The future of HW at Oracle? This is on the Oracle web site. What would you do with IP you have no idea how to build/market/sell to customers?
13 replies (most recent on top)
The SPARC CPU itself has some serious baggage in it, originating from a flawed Berkeley technology from the 1980s: register windows. It was never studied correctly in real multi-processor environments, it hurts performance, and makes it much more difficult to port modern software languages to SPARC for anything that deals with assembly language instructions, like performance or machine generation code coming from X64. Register windows can never be removed from SPARC because that would break application compatibility, thus SPARC is stuck with them. MIPS from Stanford got it right, SPARC from Sun/Berkeley didn't. Rumors had Oracle studying how to put a flat, window-less procedure calling mode into SPARC and preserving compatibility at the same time but there is no SPARC staff left to continue this work, if it ever existed.
F is already changing processor strategy and they've an announced it - see A64FX stuff. Oracle's only interest in SPARC is selling the current generation of systems for as long as they can. Most of the hardware people have already gone. The remaining remnants of Solaris are there to fix bugs, update some open source components and finish some work they were contractually obliged to. There is no roadmap, unless you accept a blank sheet of paper as a roadmap.
Oracle have had so many Solaris people voluntarily leave they've had to un-RIF (cancel the RIF) for some. The org continues to shrink.
"That number will decline as SC's spreadsheet management will ensure proper margins in supporting customer SPARC deployments until 2034."
Given that many will be retiring in the coming years, I doubt layoffs will even be necessary.
"My company is just about to invest heavily in Oracle SPARC gear."
New oracle kindergarten propaganda.
Ask you parents if you can go go outside oracle building and go to customers. Ask them about solaris.
id--t !!!
SPARC is still out there in some of the mature industries: Insurance, Financial, Transportation, etc. Heck, IBM mainframes are still running critical business code and they are a few generations back from a SPARC M7 system. I could see an existing SPARC customer buying incremental SPARC to address capacity needs but "investment" for new applications is over. ORCL is down to a few hundred people for combined SPARC and Solaris development, test, sales, marketing, support, tech writing, etc. That number will decline as SC's spreadsheet management will ensure proper margins in supporting customer SPARC deployments until 2034.
Ive been in alot of data centers, and talked to alot of companies. You never hear about or see sparcs- and its been that way for like 15 years. The guy below, who says he is buying sparcs, probably works on the sparc team, and is either a bs artist or a dinosaur. Very few, if anyone, is making any large investmemt in sparcs. Certianly not enough to be meaningful.
Cray is even having hige trouble, and they are still selling crays.
Nice try though
Who cares about Fujitsu ? May be only a few big old Japanese companies still use it because Fujitsu sales girls are cute. The rest of the world move on to cloud computing.
SPARC on servers is not relevant anymore. Note that I didn't say dead because it still runs in a lot of datacenters. Oracle isn't staffed anymore to even do a maintenance update of the chip anymore, all work done on the next SPARC rev on the roadmap (is it called M8.1?) was done prior to the September 1, 2017 RIF. Older SPARC CPUs like you mentioned were dumped by Solaris 11.4 because their memory management unit (MMU) wasn't compatible with the latest Solaris kernel virtual memory allocator which was redesigned to handle terrabyte RAM sizes. Plus, they no longer have the personnel to test Solaris against older SPARC systems.
Wise choice grasshopper.
My company is just about to invest heavily in Oracle SPARC gear. It's got a great future as the Oracle roadmap shows. Solaris is also being heavily invested in and you only need to see 11.4 being released to know this is true. Oracle wouldn't dump T3 and below systems from running 11.4 if they weren't 100% confident in their roadmap.
From an outsider...
Sorry guys, sparc, as cool as it was, is deader than dead. No amount of life support will save it.
Same for solaris
I know several people in that org and they all plan to retire in that 2-3 year time frame. So, I hope you are right.
Nope. O is no more interested in HW at all, less than ever in reselling someone else's HW with even lower margins.
Original plan was to sell SPARC IP to F and keep on developing Solaris, but it seems it is no more the way the 3 elders want to go.
New plan (still a rumour, but coming fom different trusted sources) is that we will go on until we can squeeze some more money out of the system org, then just close it.
Most probably this means another 2-3 years of surviving.
This is for the sales org, most probably the support org will be here for a little more (from India and Romenia)
F itself is abandoning SPARC. the new S64 cpu due out in 2021 is still unconfirmed (from F internal trusted source) and most of the budget has been redirected to the new ARM A64FX prj, that will see the light first in the new Post-K supercomputer, and then to the new Linux (running a virtualized Soalris) offering that will take the place of the current M12.