Skip to content
MICA: A Fast In-memory Key-Value Store (see isca2015 branch for the ISCA2015 version)
C C++ Python Shell Objective-C
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.


A fast in-memory key-value store.

Hardware Requirements

  • Dual CPU system
  • Intel 10 GbE NICs
  • Note: The current codebase has several assumptions on the hardware configuration of the server and clients. It runs ideally on a dual octa-core server with 4 dual-port 10 GbE NICs, and clients with 2 dual-port 10 GbE NICs.

Software Requrements

  • linux x86_64 >= 3.2.0
  • gcc >= 4.6.0
  • Python >= 2.7.0
  • Intel DPDK >= 1.5.0
  • bash >= 4.0.0
  • cmake >= 2.6.0
  • Hugepage (2 GiB) support


  • build/netbench_server: MICA server in cache mode (use with netbench_client)
  • build/netbench_server_store: MICA server in store mode (use with netbench_client)
  • build/netbench_server_latency: MICA server in cache mode modified for end-to-end latency measurement (use with netbench_client_latency)
  • build/netbench_server_soft_fdir: MICA server in cache mode using software-based request direction (use with netbench_client_soft_fdir)
  • build/netbench_client*: MICA clients
  • build/netbench_analysis: workload analyzer (used for generating preset configurations)
  • build/microbench: a local microbenchmark for MICA in cache mode
  • build/microbench_store: a local microbenchmark for MICA in store mode
  • build/test: a simple feature test program
  • build/load: a load factor experiment

Compiling Executables

# unpack DPDK as "DPDK" to the directory containing mica
$ cd mica/build
$ ../scripts/	# this uses sudo
$ ../
$ make

Generating Configuration Files

# conf_* files determine how MICA uses system resources. build/ generates a preset of configuration files for a 16-core server and 12-core clients
# in mica
$ ./	# this uses sudo
$ ./

Running a Server

# in mica/build
$ sudo ./netbench_server conf_machines_DATASET_CMODE_0.5 server 0 0 conf_prepopulation_empty
# DATASET=0,1,2 (used to determine how much memory to allocate); CMODE=EREW,CREW,CRCWS (specifies the data access mode)

Running a Client (e.g., client0)

# in mica/build
$ sudo ./netbench_client conf_machines_DATASET_CMODE_0.5 client0 0 0 conf_workload_DATASET_SKEW_GET_PUT_0.00_1
# DATASET=0,1,2 (specifies the dataset to use); SKEW=uniform,skewed,single (specifies the workload skew); GET/PUT=0.00,0.50,0.95,1.00 (specifies the read/write ratio)

Running a Local Microbenchmark

# in mica/build
$ sudo ./microbench CMODE SKEWNESS 0.5
# CMODE=EREW,CREW,CRCWS (specifies the data acces mode); SKEWNESS=0(uniform),0.99(skewed),99(single) (specifies the workload skew)


Copyright 2014 Carnegie Mellon University

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
See the License for the specific language governing permissions and
limitations under the License.
You can’t perform that action at this time.