*Note: This review and score is purely based on the information disclosed by the validator service and the scoring rubric.
Last Updated: Oct 12, 2019
B-Harvest is a validator service based in Korea. The team values a high standard of security, stability, and active community participation. B-Harvest was the first validator to create a passing proposal in the Cosmos network.
- IRISnet (testing)
Team Background (75/100)
- Full-Time/Part-Time (10/10)
- Prior Blockchain Dev/Impact (10/10)
- Systems Experience (5/10)
- Recognizability (5/10)
Current Voting Power (77/100)
- Total Staked: (8/10)
- Unique Self-Bonders: (10/10)
- Commissions: (5/10)
Historical Metrics (90/100)
- Uptime (8/10)
- Proposals (10/10)
- Legal Compliance/Insurance (0)
- Innovations (0)
B-Harvest, based in Korea, was started in February, co-founded by Lee Hyung Yeon and Dongsam Byun. Yeon has 10 years prior experience in investment banking as an algorithmic trader and risk manager. This has given Yeon extensive skills in low latency environments, big data analysis and machine learning applications. The CTO, Byun, has a background in security and blockchain. Byun has worked as a product manager for a prior blockchain project and has contributed to notable technologies such as Ethereum, Web3, solidity, and the Cosmos project.
B-Harvest’s current focus is in the Cosmos and Tendermint based blockchain ecosystem, currently validating the Cosmos, Terra, and Iris networks. In our discussion, it was stated that the team has a concentration on the decentralization of the blockchains they validate. Additionally, the team believes that validators have more responsibility than operating nodes. B-Harvest lives by the motto, “not just operate, but also contribute. (Codebase & Governance). ” The team is proud to have created the first passing proposal on the Cosmos network in addition to the critical bugs found in Tendermint.
B-Harvest currently validates on the Cosmos and Terra network. The team also has a “test” validator setup on the Iris network.
As #14 on the Cosmos network, B-Harvest’s stake holds a 2.27% voting power. 25.64% of the service’s stake comes from a single address, but has a nice distribution of delegators otherwise, with 21 other validators with at least 20k atoms delegated.
As #18 on the Terra network, B-Harvest’s stake holds a 1.18% voting power. However, 99.998% of their stake comes from a single address.
B-Harvest has a different approach to the service’s commission rate. The team sets the rate to the median commission rate of all validators, updated every month.
B-Harvest has been registered with the Cosmos and Terra network since genesis.
For the Cosmos Network, Hubble’s event history shows 3 incidents of the service missing 50 of 1000 precommits in the month of May. Otherwise, the service has maintained a stellar record. B-Harvest’s governance proposal activity echoes the team’s motto, with votes on all major proposals in the network.
For the Terra Network, Hubble’s event history shows no wrongful activity.
B-Harvest participated in Game of Stakes before the launch of the Cosmos mainnet. In the end, B-Harvest was #2 in accumulated stake and apart of the global “Never Jailed Crew”. This game demonstrated the services ability to run a highly available validation service.
B-Harvest’s actively analyzes and tests the Tendermint/Cosmos-SDK codebase. As a result, the team has uncovered 2 critical bugs about “Validator Withdrawal” and “Accum calculation malfunction”. The team was awarded $6,750 by the Tendermint team for both reports. Additionally, the team’s reputation score on Tendermint HackerOne is #2, followed by Certus.One.
The team’s background in Big Data has lead B-Harvest to build a data driven product in addition to their validation service. B-Harvest provides a “Blockchain Data Service” that provides data that common explorers cannot provide. The team continually indexes blockchain data to provide varying analysis and statistical information for customers and delegators.
B-Harvest expects their delegators to understand the risks of delegation. There is currently no SLA in place with validators.
- Failover (16/30)
- Private Peering (10/10)
- Agreements with other Validators (10/10)
- Sentry Scaling (10/10)
- Backup Strategy
B-Harvest makes use of an active/passive architecture for their validators, located in a datacenter. The failover procedure is kicked in manually to eliminate the risk of accidental double signing.
B-Harvest makes use of the standard sentry architecture. The sentry nodes are distributed amongst 4 different cloud providers for redundancy. There are direct connections established between the datacenter and cloud providers in Korea. The company is apart of an alliance among korean validators in addition to allies in Europe andNorth America, private peering some nodes with these validator.
The team also takes daily snapshots of the latest state, making it easy to create new sentry nodes. It was stated in our discussion that a new node is usually up an running within ~10 minutes.
As apart of B-Harvest’s data driven approach, the team is also building a tool to reconstruct historical data from the raw Cosmos DB data. Part of how this works is by storing the “deltas” of data in addition to the latest state. This allows the reconstruction of each blocks state data without needing to store it all. This project is not yet open sourced. The team is hoping to receive some funding from the ICF (InterChain Foundation) to open source and maintain the project.
Monitoring Tools (50 /100)
- Network Level (5/10)
- Hardware Level (5/10)
- Paging (5/10)
Single Point of Failure (50/100)
- Multi-Cloud (10/10)
- Multi-Region (0/10)
Key Management (50/100)
- HSM Selections (10/10)
- Smart Key Management (0/10)
Validator Access (0/100)
- Physical/Remote (0/10) – Not disclosed
B-Harvest uses the ELK (Elastic search, Logstash, and Kibana) stack to monitor the health of their nodes. Their monitoring tools are linked with PagerDuty to notify the team in case of any problems. Although not imminent, the team is also working on constructing a framework using Kubernetes to make management more structural across all the networks the service validate.
Single Points of Failure
B-Harvest’s use of a single colocation to host the service’s keys and validator nodes presents a single point of failure in the case of a catastrophic failure. The rest of the system architecture is resilient to single failures.
B-Harvest makes use of the YubiHSM in their colocation to sign blocks. Per the modules design, hardware based double sign protection is not possible. There is also no software based KMS solution in front of the HSM either. The team stated that while the Tendermint KMS is an attractive open source solution, more tests and due diligence needs to be done before it’s use. However, it was also stated that the team believes a KMS solution will be in place within 1-2 months of this writeup.
B-Harvest’s offices and team members homes are located within 15 minutes of their colocation center. Addressing any problems or triggering manual failover can be done by the team very quickly. Security Practices around remote access to these servers were not disclosed in this interview.
B-Harvest believes that their data driven approach is a unique differentiator among other validators. At the time of writing, the team stated that is not currently possible to quickly search your own transactional data without having the complete blockchain state history. B-Harvest does have a paid plan for their data service, and aims to pursue future iterations in an open-source and free fashion.
The team is also selective on the networks they validate on. Per their motto, the team keeps a balance between prioritizing validator security/operations and contributions. A depth-first approach is taken by B-Harvest instead of breadth-first.