BESPOKV - README

Dependencies

• gcc 4.8 (required by folly)
• autoconf-2.69 (requires v >= 2.69)
• google-glog
• protobuf
• libopenssl
• gflags-master
• boost_1_55_0: ./b2 install
• folly (requires double_conversion)
• libuuid
• libevent
• lz4 1.7.1 (redlock requires v >= 1.7.1)
• zookeeper (server + C binding lib)

How to compile?

Compile BESPOKV

Go to the src directory, for debugging mode, run 'make'

To enable compiler level optimization, run 'make opti'

Compile datalet application

Go to the apps directory, for debugging mode, run 'make'

To enable compiler level optimization, run 'make opti'

Compile client lib

To compile client lib, go to the libckv dir and compile proto file:

cd apps/clibs/libckv
protoc --cpp_out=. ckv_proto.proto

Move the header file generated to the include folder:

cp *.h include/

Next, create a directory to compile the lib:

mkdir build
cd build/
cmake ..
make

libckv.a will be available in build directory

Compile benchmark

First, compile the client lib as shown in previous step.

Then go to the bench directory and run 'make'

How to run?

Datalet backend

First, you should have a backend datalet running, e.g., a Redis node:
Go to the Redis dir:

$ ./redis-server --port 12346

Under apps/ , we implemented two applications for bespoKV. If you want the datalet backend to be a key-value store, type:

$ cd apps/ckv
$ ./conkv -l 192.168.0.170 -p 11111 -t 1 

BESPOKV

To run the bespoKV executable, go to the src dir:

$ ./conproxy --config /root/conrun/conf/c1.json --datalets /root/conrun/conf/d1.cfg --shard shard1 --proxyAddr 192.168.0.170 --proxyClientPort 12345 

Configuration files include:
A JSON formatted file specifying all options (num_replicas option might be a bit confusing and here it indicates how many replicas excluding the master replica), as below:

{
	"zk_host": "192.168.0.173:2181",
	"kafka_broker": "192.168.0.173:9092",
	"consistency_model": "strong",
	"consistency_tech": "cr",
	"topology": "ms",
	"num_replicas": "2",
}

And a datalet list file specifying all datalets, as below:

# 0: master; 1: slave
192.168.0.171:11111:0
192.168.0.171:11112:1
192.168.0.171:11113:1

Benchmarking

To run the YCSB trace bench:

$ cd bench
$ ./bench_client -d 40 -l trace_dir/kv1M_op1M_uniform_text.run -t 32 -m 6 -r 2 -f hosts.cfg -R 0 -W 0

Where hosts.cfg is an example host file including all hosts, -m indicates how many hosts, -r indicates how many replicas, -R specifies which replica to serve READ (-1 means any replica), and -W indicates which replica to serve WRITE (again, -1 means any replica for active/active topology). Example host file is shown as below:

192.168.0.171:12345 192.168.0.171:12348
192.168.0.171:12346 192.168.0.171:12349
192.168.0.171:12347 192.168.0.171:12350
192.168.0.172:12345 192.168.0.172:12348
192.168.0.172:12346 192.168.0.172:12349
192.168.0.172:12347 192.168.0.172:12350

To run the Redis benchmark:

$ ./redis-benchmark -h hulk0 -p 12345 -c 50 -n 100000 -t set,get -P 32 -r 1000000  

This will send requests to conproxy, which will serve as a proxy forwarding requests between benchmark clients and Redis backend datalets.

Deployment

To launch zk and MQ on cloud, run:

bin/zookeeper-server-start.sh -daemon config/zookeeper.properties
bin/MQ-server-start.sh -daemon config/server.properties

To launch a cluster of bespoKV + conkv nodes, first add the data node info in slap.sh, then run:

$ cd scripts  
$ ./slap.sh runckv  
$ ./slap.sh runcon

Docker based bespoKV is ==PARTIALLY== supported. To run containerized deployment:

$ ./slap.sh docker_runckv
$ ./slap.sh docker_runcon

Configuration file setting for different scenarios

-----Master-Slave topology with strong consistency-----

Master:
{
	"zk_host": "192.168.0.173:2181",
	"kafka_broker": "192.168.0.173:9092",
	"consistency_model": "strong",
	"consistency_tech": "cr",
	"topology": "ms",
	"num_replicas": "2",
}

Slave:
{
	"zk_host": "192.168.0.173:2181",
	"kafka_broker": "192.168.0.173:9092",
	"consistency_model": "strong",
	"consistency_tech": "cr",
	"topology": "slave",
	"num_replicas": "0",
}


-----Master-Slave topology with eventual consistency-----

Master:
{
	"zk_host": "192.168.0.173:2181",
	"kafka_broker": "192.168.0.173:9092",
	"consistency_model": "eventual",
	"consistency_tech": "cr",
	"topology": "ms",
	"num_replicas": "0",
}

Slave:
{
	"zk_host": "192.168.0.173:2181",
	"kafka_broker": "192.168.0.173:9092",
	"consistency_model": "evenual",
	"consistency_tech": "cr",
	"topology": "slave",
	"num_replicas": "0",
}


-----Master-Slave topology without consistency-----

Master:
{
	"zk_host": "192.168.0.173:2181",
	"kafka_broker": "192.168.0.173:9092",
	"consistency_model": "without",
	"consistency_tech": "none",
	"topology": "ms",
	"num_replicas": "2",
}

Slave:
{
	"zk_host": "192.168.0.173:2181",
	"kafka_broker": "192.168.0.173:9092",
	"consistency_model": "without",
	"consistency_tech": "none",
	"topology": "slave",
	"num_replicas": "0",
}

-----No topology and without consistency-----
{
	"zk_host": "172.17.0.2:2181",
	"kafka_broker": "172.17.0.3:9092",
	"consistency_model": "without",
	"consistency_tech": "none",
	"topology": "no",
	"num_replicas": "0",
}


-----Active-Active toplogy with strong consistency-----

{
	"zk_host": "172.17.0.2:2181",
	"kafka_broker": "172.17.0.3:9092",
	"consistency_model": "strong",
	"consistency_tech": "rl",
	"topology": "aa",
	"num_replicas": "0",
	"rl_host": "127.0.0.1:12121"
}

-----Active-Active toplogy with eventual consistency-----

{
	"zk_host": "192.168.0.173:2181",
	"kafka_broker": "192.168.0.173:9092",
	"consistency_model": "eventual",
	"consistency_tech": "cr",
	"topology": "aa",
	"num_replicas": "0",
} 

Support Channel

Ali Anwar: [email protected]

Yue Cheng: [email protected]

Support is also available through the github repository