The “Client-Cluster” Model: An Untapped Benefit of Software Defined Storage

Home   >   Blog

The “Client-Cluster” Model: An Untapped Benefit of Software Defined Storage

Storage analysts forecast the global software defined storage (SDS) market to grow at a CAGR of 31.6% during the period 2016-2020. While customers are drawn to its attractive Total Cost of Ownership (TCO), SDS solutions should also offer the following three benefits:

  1. Commercial off-the-shelf (COTS) hardware
  2. Seamless capacity and performance scaling
  3. Convergence of block, file, and object

AquariTM Storage provides all of these attributes and also features an innovative way for clients and applications to access their data called the client-cluster model.

The Legacy Client-Server Model

Since the early days of storage system architecture, the connections between servers and storage have been point-to-point. As storage systems scaled up by adding ports, “multi-pathing” layers hid the complexity of multiple physical connections so that client servers believed they still only had a single point-to-point connection. Even a large Storage Area Networking (SAN) storage system with 100’s of Fibre Channel ports still presented point-to-point connections to each server. In the world of storage infrastructure, the complexities and pitfalls of dealing with multi-pathing technologies such as MPIO, ALUA, NPIV, and others are well known.

The Modern Client-Cluster Model

With Aquari Storage, data is organized into objects to enable a client-cluster model. Each member of the cluster is aware of the location of objects through a consistent hashing protocol called CRUSH (Controlled Replication Under Scalable Hashing). Each cluster member locates objects within the cluster themselves and object locations are resolved down to single drives using a process known as OSD (Object Storage Daemon).


Client-Cluster Model
Aquari Storage Systems

CRUSH spreads data evenly across the participating disks (OSDs) in the cluster so all cluster resources are fully utilized. It also allows databases and applications to know the exact location of data, including the backup location of objects during failures, without having to check with other nodes.

Aquari Storage Leverages the Client-Cluster Model

Taking full advantage of the Client-Cluster model, Aquari has developed an optimized Reliable Autonomic Distributed Object Store (RADOS) interface, Aquari RADOS (AQR), that intelligently locates and retrieves visual & media streaming assets.

Origin applications that utilize this protocol can automatically and directly access new resources as storage nodes (servers with disks and CPU) are added to the Aquari Storage cluster, with data relocating to utilize the new disks and servers. Similarly, when an additional Origin server is added, it is automatically aware of all content data, the location of that data, and how to access it.

The recent release of Aquari Storage 3.0 adds the new High Availability feature, providing continuous access to files using two Aquari Services Nodes so clients can continue to access data should one services node fail.

Summary

One of the exciting benefits of software defined storage is the way that applications access and consume the storage. Aquari Storage is based on a modern software-defined storage platform that takes full advantage of the Client-Cluster model – providing a superior alternative to the Client-Server model of DAS or SAN.