Sun Open Sources "Honeycomb" Disk Array Software

Published: February 28, 2008

by Timothy Prickett Morgan

The product took two years to move from a project inside the labs to become the StorageTek 5800 disk array, but only a few months after delivering the petabyte-scale disk array that was code-named "Honeycomb" and that is a key part of the Constellation InfiniBand clustered system, Sun Microsystems has decided to open source the software that comprises the guts of the Honeycomb system.

The purpose of the Honeycomb disk array is to store fixed data--pieces of information, like digital photos, videos, and other kinds of archives--in an efficient and cheap manner. Because the data is fixed, there are some software tricks you can do to boost the performance of the arrays, since they only write once but read the data many times, unlike normal disk arrays, which have to mess around with constantly changing data sets. This may not sound like a big deal, but a study performed by computer scientist Hal Varian of the University of California at Berkeley shows that about 80 percent of data created today will not be modified, and that the pile of this unmodified data will grow by 90 percent between now and 2010, compared to 60 percent growth for mutable data. (Blame YouTube and other kinds of eye and ear candy out there on the Internet.) The Honeycomb system implements what Sun calls an object-oriented file system (meaning it thinks in terms of the binary objects it stores, not blocks and sectors below a file system) that can scale to 100 million objects comprising petabytes of disk capacity in aggregate.

With such a capability, you might think Sun would try to keep its hands on it. But selling an actual disk array that is different from other kinds of arrays is a lot harder than giving away the software for free and letting nerds play with it and then waiting for nerds to recommend a commercial-grade version of the product at some point down the road when a company opts to put it into production. (That, in a nutshell, is Sun's entire marketing plan for all of its software, including Solaris, Java, its application servers and development tools, and anything else that you would presume would be locked away in a safe somewhere in Mountain View.)

Anyway, if you want to load the Honeycomb file system onto an X86 or X64 server and see what it can do, you can download the binaries for the Honeycomb code here from the OpenSolaris site. The software, which was created under a BSD license (a lot less restrictive than the GNU GPL), is technically known as StorageTek 5800 Open Edition, and the reason why Sun probably didn't just call it Honeycomb Open Edition is that a certain cereal manufacturer's lawyers would probably not take it too kindly. Sun has also donated the code to the Java.net community in the hopes that Java nerds will work on some Java APIs in the software (which also has C APIs). The company also submitted the Honeycomb code to the Storage Networking Industry Association (SNIA), which is taking it under consideration as product meeting the organizations Extensible Access Method (XAM) standard. XAM seeks to create standards that allow interoperability between different servers, operating systems, arrays, and applications.

RELATED STORIES

Sun Puts Some Numbers on Its Constellation System

Sun Preps "Honeycomb" Storage Server

                     Post this story to del.icio.us
               Post this story to Digg
    Post this story to Slashdot