version-lock

WSC coursework project

Implementation

Hash table variants:

1. vlht.hpp => protected by version lock
2. lfht.hpp => lock-free
3. rwht.hpp => protected by reader-writer lock

Executable

ycsb.cpp => entry point for execution. Some of the important variables include

1. LOAD_SIZE : number of items in hash table before running a workload (set to 64M)

https://github.com/madhavakrishnan/version-lock/blob/main/ycsb.cpp#L160

2. RUN_SIZE : number of operations to run for a workload (set to 64M)

https://github.com/madhavakrishnan/version-lock/blob/main/ycsb.cpp#L161

3. HASH_BUCKETS : number of buckets for each hash table (set to 8M)

https://github.com/madhavakrishnan/version-lock/blob/main/ycsb.cpp#L163

To build the executable, run make and it will generate a binary named ycsb

To run the executable, specify four parameters on the command line as shown below

./ycsb [hash table type] [ycsb workload type] [access pattern] [number of threads]

These parameters are

• hash table type : specify either of vlht, lfht, rwht for hash table based on version lock, lock-free or reader-writer lock
• ycsb workload type : specify either of a, b , c or d
• access pattern : either uniform or zipfian
• number of threads : number of threads to use for a particular evaluation

NOTE:: before running the binary, make sure to generate the YCSB workloads as shown next

Generating YCSB workloads

Download YCSB source code

$ cd ./index-microbench
$ curl -O --location https://github.com/brianfrankcooper/YCSB/releases/download/0.11.0/ycsb-0.11.0.tar.gz
$ tar xfvz ycsb-0.11.0.tar.gz
$ mv ycsb-0.11.0 YCSB

How to configure and generate workloads

Configure the options of each workload (a, b, c, d), change $recordcount and $operationcount ($recordcount should be same as LOAD_SIZE and $operationcount should be same as RUN_SIZE in ycsb.cpp as described above).

$ vi ./index-microbench/workload_spec/<workloada or workloadb or workloadc or workloadd>

In the same file, change the key distribution in $requestdistribution (set to either uniform or zipfian, this parameter is the same as access pattern when running the ycsb binary generated above)

Generate the workloads. This will generate random integer keys with the specified key distribution.

$ cd ./index-microbench/
$ mkdir workloads
$ bash generate_all_workloads.sh

This will produce .dat files for the load and run phase within workloads folder. We will move the generated .dat files into folders based on distribution and size of workload as follows

$ mkdir workloads/uniform 
$ mkdir workloads/zipfian

If we generated workload for 64M operations, create a new folder for the size

$ mkdir workloads/uniform/64M
$ mkdir workloads/zipfian/64M

Finally, move the generated .dat files from the workloads folder into the correct location. As an example, if we generated workload for 64M operations with uniform distribution, move the files by running

$ mv workloads/*.dat workloads/uniform/64M/

Once the workload is moved to the right location, run the ycsb executable as described above.