Verdi Raft

An implementation of the Raft distributed consensus protocol, verified in Coq using the Verdi framework.

Requirements

Building

First run ./configure in the root directory. This will check for the appropriate version of Coq and ensure all necessary dependencies can be located. By default, it checks for verdi and StructTact in the current parent directory, but this can be overridden by setting the Verdi_PATH and StructTact_PATH environment variables.

Then run make in the root directory. This will compile the Raft implementation and proof interfaces, and check all the proofs.

Files

The raft and raft-proofs subdirectories contain the implementation and verification of Raft. For each proof interface file in raft, there is a corresponding proof file in raft-proofs. The files in the raft subdirectory include:

Raft.v: an implementation of Raft in Verdi
RaftRefinementInterface.v: an application of the ghost-variable transformer to Raft which tracks several ghost variables used in the verification of Raft
CommonTheorems.v: several useful theorems about functions used by the Raft implementation
OneLeaderPerTermInterface: a statement of Raft's Election safety property. See also the corresponding proof file in raft-proofs.
- CandidatesVoteForSelvesInterface.v, VotesCorrectInterface.v, and CroniesCorrectInterface.v: statements of properties used by the proof OneLeaderPerTermProof.v
LogMatchingInterface.v: a statement of Raft's Log Matching property. See also raft-proofs/LogMatchingProof.v
- LeaderSublogInterface.v, SortedInterface.v, and UniqueIndicesInterface.v: statements of properties used by LogMatchingProof.v

The vard Key-Value Store

Requirements:

Coq 8.5
ocaml, ocamlbuild

vard is a simple key-value store implemented using Verdi. vard is specified and verified against Verdi's state-machine semantics in VarD.v. When the Raft transformer is applied, vard can be run as a strongly-consistent, fault-tolerant key-value store along the lines of etcd. The Coq and Ocaml files necessary to run vard on real hardware are in extraction/vard. Running make in that directory will extract the vard and Raft code, link it against the Verdi shim and some vard-specific serialization/debugging code, and compile it all into a vard.native binary. Running make bench-vard will produce some benchmark numbers, which are largely meaningless on localhost (multiple processes writing and fsync-ing to the same disk and communicating over loopback doesn't accurately model real-world use cases). Running make debug will get you a tmux session where you can play around with a vard cluster in debug mode; look in bench/vard.py for a simple python vard client.

As the name suggests, vard is designed to be comparable to the etcd key-value store (although it currently supports many fewer features). To that end, we include a very simple etcd "client" which can be used for benchmarking. Running make bench-etcd will run the vard benchmarks against etcd (although see above for why these results are not particularly meaningful). See below for instructions to run both stores on a cluster in order to get a more useful performance comparison.

Running vard on a cluster

vard doesn't support run-time configuration, so in order to run vard in another configuration (i.e. on multiple hosts) you'll have to edit the extraction/vard/ml/vard.ml file, specifically the 4 lines starting with let nodes = .... For instance, to run it on a cluster with ip addresses 192.168.0.1, 192.168.0.2, 192.168.0.3 you'd edit those lines to read

let nodes = [ (1, ("192.168.0.1", 9001))
            ; (2, ("192.168.0.2", 9001))
            ; (3, ("192.168.0.3", 9001))
            ]

The port is the port used for internal communication; the client port is hard-coded to be internal-port - 1000 (8001 in this example). After recompiling with this config, you could run the benchmarks on a cluster as follows:

# on 192.168.0.1
$ ./vard.native 1

# on 192.168.0.2
$ ./vard.native 2

# on 192.168.0.3
$ ./vard.native 3

# on the client machine
$ python2 bench/setup.py --service vard --keys 50 \
                         --cluster "192.168.0.1:8001,192.168.0.2:8001,192.168.0.3:8001"
$ python2 bench/bench.py --service vard --keys 50 \
                         --cluster "192.168.0.1:8001,192.168.0.2:8001,192.168.0.3:8001" \
                         --threads 8 --requests 100

Running etcd on a cluster

We can compare vard's numbers to etcd running on the same cluster as follows:

# on 192.168.0.1
$ etcd --name=one \
 --listen-client-urls http://192.168.0.1:8001 \
 --advertise-client-urls http://192.168.0.1:8001 \
 --initial-advertise-peer-urls http://192.168.0.1:9001 \
 --listen-peer-urls http://192.168.0.1:9001 \
 --data-dir=/tmp/etcd \
 --initial-cluster "one=http://192.168.0.1:9001,two=http://192.168.0.2:9001,three=http://192.168.0.3:9001"

# on 192.168.0.2
$ etcd --name=two \
 --listen-client-urls http://192.168.0.2:8001 \
 --advertise-client-urls http://192.168.0.2:8001 \
 --initial-advertise-peer-urls http://192.168.0.2:9001 \
 --listen-peer-urls http://192.168.0.2:9001 \
 --data-dir=/tmp/etcd \
 --initial-cluster "one=http://192.168.0.1:9001,two=http://192.168.0.2:9001,three=http://192.168.0.3:9001"

# on 192.168.0.3
$ etcd --name=three \
 --listen-client-urls http://192.168.0.3:8001 \
 --advertise-client-urls http://192.168.0.3:8001 \
 --initial-advertise-peer-urls http://192.168.0.3:9001 \
 --listen-peer-urls http://192.168.0.3:9001 \
 --data-dir=/tmp/etcd \
 --initial-cluster "one=http://192.168.0.1:9001,two=http://192.168.0.2:9001,three=http://192.168.0.3:9001"

# on the client machine
$ python2 bench/setup.py --service etcd --keys 50 \
                         --cluster "192.168.0.1:8001,192.168.0.2:8001,192.168.0.3:8001"
$ python2 bench/bench.py --service etcd --keys 50 \
                         --cluster "192.168.0.1:8001,192.168.0.2:8001,192.168.0.3:8001" \
                         --threads 8 --requests 100