edit-icon download-icon

System architecture

Last Updated: Jul 16, 2018

Introduction

ApsaraDB for Memcache uses a cluster-based architecture. It is embedded with data partitioning and reading algorithms. The whole process is transparent to the users and saves development and O&M hassles. Each partition node uses a master-slave architecture to guarantee high availability of services.

Architecture

ApsaraDB for Memcache is comprised of three components, namely, the proxy server (service proxy), the partitioning server, and the configuration server.

  • Proxy server

    Single-node. A cluster structure may contain multiple proxies and the system implements load balance and failover for the proxies.

  • Partitioning server

    Each partitioning server is in a dual-copy highly available architecture. The system automatically implements the master-slave switchover in case of a fault in the master node to guarantee high availability of services.

  • Configuration server

    The server is used to store cluster configuration information and partitioning policies. It currently adopts dual-copy architecture to guarantee high availability.

Notes

  • The quantities of proxy servers, sharding servers, and configuration servers depend on the purchased specification. The following table lists the details.
Specification Number of poxy servers Number of sharding servers Memory of a single sharding server
1 GB 1 1 1 GB
2 GB 1 1 2 GB
4 GB 1 1 4 GB
8 GB 1 1 8 GB
16 GB 2 2 8 GB
32 GB 4 4 8 GB
64 GB 8 8 8 GB
128 GB 16 16 8 GB
256 GB 16 16 16 GB
512 GB 32 32 16 GB
  • A Memcache cluster exposes a uniform domain for access. You can visit this domain for normal access to and data operations on Memcache. The proxy server, the partitioning server and the configuration server do not provide domain access and cannot be directly accessed for operations.
Thank you! We've received your feedback.