Patch availability change
in [Solaris]
|
Prev: <Beginner Question> - lu /sst
Next: difference between sun cluster "/dev/did" and "/dev/global" devices
From: jimp on 12 Feb 2010 17:13 Tim Bradshaw <tfb(a)tfeb.org> wrote: > On 2010-02-12 19:50:02 +0000, jimp(a)specsol.spam.sux.com said: > >> Yeah, so what? > > You will almost certainly be paying a license per core (possibly socket > if you are lucky) for the RDBMS. That's just how life is: people with > big machines pay more for their software support as well as their > hardware support. It may not be fair, and it's particularly painful > for hobbyists (generally) who discover they can buy old, many-socket, > systems but can't get a cheap patch-download contract for them. > Unfortunately basically no vendors care about those people. It's also > hard to see how they would go about making exceptions for them - how do > they distinguish between some big organisation with a couple of E4500s > sitting somewhere running some legacy thing, who probably should not be > getting a cheapo license, and me with the same machines sitting in my > garage, which I bought for peanuts and like to play with? > > There's actually no real reason they *should* care - I'm not really > helping them any with my vintage hardware, so why should they bother? > The case where they should care is if I *am* helping them - perhaps I'm > a software developer who it would be in their interests to support. > I'm guessing (I don't know, though I think this used to be the case) > that you can get special developer contracts which allow for just this > sort of thing. Or in other words, set the pricing by the apparent deepness of the customer's pocket. FYI, most businesses don't run out and upgrade their computers when a new model comes out. There are still a lot of places running things like the V440 which were purchased new. -- Jim Pennino Remove .spam.sux to reply.
From: Tim Bradshaw on 13 Feb 2010 11:43 On 2010-02-12 22:13:11 +0000, jimp(a)specsol.spam.sux.com said: > Or in other words, set the pricing by the apparent deepness of the > customer's pocket. Well, I think more accurately set the pricing to correspond to some approximation to the "amount of computing power" the system had when it was new. (The reason for the "when it was new" is that otherwise old machines end up being very cheap to support which both discourges people from upgrading, which is bad for business, and also does not correspond to the cost of supporting these old systems, which generally goes up over time.) > > FYI, most businesses don't run out and upgrade their computers when > a new model comes out. > > There are still a lot of places running things like the V440 which were > purchased new. I'm aware of this, of course: I don't know what the age-curve looks like but I suspect the median would be 3-5 years old, with a very long tail (somewhere there will be Sun3s in production use I am sure: fairly recently someone I know was involved in decomissioning a system which turned out to be a PDP11). I don;t think it affects the pricing model particularly: in particular you *don't* want to make old machines very cheap for commercial use, for the reasons given above.
From: jimp on 13 Feb 2010 12:15 Tim Bradshaw <tfb(a)tfeb.org> wrote: > On 2010-02-12 22:13:11 +0000, jimp(a)specsol.spam.sux.com said: > >> Or in other words, set the pricing by the apparent deepness of the >> customer's pocket. > > Well, I think more accurately set the pricing to correspond to some > approximation to the "amount of computing power" the system had when it > was new. Or in other words, set the pricing by the apparent deepness of the customer's pocket. > (The reason for the "when it was new" is that otherwise old machines > end up being very cheap to support which both discourges people from > upgrading, which is bad for business, and also does not correspond to > the cost of supporting these old systems, which generally goes up over > time.) Nonsense. The fact that Solaris 10 patches work on SunFire V100's (likely the cheapest SPARC machine Sun ever made), V440's, as well as the latest M9000's doesn't add any cost. And, since the software support contract for an X CPU V440 is the same as an X CPU T5120, OS support cost is irrelevant to upgrading to the end user. -- Jim Pennino Remove .spam.sux to reply.
From: Tim Bradshaw on 13 Feb 2010 12:57 On 2010-02-13 17:15:59 +0000, jimp(a)specsol.spam.sux.com said: > The fact that Solaris 10 patches work on SunFire V100's (likely the > cheapest SPARC machine Sun ever made), V440's, as well as the latest M9000's > doesn't add any cost. Yes it does. Any patch which may depend on the hardware will have to be tested on all the platforms it is intended for. New bugs may also be found in old hardware which require patches. Obviously most patches do not depend on the hardware platform, but some do, which requires you to have instance of that platform around. > > And, since the software support contract for an X CPU V440 is the same as > an X CPU T5120, OS support cost is irrelevant to upgrading to the end user. This would be true if a v440 with 4 sockets had the same performance as (say) a T5440 with 4 sockets. My guess is it doesn't, and the question would really be: how many V440s could you consolidate onto the T5440, which will then cost you the same in software support as a single one of the systems you've replaced?
From: jimp on 13 Feb 2010 19:30
Tim Bradshaw <tfb(a)tfeb.org> wrote: > On 2010-02-13 17:15:59 +0000, jimp(a)specsol.spam.sux.com said: > >> The fact that Solaris 10 patches work on SunFire V100's (likely the >> cheapest SPARC machine Sun ever made), V440's, as well as the latest M9000's >> doesn't add any cost. > > Yes it does. Any patch which may depend on the hardware will have to > be tested on all the platforms it is intended for. New bugs may also > be found in old hardware which require patches. Obviously most patches > do not depend on the hardware platform, but some do, which requires you > to have instance of that platform around. No, it doesn't. There is no reason to test patches on all the machines that have reached EOL, nor much motivation to develop hardware patches for machines that have reached EOL. >> And, since the software support contract for an X CPU V440 is the same as >> an X CPU T5120, OS support cost is irrelevant to upgrading to the end user. > > This would be true if a v440 with 4 sockets had the same performance as > (say) a T5440 with 4 sockets. My guess is it doesn't, and the question > would really be: how many V440s could you consolidate onto the T5440, > which will then cost you the same in software support as a single one > of the systems you've replaced? A V440 and a T5440 are not the same class machine. The entry level T5440 is 2 X 8 cores at the same speed (basically) as a V440, which could replace 4 4 CPU V440's. The support cost for the T5540 is $324 while the 4 V440's would cost $5,760, for a savings of $5,436. However the T5540 start at $40,295. So for $40,295 you can get the same performance and save $5,436/year on support and the T5540 will be long past EOL when you break even neglecting all the associated costs with moving to a different machine. That is NOT the way most people decide to do upgrades. ----------- Jim Pennino Remove .spam.sux to reply. |