Sun Microsystems Inc. plans to introduce a fixed-content appliance, code-named Honeycomb, that will compete head-to-head with EMC Corp.'s Centera, in the third quarter of this year.
The object-based storage device will use a proprietary API just like Centera that third-party applications will use to talk to the device. It will be able to store multiple different data types including JPEGs, medical images, e-mail and file data, enabling users to easily locate specific files from a giant data pool.
"The unique part about this product is its distributed search engine design, which allows users to add capacity and linearly add compute power so that as the box fills up, it doesn't slow down," said Chris Woods, chief technology officer of Sun's storage practice.
The final specification for Honeycomb is still being worked out, but the device will be modular with 8 to12
Sun is using a variant of Reed-Solomon encoding to algorithmically generate redundant bits that can reconstruct data when drives die, or data is corrupted and lost. EMC uses the MD5 algorithm for this purpose.
Woods claims Reed-Solomon is superior to the current RAID options available today, as it will be able to recover data following triple or quadruple drive failures. Reed-Solomon encoding was devised in the 1960s and was first used by Voyager to encode digital images sent back from Uranus. Variants of it are more commonly used today in CDs and DVDs to spread data over the media to protect against data loss.
Reed-Solomon is also a way to give each stored object a unique content address, derived from the content itself, to save storage space and ensure the authenticity of the data. No duplicates of the same content are stored. Eventually, users will be able to choose the data protection option that best suits that object.
"This object interface could replace the traditional file interface over time," Woods said. He painted a future where objects would be given an service level agreement number, ranging from, let's say, one to five. One would mean the object was of the highest priority and therefore should be mirrored twice and replicated to a disaster recovery site, while five would be the lowest priority, meaning the data should be sent to cheap storage and deleted within three months. This system would replace the need for multiple different data protection applications like snapshot, backup, replication etc. "You will be able to dial-in the degrees of reliability and availability," Woods said.
The concept of Sun's SAM FS multi-tiered archiving software that provides HSM-like functions will become a part of Honeycomb, and it's QFS file system will eventually be able to hook two Honeycombs together to share a single image. Woods said the Honeycomb software could be ported to other devices: like NAS for general purpose file serving using CIFS and NFS; or it could live in purpose-built hardware or on a large Sun server with generic storage bolted to the back.
Sun will work with third-party software developers to provide the applications for Honeycomb. It is already working with AXS-One on an e-mail archiving package for compliance purposes. This is expected to replace the company's current e-mail archiving software, Infinite Mailbox.
"Too little, too late"
Most analysts think Sun's ingenuity is admirable, but believe Honeycomb will be too late to make a real impact in the market. "It's brilliant they are doing it, but they should have done it three years ago -- it's too little too late," said Todd Holcomb, director of professional services at Evolving Solutions Inc., an IT consulting firm in Hamel, Minn. He also pointed out that it doesn't gel with the company's current directive to slash the prices of its products. "How does a high-end product fit in to this?" he said.
Sun will have a tough job catching up with the rest of the market, too. EMC will have been shipping Centera for several years by the time Honeycomb is ready as will IBM with its DR 550. Hewlett-Packard Co.'s RISS product is relatively new, but starting to gain interest.