Gluster

Gluster Inc.
Company type	Private
Industry	Software
Predecessor	Z RESEARCH
Founder	Anand Babu Periasamy
Defunct	2011-10-07
Fate	Acquired by Red Hat
Successor	Red Hat
Headquarters	Sunnyvale, California, United States
Area served	Global
Key people	Anand Babu Periasamy (Founder, CTO), Ben Golub (President and CEO from May 2010)
Products	GlusterFS (open-source network-attached file system)
Owner	Nexus Venture Partners, Index Ventures (investors)
Parent	Red Hat (from 2011)

Gluster Inc. (previously known as Z RESEARCH) was a privately held software firm focused on delivering an open-source platform tailored for scalable public and private cloud storage solutions. Based in Sunnyvale, California, with an engineering hub located in Bangalore, India, Gluster received investment from Nexus Venture Partners and Index Ventures. The company was acquired by Red Hat on October 7, 2011.

History

The term Gluster is derived from a combination of GNU and cluster. Despite the similarity in naming, Gluster has no connection to the Lustre file system and does not incorporate any Lustre code. Its flagship offering, GlusterFS, is an open-source network-attached file system engineered to operate on standard hardware. Anand Babu Periasamy, who founded Gluster and served as CTO, created the original version of GlusterFS.

In May 2010, Ben Golub was appointed president and CEO. Following the acquisition by Red Hat in 2011, Gluster became the main steward of GlusterFS. Initially branded as Red Hat Storage Server, the product was renamed Red Hat Gluster Storage in early 2015 after Red Hat acquired the Ceph file system technology.

Red Hat Gluster Storage is currently being phased out, with support scheduled to end on December 31, 2024.

Architecture

The architecture of GlusterFS consolidates computing power, storage capacity, and I/O resources into a unified global namespace. Each server equipped with commodity storage—configured as direct-attached storage, JBOD, or connected via a storage area network—functions as a node. Scalability in capacity is achieved by adding more nodes or increasing storage within existing nodes, while performance enhancements result from distributing data across additional nodes. High availability is maintained through n-way replication of data among these nodes.

Public cloud deployment

Within public cloud environments, GlusterFS offers an Amazon Web Services (AWS) Amazon Machine Image (AMI) that can be deployed on Elastic Compute Cloud (EC2) instances utilizing Amazon's Elastic Block Storage (EBS). Storage capacity expands by attaching more EBS volumes; performance scales by increasing EC2 instances; and availability is ensured through n-way replication across multiple AWS availability zones.

Private cloud deployment

For private cloud or on-premises implementations, GlusterFS operates as a virtual appliance on commodity servers running hypervisors such as KVM, Xen, or VMware, or it can be installed directly on bare-metal servers.

GlusterFS

GlusterFS is a scale-out network-attached storage file system widely utilized in cloud computing, streaming media, and content delivery networks. Originally developed by Gluster, Inc., it became part of Red Hat, Inc. following the 2011 acquisition.

In June 2012, Red Hat Storage Server was introduced as a commercially supported integration of GlusterFS with Red Hat Enterprise Linux. After Red Hat acquired Inktank Storage—the company behind the Ceph distributed file system—in April 2014, it rebranded its GlusterFS-based product as "Red Hat Gluster Storage".

Design

GlusterFS combines storage servers over networks such as Ethernet or Infiniband using RDMA to form a large-scale parallel network file system. It is distributed under free software licenses including GNU General Public License (GPL) versions 2 and 3, as well as the Lesser General Public License (LGPL) version 3. The design employs a stackable user-space architecture.

The system consists of client and server components. Servers run a daemon that exports local file systems as volumes known as storage bricks. Clients connect using a proprietary protocol over TCP/IP, InfiniBand, or Sockets Direct Protocol, assembling composite virtual volumes from multiple servers via stackable translators. By default, files are stored intact; however, support for file striping across volumes exists. Clients mount these volumes through FUSE using the native GlusterFS protocol, via NFSv3 with an integrated translator, or access them using the client library. Furthermore, clients can re-export mounts through kernel NFSv4, SAMBA, or object storage interfaces like OpenStack Swift by employing the "UFO" translator.

Core features are implemented through translators that provide functionalities such as file mirroring and replication, striping, load balancing, volume failover mechanisms, I/O scheduling, disk caching, enforcing storage quotas, and user-managed volume snapshots starting from version 3.6.

GlusterFS servers export directories without modification; client-side translators organize data presentation. Clients remain stateless and do not communicate directly but require consistent translator configurations across nodes. Rather than relying on centralized or distributed metadata services, the system utilizes an elastic hashing algorithm. Volumes can be dynamically added, removed, or migrated without downtime, allowing scalability to several petabytes on commodity hardware while avoiding bottlenecks common in tightly coupled distributed file systems.

Data durability and availability are achieved through replicated volumes and geo-replication. Replicated volumes maintain multiple copies of each file across bricks to ensure fault tolerance. Geo-replication asynchronously synchronizes volumes across different geographic locations to support disaster recovery strategies.

Red Hat positions GlusterFS for three principal markets: on-premises deployments, public cloud infrastructures, and private cloud environments.

References