Original URL: http://www.theregister.co.uk/2008/04/24/sun_oracle_t2/

Sun chum Oracle pushes database buyers to IBM

T2+ blunder spells Rock disaster in waiting

By Ashlee Vance

Posted in Servers, 24th April 2008 17:02 GMT

Comment A couple of years back, Oracle chief Larry Ellison and then Sun CEO Scott McNealy held an event in Redwood City to renew their vows. Oracle signed on to ship Java for ten more years, and Sun started bundling Oracle's database on its servers at no charge. That last bit was meant to give Sun an edge over hardware rivals, although we can't claim to have heard of it ever making a difference in the market and aren't even sure the deal is still going.

What we do know is that Oracle is working over Sun customers who have adopted the multi-core T2 processor and its recent successor the T2+. In fact, Oracle looks set on exacting some measure of punishment on Sun customers who - dare we say it - want to use the T2 boxes for databases.

Oracle, like IBM, continues to fight the shift toward multi-core chips with all its might, since the beefier silicon threatens to erode per processor licensing revenue. So, the database maker has come up with a system for pricing different processors at different levels based on how many cores they have and how many threads they handle.

In the past, Oracle actually threw Sun a teeny bone, agreeing to price four-, six- and eight-core versions of the T1 processor on a .25 X number of cores basis. So, an eight-core T1 counted as a two processors in the per processor licensing scheme. Oracle, however, kicked Sun's highest-end T1-based server - an eight-core unit running on 1.4GHz chips - up to a factor of .50.

Meanwhile, AMD and Intel chips are meant to be priced by a factor of .50, and all other chips are meant to be priced at a factor of .75. These figures apply to Oracle Enterprise Edition, since Edition One and Standard Edition work on the more sane per-socket basis.

The thing about Oracle's processor pricing chart (PDF) is that it's not terribly good at keeping up with the processor game. For example, IBM's now selling Power6-based systems running at 5.0GHz, and we can only assume they're priced at the same level as the older Power5-based boxes. Pretty soon, HP and Sun will start shipping Unix boxes based on four-core chips, and we suspect Oracle won't get around to updating its list for those suckers, just as it hasn't updated its list for four-core x86 parts.

Sun customers that we're hearing from are particularly aggravated because they thought Oracle and Sun had a good thing going around the UltraSPARC T1 chips. But those joyous days of yore have passed.

Oracle, we hear, is charging a factor of .75 for Sun's T2 and T2+ systems even though they're running at about the same speed as the T1s. The major difference with the new chips is their support for more threads and the fact that the T2+s can go into multi-socket servers making them more useful for, er, databases. And by "more useful" here we mean "useful at all" since no one in their right mind would have thrown the older T1 systems at Oracle.

The .75 T2 factor comes as quite a shock to Sun customers who have upgraded their hardware only to have the Oracle tax man come along and tell them that the solid price/performance they were expecting via the hardware will be eroded via the software.

Gentle IBM? Surely not.

According to our sources, Sun's sales reps have been trying to deal with this situation ever since the T2 boxes came out last year. We understand that Sun is desperately trying to work out a more satisfying arrangement with Oracle.

At this point, just parity with the x86 world would seem nice. Sure, the T2 chips have eight cores and support for 64 threads, but, as you all know, Xeons and Opterons run awfully fast.

Warring over how a T2 should be priced versus a Xeon feels like an inane endeavor at this point. This isn't to say that Sun deserves special treatment. It's to say that all of the processor makers deserve some measure of consistency.

For its part, Sun appears unwilling to gripe in public about Oracle's slight. Both CEO Jonathan Schwartz and server chief John Fowler were silent on the matter when we polled them, and Sun's public relations team has declined to issue a canned statement.

If Oracle and Sun do enjoy a special relationship, we're struggling to find evidence of the love. Sun hopes that the T2+ chips will carry it into the database tier and expand the market for the hardware. Well, that's not going to happen with this kind of pricing structure.

And imagine what will occur next year when - and if - Sun finally rolls out the Rock family of chips. If ever there were a chip meant to run a database, this is it. The 16-core part can handle 32 threads and form servers which share unprecedented amounts of memory.

The funny thing about the Rock design is that you can easily argue it's really a four-core chip, since it divides into four, tidy segments each with additional processing units. Sun, however, wants the glory that comes with delivering a 16-core part. Such braggadocio will do little for Sun's customers multiplying all of those cores by a factor of .75.

As mentioned, IBM has a complex per processor licensing model too, although we must give the vendor credit for taking the whole thing more seriously than Oracle. IBM updates its licensing chart at speed to deal with the latest and greatest chips and seems to have a model that reflects reality better than Larry's shop.

For example, IBM prices its new Power6 chips on a 120 PVU (Processor Value Unit) rate, while older Power5 chips come in at 100 PVUs. So, jumping up to 5.0GHz costs you just a little bit extra. IBM is understandably taking care of its own hardware customers, while placating DB2 and middleware salesmen to a degree. IBM also breaks out the T1 and T2 pricing, so you see T1 systems on 30 PVUs and T2 boxes on 50 PVUs, just like Opterons and Xeons.

If there's a special relationship in play here, it may actually be between Sun and IBM. ®