Technical and economic design of L2 proof of storage on the Dynamo blockchain

Introduction

Dynamo is a proposal for a hybrid proof of work / proof of stake cryptocurrency which incorporates native NFTs and smart contracts. Early in the project’s lifecycle many community users requested proof of storage as an alternative security mechanism. After further discussion, it was collectively decided to implement a utilitarian distributed storage facility which would allow third party users to purchase “pay as you go” decentralized storage on based on consumer grade hardware and hosted by the Dynamo community members. This document provides an overview of the technical design and economics of the distributed storage model which Dynamo will implement.

Storage design

Dynamo Distributed Storage, DDS, will be based on an (at least)8 way mirror RAID 5 design. Each RAID 5 instance will be based on 4 virtual drives, for a total of 32 distributed drives per logical drive. In order for data loss to occur, one virtual drive from every one of the 8 mirrors would need to fail simultaneously. Typically failures are measured based on MTBF on hardware, however in distributed storage systems, failures are expected to occur frequently due to many factors beyond simple hardware failure such as local internet outages, changes in individual preferences of storage providers, accidental power down of consumer grade equipment and malicious actors seeking to disrupt or challenge a project’s legitimacy.

RAID5 is superior to RAID6 and xfs based RAIDZ1/Z2 because of its simplicity in implementation. When considering consensus based rewards for distributed storage, simplicity is critical for time to market, development and debug complexity, security attack vector surface and support.

RAID5 also allows for rapid resilver due to the loss of a node. This is critical in distributed systems where nodes may go offline/online in short periods of time. A contemplated DDOS attack is simply turning on/off a node in quick succession to force multiple resilvers, which would adversely impact IOPS of the overall system. Therefore, it is important to implement an “at least” 8 way mirror, which can effectively ignore nodes with high availability hysteresis.

Storage will be available as a layer 2 service based on Dynamo mainnet. Distributed storage requires high bandwidth, low latency nodes which are fundamentally incompatible with the fault tolerant, consensus based peer to peer node communication of Dynamo. In order to offer commercial grade SLAs it is necessary to decouple storage access from the transactional distributed ledger. Payment for, and verification of, storage can still be accomplished in an anonymous and decentralized manner, however it cannot occur directly on mainnet.

Interface design

Distributed storage should be available in at least 3 forms.

1 — Consumer oriented web based document storage and archival backup mediums. This would take the form of a web based interface similar to Dropbox or Google drive or whole machine backup applications such as Carbon Copy. Both interfaces can leverage existing open source technologies which act as gateways to managed storage provided by third party storage aggregators.

2 — File based storage expressed as NFS/SMB shares or S3 compatible interfaces. These services are best offered via brokers who offer registration and authentication services. This would be the primary commercial interface used for cold storage/archival purposes.

3 — Direct virtual disk access via iSCSI interfaces. This would be a premium service offered by data storage brokers with high bandwidth and negotiated SLAs with storage providers.

Marketplace connector

The Dynamo Foundation will publish publicly available connectors between the Dynamo storage network and raw storage interfaces, NFS, SMB and iSCSI servers for use by storage marketplace vendors. The foundation will also create a public reference storage platform which can be used by third parties to create storage as a service platforms.

Economics

Storage providers can offer some amount of storage available for a period of time. Storage consumers can offer to purchase some amount of storage of a specific price. Storage brokers will act as intermediaries matching buyers and sellers of storage, taking a spread to make a market and create SLAs. Based on market conditions, buyers and sellers will be able to offer or sell their storage

Conclusion

This article presents a proposal for decentralized and distributed fault tolerant storage which can be be offered as a commercial product via storage marketplace brokers who interface with storage providers (both consumer and commercially based) and storage consumers (also both consumer and commercial). This design allows for a robust third party market that enhances the overall value proposition of Dynamo Coin and creates tangible economic value for storage providers, brokers and consumers.