Home Page for the TeradataForum
 

Archives of the TeradataForum

Message Posted: Fri, 01 Jun 2001 @ 21:36:44 GMT


     
  <Prev Next>   <<First <Prev Next> Last>>  


Subj:   Re: Linux and Teradata
 
From:   Charles Farley

--- "Tom Slavinsky" wrote:

  Concerning MP-RAS replacement (a separate response to the ODBC question will be forthcoming), here is the current OS strategy for Teradata :  


  32-bit:  


Continued enhancement (primarily platform enabling stuff) and support for MP-RAS (Unix), as well as Windows 2000, well beyond the availability of 64-bit platforms. MP-RAS continues to provide a robust, highly stable environment for Teradata. There really is no compelling reason to support additional 32-bit Unix flavors. If there is, I'd love to hear it.

No, I don't have any intentions here, I just was made to understand by our NCR people that NCR would be phasing out MP-RAS and would be doing less and less in fixes and development. That's fine with me, but I expect to probably (due to my budget) still be on 32-bit hardware for some time to come. If the phase out means no more etapes or other goodies for the 32 bit version, then I'd really like to see a replacement I can get that for. SGI was rumored (not without merit) that they would be dropping IRIX within 3 years, they already clipped their NT support (that is gone at the end of the year), they see Linux as a great thing, they don't need as many programmers and such because they are no longer the only ones developing the OS, they have a large community to draw on, they would then focus on the parts of Linux that matter to them and release their own dist with those enhancements. I can see the benefits here for NCR, less staff, less infrastructure to support a proprietary OS. I'm not knocking this, I like the proprietary OSes, but I see the business sense this can make.


  To continue to expand the reach of Teradata and provide seamless access to Teradata from industry standard, as well as emerging standards, the Teradata Utilities Foundation (TUF, which is the tools and utilities) are/will be incorporating support for Linux.  


That sounds really awesome, any ideas about when on those tools?


  64-bit:  


  Like 32-bit, support for both Unix and Windows 2000. For Unix, Teradata will have an Open Unix strategy, supporting one or more enterprise-class, industry accepted/demanded Unix variants. This was described at last years Partners Conference. NCR expects to make an announcement in the very near future.  


I missed the announcement, but it sounds as though they are looking to move the teradata system to a package of bynet hardware and the Teradata software w/drivers for multiple OSes. That would be really interesting in my opinion. I understand that you would still have to buy disk and cpu in a given config to get scalability, but I think that would be fairly straightforward given a certain platform and certain requirements.

This is partly where my interest in the Linux on S/390 comes into play.


  NCR is closely monitoring Linux, specifically for its ability to support a business-critical process such as Data Warehousing. The industry is pretty unanimous in their view that Linux may have some shortcomings (e.g. RAS) that keep it from being considered enterprise-class. The Linux Open Community, with the help of major vendors, are actively working to take Linux up the food chain. NCR will continue monitoring this progress.  


This is where I'll ask of NCR what was asked of SGI, if you want it to change, then why not change it? SGI went out and changed some things in the kernel, they added a port of their filesystem (xfs), and published a set of power tools that just install with the Red Hat distro. These are openly available, and they work quite well. The other part of the equation here is that SGI also worked on the kernel and submitted changes that were integrated into the kernel as a whole. Their work allowed for the new memory limits in machines (I think it's around 64 gig., now), they also made other enhancements (see oss.sgi.com for a listing). NCR obviously could do this and also produce their own set of extensions to the OS to make it what they want it to be. They obviously have the ability to produce an excellent OS, I don't think they would blink at improving one that could use the work.

Here is the rub, the only payoff here is if people use and propagate your extensions, and use your software with them. Otherwise, there really isn't any money in working on the Linux kernel, it's a (I just don't like this word) "community" deal.

I think NCR has some EXTREMELY valuable knowledge in these areas and it would be great to see that applied to something as promising as Linux. But I also go along with Marcus J. Ranum, he said at a talk I attended that he felt that, "Linux is killing innovation in OS development. No one goes out andwrites an OS anymore, they just go out and put together their own Linux distro." (that's as close as I remember) And I do agree with that, but I do see a lot of acceptance of Linux and it's usefulness is growing by the day. I will rate it as much more of a data center OS than Windows NT/2K, but I still see gaps in the walls. But again, as the Linux "community" evangelizes, if you want something to change, don't wait for someone else to do it for you....

No harm or malice meant here, my tone is one of curiousity and hope. Thanks for listening,

loadc



     
  <Prev Next>   <<First <Prev Next> Last>>  
 
 
 
 
 
 
 
 
  
  Top Home Privacy Feedback  
 
 
Copyright for the TeradataForum (TDATA-L), Manta BlueSky    
Copyright 2016 - All Rights Reserved    
Last Modified: 27 Dec 2016