boost smr with parallel execution:
- early model: decision to execute the command in parallel is made at the leader
- late model: decision to execute the command in parallel is made at the each follower
- Parallel execution strategy: break the limitation of command sequence execution based on paxos/raft, used hash+queue to construct DAG logically
- raft package used in etcd
- redis
- goredis
- performance tests
- code review
- integrate into Docker/k8s
- 4 HP nodes with four-way E7-4820 CPU, 2.0GHz, 8 cores per channel, a total of 64 threads
only used for validation of algorithm correctness and basic performance testing, not for production environment