Cassandra Client

Installation

Java8

sudo apt-get update
printf "Y" | sudo apt-get install software-properties-common
printf "\n" | sudo add-apt-repository ppa:webupd8team/java
sudo apt-get update
sudo apt-get install oracle-java8-set-default

sudo apt-get update
printf "Y" | sudo apt-get install software-properties-common
sudo apt install oracle-java8-set-default

Cassandra's Python Client

sudo dpkg-reconfigure dash
sudo apt-get install python3-tk
sudo apt-get update
bash
export LC_ALL=C
sudo apt-get install python3-pip
pip3 install cassandra-driver
pip3 install Faker

pip3 install pandas
pip3 install numpy
pip3 install matplotlib
pip3 install datetime

Get Cassandra Database (on client and replicas)

cd /tmp/
git clone https://github.com/hyperpro/CassRepo.git

#client 1
cd /tmp/CassRepo/cassandra-3.9/
cp conf/cassandra1.yaml conf/cassandra.yaml
cd bin
sudo chmod 777 cassandra
cd /tmp/CassRepo/cassandra-3.9/bin
./cassandra -f

#client 2
cd /tmp/CassRepo/cassandra-3.9/
cp conf/cassandra2.yaml conf/cassandra.yaml
cd bin
sudo chmod 777 cassandra
cd /tmp/CassRepo/cassandra-3.9/bin
./cassandra -f

#client 3
cd /tmp/CassRepo/cassandra-3.9/
cp conf/cassandra3.yaml conf/cassandra.yaml
cd bin
sudo chmod 777 cassandra
cd /tmp/CassRepo/cassandra-3.9/bin
./cassandra -f

Cassndra Interface

cd /tmp/CassRepo/cassandra-3.9/bin
chmod 777 cqlsh
./cqlsh $thisIP 

Download tools

sudo apt-get install lrzsz
sz <filename>

Location

The client script can be found in /script/client_sender.py

How to use

Setup Cassandra

To make cassandra read from local only, we have to add read_from_local_only: true to cassandra-3.9/conf/cassandra.yaml

Setup KeySpace

We use same keyspace with YCSB, so we have to run YCSB once before use this client workload generator.

In Cassandra interface, we set

create keyspace ycsb with replication = {'class':'SimpleStrategy', 'replication_factor':3};
use ycsb;
create table usertable (y_id varchar primary key,field0 varchar,field1 varchar,field2 varchar,field3 varchar,field4 varchar,field5 varchar,field6 varchar,field7 varchar,field8 varchar,field9 varchar);

Clear old database

#in cqlsh
use ycsb;
truncate usertable;
drop keyspace ycsb;

#in linux command line
./nodetool -h localhost -p 7199 clearsnapshot

Insert data

cd /tmp/CassRepo/ycsb-0.15.0
./bin/ycsb load cassandra-cql -P workloads/workload_test -p hosts=IP -threads 10

Create Client

ClientSender constructor receive list of hosts. The hosts is IP address of cassandra replicas

from client_sender import ClientSender
from constant import QueryType

hosts = ['127.0.0.1', '127.0.0.2']
sender = ClientSender(hosts)

Send A Request

To send a request we use this method sendOneRequest(host, type, interval, req_id). host is IP address of target cassandra replica that we want to send a request, type is type of request (read, scan, insert, update) we use defined type from QueryType in constants.py, interval is interval between request, and req_id is the request id which is unique.

from client_sender import ClientSender
from constant import QueryType

hosts = ['127.0.0.1', '127.0.0.2']
sender = ClientSender(hosts)
#next line instruction will be sent after 0.01 second, 20-23 are the IDs of requests.
sender.sendOneRequest(hosts[0], QueryType.READ, 0.01, 20) 
sender.sendOneRequest(hosts[1], QueryType.SCAN, 0.01, 21)
sender.sendOneRequest(hosts[0], QueryType.INSERT, 0.01, 22)
sender.sendOneRequest(hosts[1], QueryType.UPDATE, 0.01, 23)

Get Latency

ClientSender class save latency of each requests. ClientSender class will wait until all request is finished. The result is array latency for each request.

from client_sender import ClientSender
from constant import QueryType

hosts = ['127.0.0.1', '127.0.0.2']
sender = ClientSender(hosts)

sender.sendOneRequest(hosts[0], QueryType.READ, 0.01, 20) 
sender.sendOneRequest(hosts[1], QueryType.SCAN, 0.01, 21)
sender.sendOneRequest(hosts[0], QueryType.INSERT, 0.01, 22)
sender.sendOneRequest(hosts[1], QueryType.UPDATE, 0.01, 23)
latency = sender.getLatencies()
print(latency[20], latency[21], latency[22], latency[23])
'''
latency is a dictionary, where key is req_id and value is the latency.
'''

For the code above, we sent 4 requests so we will get 4 latency in an array.

Run experiments

All the programs are inside the scripts folder.

Run a full experiment

python3 ourclient.py THROUGHPUT

The parameter THROUGHPUT is the target throughput which ranges from [90, 150].

Fast run

If you already build the workload-latency profile by wlp.build_profile(sender = sender, hosts=hosts) in scripts/ourclient.py. When running the next experiment just after the previous one, you don't have to re-build the profile. Just comment it is OK.

Use emulator

The above 'fast run' method is still not fast enough, since it has to go through all the traces. Our simulator output the similar (actually is the same) as the real client. Its usage is also similar to run the real experiment.

python3 clientsimulator.py THROUGHPUT