally not at fixed known sites
- 10 min 26sec: Blockchain as decentralized ledger to avoid central store
- Join the infrastructure without a formal registration on a single database; only listening for bids not publishing what they offer
- History of the infrastructure is public for customers to evaluate
- 12 min 31 sec: How do I know where I am pushing my workloads? Can I trust the infrastructure provider?
- Issues arise with receiving workloads that are unknown to you
- 16 min 38 sec: What do I do to add a rack of servers into Akash network?
- What you need to do vs what you should do
- Management Server, Network Isolation, Monitoring
- Seasonal Load Model ~ Electric Grid Analogy
- 20 min 23 sec: Identify Geography and Latency to Customers?
- 21 min 23sec: What do I ensure I am not getting dropped or pulled into a bidding trap?
- Workload goes down it automates a bid elsewhere in the network
- Conditions can be set should you need a long running workload and it is terminated by the host
- Do you expect providers to monitor machines or do you as a service?
- 27 min 45 sec: Kubernetes Cluster across providers?
- Kubernetes Federation
- Why Kubernetes? Writing our own Kubernetes using Kubernetes
- 31 min 55 sec: Why not do this with Virtual Machines?
- Containers makes sense
- 32 min 40 sec: How long have you been working on this project?
- Why of the project? DISCO → Scheduler
- Decentralization is the key
- Can build a private infrastructure using their system
- 37 min 01 sec: Wrap-Up