At its Global Summit Red Hat Brian Stevens announced Storage 2.0, using the technology acquired with Gluster in 2011. Its approach to storage is important, given its reliance on Open Source and its potential for influencing the direction storage hypervising will take. We know you’ll be interested in its first steps into this complex market.
No need for directory nodes with Gluster’s elastic hosting algorithm
For Red Hat Gluster provides RHEL-based nodes for scale out storage and designed to handle the masses of unstructured data entering organisations and often described as ‘Big Data‘. The data is held in RHEL XFS file systems without the need for users to know which nodes it runs on. The most important feature of Gluster is its ‘elastic hashing algorithm’, which gives instant data access without using a directory (or meta) node. It provides a global name-space across a set of systems.
Brian talked about the challenge of ‘hot data’, especially where something written once is accessed multiple times. The advantage of the Gluster approach is in the ability to add nodes easily – valuable data can be replicated to extra nodes when needed. You can also add more nodes for aggregated access.
Harvesting intelligence from data via Hadoop
Brian claims that Apache’s Hadoop is synonymous with Big Data. Its MapReduce algorithm is used by Public Cloud providers such as Facebook, Rackspace, eBay, Spotify, Yahoo! and others to run queries across (often thousands) of distributed nodes – Twitter’s handling of 12TB of data per day underlines the scale of the challenge. Hadoop queries can be run across a number of Gluster nodes, avoiding the need to build a separate cluster for storage. By integrating Hadoop and Gluster, users will aso be able to extract real time knowledge from active data. RHEV already offers live migration of applications and databases to a new node and will include the ability to do the same for storage later: all of which provides significant help in load-balancing across a wide infrastructure.
The opportunities for Red Hat and Gluster
Red Hat currently chooses not to use the term ‘storage hypervising’ when discussing Gluster, perhaps because in its certification of x86 servers from leading vendors it has not yet adopted a heterogeneous attachment of arrays: however its approach is highly relevant.
Its initial focus is on large Public Cloud suppliers. They illustrate how important data analytics and the use of Hadoop are for running a tiny number of applications at massive scale and the advantages that Gluster brings. However, given their technical expertise and use of Open Source projects and components, they may prove difficult for Red Hat to sell to.
It may also be some time for Red Hat’s approach becomes relevant to humbler organisations running server and client virtualisation on top of traditional IT and looking to data analytics to the mix – not least because the features of RHEV currently lag behind those of vCenter and System Center. Nevertheless this is an impressive development and it will be interesting to see how its Gluster’s hypervising credentials develop.
Image Credit: Lizzie Hingley