This content has been marked as final. Show 5 replies
Billy Verreynne wrote:There is no Oracle Database for Linux on SPARC, so even if you did get (any) Linux installed on that box, you wouldn't be able to run Oracle DB on it. You're better off going with Solaris, or just mothballing the machine.
Solaris is not an option due to cost implications that management consider unwarranted for using a mothballed server for some basic Oracle RDBMS development
Thanks. (facepalm for the obvious and staring simply and blindly at the kernel issue)
PS. Solaris is not an option as Oracle db has several patch requirements for Solaris, with patches only available when the machine is on a maintenance contract (which is in the same cost ballpark than getting DB Appliance - my original recommendation to the powers-to-be)
Edited by: Billy Verreynne on Jan 9, 2013 9:55 AM
Sorry I'd like to comment on that. Isn't it strange, from a technical standpoint, that all development for server CPU systems based on RISC and IA-64 has been ceased? Oracle and many other vendors stopped supporting the RISC and Titanium CPU already a couple of years ago. Finding any recent software to run on these server systems is a problem. Sparc, Titanium, PowerPC can all go for scrap.
Now with the Intel x86_64 CPU dominating the market to support legacy desktop hardware and to run 32-bit CSIC code efficiently (MS-Windows), the IT industry finally started to outphase 32-bit systems, moving to 64-bit.
As far as I'm aware, the x86_64 technology provides a limited 64-bit CPU mainly to support registers to address larger memory. Most CPU's today are internally RISC based, and have been for decades though. It seems the Intel x86_64 bit CPU killed the prestigious RISC based 64-bit server market in favor to support the end-user desktop market (Intel/Microsoft), to finally run 64-bit systems on x86_64 CPU's?
Also a fan of RISC based processors. (used to be a product manager when I was still a corporate employee and not a contractor - and RISC-based h/w product ranges were part of the product basket I looked after)
So yeah - kind of disappointing to see RISC based CPUs declining ifo Intel. (even prefer AMD instead of Intel and recently bought a couple of 8 core bulldozer systems for the home).
But one of the drawbacks for RISC was the cost of developing smart compilers that gave the RISC design its due. Whereas CISC based compilers were a lot cheaper to develop. To fully optimise CPU pipelining, the complexity moved from CPU design to compiler design. We had some seriously fast benchmarks on our RISC systems back then, in comparison with it running on "faster" Intel CPUs...
But then CPU technology changed a lot the last 2 decades.
Maybe ARM is changing the x86 stranglehold? It is the CPU for smart devices. Microsoft's Surface is using it. AMD is planning to release a new 64bit server CPU range in 2014 based on ARM.
Reading about ARM on http://en.wikipedia.org/wiki/ARM_architecture I find a lot of interesting information, but I don't see much groundbreaking, though I must admit that I'm not a CPU engineer or have lots in-depth knowledge about it.
Am I a RISC fan? Well, I like good hardware. And I like to test hardware to see what performance is possible, storage systems in particular.
Personally I regret the loss of the Alpha chip and the loss of the Apple PowerPC. I had some interesting talks with a couple of HP engineers 10 years ago, complaining about the Alpha chip being superior to the IA-64 technology. I have forgotten the details. What I know is that video compressing tasks on my 8 year old PowerMac quad-core G5 with 4 x 2.5 GHz PPC 604e is noticeably faster than doing the same job on my Mac Pro 2009 with dual quad-core 2.3 GHz Intel Nehalem with all the latest and greatest.
Regarding RISC based and CISC based compilers. I think the advantage of CISC lies in simplicity for the programmer. I remember Apple developer guidelines from the early 90's to rely on system libraries residing in ROM. It had its advantages in terms of programming ergonomics, but caused more complexity and overhead in the CPU itself and thus making programs less efficient. I think the same applies to CISC. RISC is less efficient in terms of coding, but more CPU efficient, though shifting more work to the compiler. CISC versus RISC compiling was a valid argument in the past, but does it still apply?
Somewhere I read about the huge progress of the x86_64 architecture, which claims to outperform any existing RISC architecture, but I have not been able to experience that in practice. My experience with replacing old legacy RISC server systems is that it requires 4 times the hardware. But of course it is impossible to seriously compare old with new or different systems.
Looking forward by looking back, the best way to predict the future is to invent it. I wonder if some of the past developments are really coincidence. How does my future look and why I still have plans? ;-)
Edited by: Dude on Jan 9, 2013 6:50 AM