Opinion: Vendors should follow the chip design industry's lead and create new licensing schemes to deal with the variable usage patterns of grid computing.
Software licensing is a mess.
As systems grow more open to linking with other grids and more apt to be used by users across the globe, resources are being virtualized like mad.
Web services use is growing, grid computing is leaping beyond academic and scientific usage to emerge in the enterprise, and multicore processors will muddy the definition of just what youre paying for—chip? Core? Machine?
It used to be that you simply bought software and licensed it, paying up front for the maximum number of licenses you thought youd need. That concept is already getting confusing, with the concept of multiple cores on one chip putting the definition of just what is a licensed resource up for grabs.
Grid just adds to the confusion. As Nick Gall of Meta Group put it to me recently, the concept of grid, with its introduction of dynamic spin-up and spin-down of resources, skewers the traditional notion of paying for something in full, up front, and then running it 24x7 in order to get your moneys worth.
Click here to read more about grid computing.
At this point, if you want a 100-node grid and want the capability to run your software on 100 nodes if things really scale up, you have to buy 100 copies up front, at full price, to cover it, although in a normal day-to-day situation, for maybe 300 days out of the year, you only use 10 copies.
Thats great—for vendors. But with grids, it just seems wrong. Users want dynamic pricing to reflect their actual usage, which in most cases will be a fraction of the maximum computing cycles they consume at peak times.
This isnt a problem that arose with Oracles 10g grid technology or the prospect of AMDs multicore processors. Grids have been used, for example, in computer-assisted circuit design for over 10 years. In that sector, highly specialized, compute-intensive software aids in chip layout and wiring algorithms for major semiconductor and chip design.
Its the best-kept secret out there: the chip design industry has figured out the grid licensing dilemma, as have telcos, automotive or aerospace companies—pretty much any company thats involved in compute-intensive design.
for one, works with software companies in this space to help them grid-enable software and, in so doing, works through their license structures so that licenses are paid for on a per-job or "token" model, governed by a license management server, which in turn is powered by license management software such as that from Macrovision.
Platform has worked with design software companies including Cadence Design Systems, Synopsys
and Mentor Graphics
in this way.
Vendors drag their heels on alternate licensing schemes.