Lucid DB

Implementation of Spanner and Replicated Commit

Config

'lucid.config' is the configuration file which contains the following:

1. IP address of all the shards (with distinct paxos, server and client ports)

2. Number of datacenters

3. Timeouts and Latencies

Each shard is fully replicated. All the replicas of a shard are assumed to be listed together. Order of shards is same for all datacenters. For eg. X1, X2, Y1, Y2 where 1 & 2 are datacenters and X & Y are shards.

Refer to lucid-rc.config and lucid-spanner.config for examples.

The implementation assumes that lucid.config contains valid information at all times.

Running on Eucalyptus:

• 'lucid.config' should have internal IPs of the servers.

• Start servers: all on same instance: java -cp lucid-db.jar com.lucid.ycsb.StartServers 0 -> for SpannerDB [or] java -cp lucid-db.jar com.lucid.ycsb.StartServers 1 -> for RCDB

  Each datacenter on same instance: java -cp lucid-db.jar com.lucid.ycsb.StartDatacenter 0 0 -> for SpannerDB-datacenter 0 java -cp lucid-db.jar com.lucid.ycsb.StartDatacenter 0 1 -> for SpannerDB-datacenter 1 [or] java -cp lucid-db.jar com.lucid.ycsb.StartDatacenter 1 0 -> for RCDB-datacenter 0 java -cp lucid-db.jar com.lucid.ycsb.StartDatacenter 1 1 -> for RCDB-datacenter 1

• Start client as: java -cp lucid-db.jar com.yahoo.ycsb.Client -t -db com.lucid.ycsb.SpannerDB -P workloads/readheavy1k [or] java -cp lucid-db.jar com.yahoo.ycsb.Client -t -db com.lucid.ycsb.RCDB -P workloads/readheavy1k