For the original implementation of memcached, please refer the the github link below: https://github.com/memcached/memcached

Hillclimbing implementation is derived from Cllifhanger's github:
https://github.com/assafein/memcached_hillclimbing

• libevent, http://www.monkey.org/~provos/libevent/ (libevent-dev)
• librt, in order to allow the use of POSIX shared memory
• automake, autotools-dev, and autoconf

Implemented a new mechanism that allocates from a shared pool of memory reserved only for memcached instances. This allows for a global view of memcached information and allows for future optimizations and more efficient slab allocation algorithms. Currently, if instances are run using -G (greedy) they will request to take memory as much as they can, ignoring the memory limit set for them when starting instances.

• ./autogen.sh
• ./configure (Edit the LIBS in generated makefile to include -lrt and add -w to CFLAGS)
• make
• Compile init_share and stop_share (i.e. gcc -o init_share init_share.c -lrt -pthread).
• Run ./init_share XXXX (MB) with the total amount of shared memory you need (i.e. ./init_share 10000).
• Run memcached instances (i.e. ./memcached -p 11212 -t 4 -m 4096 -n 550 -G).
  Option G indicates the greedy approach of memory allocation, if not specified, instances will get memory equal to the limit set with -m.
• Once finished run ./stop_share to unlink the shared memory segment.

For benchmarking, the data caching package provided in Cloudsuite can be used. You have to have docker installed on your system.

• Pull the client image from the docker hub (i.e. docker pull aamh/data-caching:client)

docker run -it --name dc-client --network="host" aamh/data-caching:client bash
cd usr/src/memcached/Cloudsuite-Client/memcached_client/
vim docker_servers.txt 

• Edit this file to look like this:
  127.0.0.1, 11212, 1
  127.0.0.1, 11213, 2
  ....
  

  This file should contain one line for each memcached instances you want to test.
  First column is the IP address (127.0.0.1 is the localhost in the client container).
  Second column specifies the port the server is listening to.
  Third column specifies the reverse ratio of number of request sent to the instance (1 means full speed, 2 means half,..).

• To scale the dataset and warm up the servers:

./loader -a ../twitter_dataset/twitter_dataset_unscaled -o ../twitter_dataset/twitter_dataset_30x -s docker_servers.txt -w 4 -S 30 -D 6000 -j -T 1 -r 400000

• If dataset is already scaled, warm up the servers:

./loader -a ../twitter_dataset/twitter_dataset_30x -s docker_servers.txt -w 4 -S 1 -D 6000 -j -T 1 -r 400000

• Run the benchmark:

./loader -a ../twitter_dataset/twitter_dataset_30x -s docker_servers.txt -g 0.8 -T 1 -c 200 -w 8 -r 400000

For the full description of arguments and a more comprehensive guide, please refer to the Cloudsuite documentation.

See https://github.com/memcached/memcached/wiki/DevelopmentRepos