% fortune -ae paul murphy


Oracle, long a holdout on multi-core pricing, just announced an interesting compromise. What they're going to do, at least for both of the customers likely to buy at list price this year, is assume that the additional cores will weigh in at 75% of the first one. In other words, that performance drops by 25% for the second and subsequent cores. Fractions are, of course, rounded up.

This gives an odd, and probably unintended, pricing break to Power5 and UtraSPARC IV users because that 25% performance drop-off doesn't apply until you get bus or memory saturation.

On the other hand, it doesn't meet Sun's objective in pushing Oracle to change because licensing costs continue to stand in the way of sales for hardware based on their throughput computing initiative.

Consider, for example, the choice between a Niagara server with a single eight way CPU and a machine with two dual core AMD chips. Under existing Oracle licensing the Niagara user could be asked to pay $320,000, the AMD user only $180,000. With the new pricing the amounts drop marginally (to $280,000 and $140,000 respectively) but the ratio doesn't change -and neither would the insanity of buying a Niagara for an Oracle host when you can get an AMD.

In reality, of course, Oracle generally puts getting the sale ahead of hardware chauvenism - so you're likely to pay the same for either box, but that's not guaranteed.

IBM doesn't have as much of a problem on this as Sun does. Their grid-on-a-chip, aka the Cell processor, probably qualifies as one logical processor -particularly for something like an Oracle database or business application mainly dependent on the 3.X Ghz G5 (yes, the same one IBM couldn't make work for Apple) master and not the SPUs.

What Sun has to do, of course, is bring competitive presure to bear on Oracle either, or both, by visibily supporting efforts to develop Oracle compatible open source products like EnterpriseDB or buying and open sourcing a fast, cheap, transactions database to compete directly with Oracle for application developer loyalty.

On the other hand, I used to make a fair amount of easy money by pretending to know how to tune Oracle on Unix and it occurs to me that there may be an exciting new business opportunity here.

(In reality, I never did understand Oracle tuning, but just undoing the other guy's optimizations, re-creating the database on raw disk, shuffling log and sort files to separate devices, and minimizing every other memory user (like OS buffers on HP-UX) reliabily produced a performance miracle every time.)

The new opportunity would be to create a true web consultancy - a license management server. We'd collect information on licensing - what people paid for, what rights, expiries, and obligations they have, and where these are used. At one level of service we'd bring industry wide information in on the client's side of the table in licensing negotiations with Oracle or anyone else and on the second level we could track a client's licensing obligations for them - even better, we could collect lump sum licensing payments from clients and parcel those out on a monthly or annual basis to the software supplier.


% whois licemanager.com
No match for "LICEMANAGER.COM".

Hey, it's a natural!

Paul Murphy wrote and published The Unix Guide to Defenestration. Murphy is a 25-year veteran of the I.T. consulting industry, specializing in Unix and Unix-related management issues.