Arcus is a memcached-based cache cloud developed by NAVER Corp. Memcached has been heavily modified to support functional and performance requirements of NAVER services. Arcus is normally positioned in front of a back-end storage, such as DB, where it stores/retrieves hot-spot data of service applications using fast main memory. Therefore, Arcus provides very fast data access for service applications and reduces the load on back-end storage. In addition, Arcus can serve as a fast shared storage. Remote service processes usually store complex query results, computation-intensive results, and intermediate results of web processing in Arcus cache cloud so any process can access them.
Arcus provides two new significant features over the existing memcached.
First, Arcus supports collection types. A single key can have a collection of data as its value. Arcus currently supports three collection data structures.
- List - a doubly-linked list structure.
- Set - an unordered set of unique data.
- Map - an unordered set of <field, value>.
- B+tree - a b+tree structure, conceptually similar to sorted map.
Second, Arcus can operate as a cache cloud using Zookeeper. When a cache node fails, it is quickly detected and is excluded from the cache cloud. The remaining cache nodes keep on serving using the new cloud configuration. Likewise, a cache node may join the cloud on the fly. Keys are automatically re-distributed to maintain load balancing.
arcus-memcached builds successfully on 64-bit Linux (CentOS, Ubuntu) environments. Other Linux distributions have not been tested.
Dependencies
Install tools for building by using package manager in linux system.
(CentOS) sudo yum install gcc make which libtool
(Ubuntu) sudo apt-get install build-essential libtool
Also, arcus-memcached has the following library dependencies.
- libevent - An event notification library
- arcus-zookeeper - Zookeeper c library with Arcus modification
To install them easily, run the command below from the git-cloned source code.
$ ./deps/install.sh [<arcus_install_path>]
If the arcus_install_path was not given, dependencies will be installed in the system path(/usr/local).
Therefore, giving the arcus_install_path preferably is recommended.
Compile
To build arcus-memcached from the git-cloned source code,
$ ./config/autorun.sh
$ ./configure [--prefix=<arcus_install_path>] [--with-libevent=<libevent_install_path>]
$ make
$ make install
The use of ZooKeeper based clustering is optional.
To enable it, use --enable-zk-integration along with --with-zookeeper when running configure.
Note that ZooKeeper dynamic reconfig was included in it.
So, you must use the ZooKeeper library 3.5.8 or higher version with Arcus modifications.
If you want to turn on the ZooKeeper dynamic reconfig, add --with-zk-reconfig to the above configure options.
To test arcus-memcached, you can execute make test. If any problem exists in compilation, please refer to compilation FAQ.
arcus-memcached has a pluggable engine structure. Only the default engine provides all of the above Arcus functionalities.
To start Arcus cache server with the default engine, run the following command line with non-root user.
$ <arcus_install_path>/bin/memcached -m 4000 -p 11211 -E <arcus_install_path>/lib/default_engine.so
Important start options are here.
- To enable Zookeeper-based clustering, use
-zto specify the Zookeeper ensemble ip:port list. - The scrub command is provided as an ASCII command extension.
To use the command, use
-Xto specify the location of ascii_scrub.so library.
To see details on arcus-memcached start options, run memcached with -h option like below.
$ <arcus_install_path>/bin/memcached -h
To stop the running Arcus cache server, use kill command like below.
$ kill -INT <pid>
Please refer to Arcus cache server ascii protocol for details on Arcus ASCII commands.
You can use telnet interface to test simply Arcus ASCII commands on Arcus cache server. To know the usage of telnet interface, refer to Arcus telnet interface guide.
If you find a bug, please report it via the GitHub issues page.
https://github.com/naver/arcus-memcached/issues
In addition to those who had contributed to the original memcached, the following people at NAVER have contributed to arcus-memcached.
JunHyun Park [email protected]; [email protected]
HyongYoub Kim [email protected]
YeaSol Kim (ngleader) [email protected]; [email protected]
HoonMin Kim (harebox) [email protected]; [email protected]
SeongHwan Jeong (scryner) [email protected]
Chang Song [email protected]
Licensed under the Apache License, Version 2.0: http://www.apache.org/licenses/LICENSE-2.0
Arcus has patents on b+tree smget operation. Refer to PATENTS file in this directory to get the patent information.
Under the Apache License 2.0, a perpetual, worldwide, non-exclusive, no-charge, royalty-free, irrevocable patent license is granted to any user for any usage. You can see the specifics on the grant of patent license in LICENSE file in this directory.