Oct 26, 2010 (11:10 AM EDT)
Caringo Object Storage Software Puts SMBs On Cloud
Read the Original Article at InformationWeek
Slideshow: Amazon's Case For Enterprise Cloud Computing(click image for larger view and for full slideshow)
Version 5.0 of CAStor adds features that make it easy for SMBs and hosted services firms, as well as enterprises, to create and provision highly reliable and available scalable storage from commodity x86 servers, as opposed to RAID or traditional NAS/SAN solutions.
Objected-based storage means "no file system," said Mark Goros, chief executive officer. "A file system has limits on the size and number of files, and other concerns. CAStor manages the files and their metadata, which includes system metadata that users don't touch, and user metadata such as tags to describe the data. To find an object, you give the UID for a file. The metadata can be used to 'drive' the data, like create rules based on metadata." (Storage “objects” are unstructured digital data, that is documents created by user applications such as Microsoft Office, not databases like email, or for transactions.)
A CAStor cloud storage environment can be used as a private cloud, or as a service, said Goros. "The one or more x86 servers running CAStor create a self-managing, self-healing, self-balancing storage infrastructure," said Goros. "To add more space, you just add another x86 box to the sub-net. Adding more disks or bigger disks means more space. Adding another node -- another machine -- also adds more access paths, which provides more performance. And CAStor is symmetric software -- all machines run the same code, which reduces the administration."
"A CAStor-based cloud is perfect for SMBs, because you can start with as little as one terabyte of storage, and grow seamlessly," said Goros. "We mostly focus on private and enterprise clouds, inside the firewall, used as a storage pool. You can use this to make a public cloud. And many make a hybrid, having a replica in a public cloud, for disaster recovery."
"NAS and SAN aren't very scalable, they don’t store metadata, and they don’t automatically migrate or scale as easily," said Goros. "If you want to store a few TB, these are fine. CAStor is good for storing more data over time. For example, a CAStor installation at John Hopkins Hospital went from 30 terabytes to one petabyte without an IT person to manage it."
A CAStor cloud can be used as primary storage, as secondary/archive storage, or as backup, and can be off-site for business continuity/disaster recovery -- any mix -- according to Goros. "A law firm, for example, can use CAStor to assure that data is available, but protected, can't be changed, and supports multiple locations."
New features in CAStor 5.0 software include named objects, Domains and Buckets, dynamic caching, authorization lists, and multi-tenancy.
Unlike traditional file systems, which "use symbolic names for folders and files in b-tree based directory structures," according to Caringo, Unique IDs (UIDs) assigned to objects can be used without changing, regardless of on what disk or node the object is being stored.
"Named Objects," in turn, are "like user-assigned names versus system names," said Caringo. By using UIDs, a company may have billions of objects in a storage environment, each uniquely named and identified. "Most file systems can't scale to this many entries and slow down as you have more things," said Goros. "Our approach doesn't slow down the cluster as the number of objects grows."
"Named objects are ideal for applications requiring a self-generated or symbolic name to store and retrieve a file or object," according to the company. For example, said Goros, "If you are dealing with a large photo-sharing site that automatically generates filenames, CAStor can take and use these names."
Caringo's new Named Objects semantics "are consistent with the emerging de facto standards of popular public cloud storage services, such as Amazon S3 services, enhancing Web 2.0 usability of a Caringo CAStor cloud environment," according to the company.
For larger SMBs and enterprises, CAStor 5.0's new multi-tenancy feature lets multiple departments or divisions be hosted within one physical private cloud, "with each maintaining total security and authenticated separation without requiring multiple hardware configurations," according to the company.
Multi-tenancy lets enterprise IT give each tenant its own domain for user content, and lets storage service companies provide each client company with its own domain, within a common storage infrastructure.
According to Caringo, unlike Amazon S3, which only allows one domain, CAStor 5.0 lets a company create multiple domains in the same storage infrastructure, and "within a domain, named objects are organized into multiple storage buckets that may be protected with their own security realms and Access Control Lists (ACLs). Users can also provide ACL control at the individual object level. CAStor multi-tenancy makes it a straightforward process for Web 2.0 and individual companies to easily host a CAStor storage cloud."
"Domains let you have one physical storage cloud with multiple virtual clouds in it, like Marketing, Administration, Sales, and Accounting, with each domain having control over its security," said Goros. Users can provide ACL (Access Control List) control down to the individual object level.
CAStor 5.0's new dynamic caching feature can accelerate read access of high-demand objects, somewhat like how a Content Delivery Network (CDN) makes online content more quickly and effectively available to web users. "A company can put clusters in seprate locations, to improve performance for when an object is 'popular' -- being heavily requested -- and to provide disaster recovery among sites."
CAStor software has an embedded Linux kernel, and installs on "bare metal," according to Goros. "You download the software and install it onto x86 servers, and it automatically creates an easy-to-manage, scalable storage cloud. We use replication to other nodes, instead of RAID, and there is always a minimum of two replicas for an object. So recovery is minutes instead of hours; CAStor scales, which RAID doesn't; and you never lose access while rebuilding a node."
CAStor 5.0 is available now, through channel partners, as software and/or hardware/software appliances. Pricing is determined by channel partners. A version that will work with up to 4TB is available free.