DISTRIBUTED Yahoo! Cloud System Benchmark (YCSB)

Requirements

1. ssh client

2. check that the command line utility "expect" is installed on the machines, that will run your cluster

Getting Started

1. the original readme is renamed as

    original-README.md
   

2. Set the USER variable of the global.sh

    you will need to set the USER variable to allow ssh access on to the remote machines
    you will need to set up your public/private key pair so that ssh will not prompt for a password
   

3. Installing DISTRIBUTED YCSB run

    ./install.sh
   
    this will install the YCSB system on all machines required to run the performance tests
    default to localhost
   

4. Run YCSB on DISTRIBUTED machines.

    ./go.sh [outputDirectory]
   
    Running the `go.sh` script with an optional output directory will...
   
        for each type of DB/CLUSTER system listed in SYSTEMS variable of global.sh
            Create a cluster as defined by the variables in the CLUSTER VARIABLES section of global.sh
            Check that the cluster has formed correctly
            Run the original YCSB performance test as defined by the variables in the LOAD PRODUCER VARIABLES section of global.sh
        Produce a directory (default /report) containing the results of the run
   

5. Configure the CLUSTER VARIABLES section of global.sh

    Edit these variables in global.sh to control how the cluster is formed
   
    CLUSTER_MACHINES
        List of IP addresses that will run the cluster nodes
   
    CLUSTER_JVMS_PER_BOX
        each of the machines can run multiple "cluster" JVM's
   
    CLUSTER_NODES_PER_JVM=1
        each of the "cluster" JVM's can run multiple Nodes
   

6. Configure the LOAD PRODUCER VARIABLES section of global.sh

    Edit these variables in global.sh to control how much load will be targeted at the cluster
   
    LOAD_MACHINES
        List of IP addresses that will run the load producing DbClients of YCSB
   
    DB_CLIENTS_PER_BOX
        each machine can run multiple DbClients
   
    NODES_PER_DB_CLIENT
        number of nodes to be started in each DBClient
   
    CLIENT_NODE
        boolean value showing type of nodes to be started in each DBClient, client or cluster member
   
    INSERTS_PER_DB_CLIENT
        How much load each DBClient will produce in the "LOAD" phase of YCSB
   
    OPERATIONS_PER_DB_CLIENT=1000
        How much load each DBClient will produce in the "TRANSACTION" phase of the YCSB system
   
    WORKLOAD
        the name of the YCSB workload file that YCSB will use to control the "TRANSACTION" phase
        see original-README.md for more information about the workload file
   

Adding a new system

1. see original-README.md for how to implement a YCSB DB client interface.

2. DISTRIBUTED YCSB extra requirements

   requires that your target jar will be runnable. and define a static void main which will be passed 2 command line arguments:

        1) the number of cluster nodes to start in this JVM
        2) the target cluster size, the final size of the cluster that will be formed
   

   you are required to check the size of the cluster that forms and on equaling the target size print to std out

        ===>>CLUSTERED<<===
   

   this way DISTRIBUTED YCSB can coordinate the YCSB load and run phases