Friday, November 14, 2008

Cool Technology of the Week

In my quest for creating a storage utility service for thousands of users, I've explored SAN, NAS, Content Addressable Storage, compression/de-duplication, and file virtualization.

One area that has been a struggle is providing a cloud of mid-tier storage at price that the community is willing to pay.

This week, EMC introduced Atmos, a policy-based information management solution for building cloud storage infrastructures. Atmos is a multi-petabyte information management solution designed to optimize the delivery of unstructured information (files rather than databases) across large-scale, global storage environments. Cloud Optimized Storage (COS), EMC's name for this new class of storage, is massively scalable and initially designed for Web 2.0, telecommunications, and media companies with significant storage demands such as YouTube, Picassa, Flickr, and Akamai.

EMC calculates that for every $1.00 spent on storage, $6.00 is spent on backup and recovery over the life of the information. The Atmos approach is to use large arrays of inexpensive drives without RAID - data is simply replicated within the Atmos cabinet or in a different cabinet across the coutry. Eliminating RAID enhances performance and reduces complexity. EMC also realized that archiving/restoring a petabyte of data is extremely challenging, but replication works well as a backup strategy.

EMC Atmos is made up of management software that is packaged with a low-cost, high-density storage system that offers ease of implementation and service in an efficient footprint (comprised of X86 servers and high-capacity SATA drives).

Per EMC, the features of Atmos include:

* Massive scalability: EMC Atmos scales effortlessly to accommodate the ever-expanding need for storage.
* Policy-based information management: EMC Atmos improves operational efficiency by automatically distributing information based on business policy. The user-defined policies dictate how, when, and where the information resides.
* Object metadata: EMC Atmos uses metadata to refine the content distribution and retention policy, improve searches, or build custom queries for cloud-based services.
* All-in-one data services: EMC Atmos includes built-in storage and information management features that include replication, versioning, compression, de-duplication, and disk spin-down. These features are native to the platform.
* Choice of access mechanisms: EMC Atmos provides flexible Web service APIs (REST/SOAP) for Internet-based applications or legacy protocols (CIFS/NFS/IFS) for file-based systems. This enables integration with virtually any application.
* Automated system management: EMC Atmos provides auto-configuring, auto-managing, and auto-healing capabilities to reduce administration and downtime.
* Multi-tenancy: EMC Atmos can allow multiple applications to be served from the same infrastructure. Each application is securely partitioned and can never access another application's data. Multi-tenancy is ideal for service providers or large enterprises who wish to provide cloud services for many customers or departments.
* Unified namespace: EMC Atmos uses a single view of the information to provide universal access regardless of location, reducing complexity and improving productivity.

These features seem to be a good fit for the image storage requirements of a large academic medical center and the increasing need for storing genomics data for the research community.

A cloud of inexpensive, massively scalable storage. That's cool!

No comments: