[Info-vax] rx2800i2 sales/support window changes

[Paul Sture]

On 2014-02-14, johnwallace4 at yahoo.co.uk <johnwallace4 at yahoo.co.uk> wrote:
> On Friday, 14 February 2014 07:33:05 UTC, Paul Sture  wrote:
>> On 2014-02-14, johnson.eric at gmail.com <johnson.eric at gmail.com> wrote:
>>
>> > On Thursday, February 13, 2014 6:33:36 PM UTC-5, JF Mezei wrote:
>>
>> >
>>
>> >> Also, while the Tukwila system may have used many components that were
>> >> easily available (commodity) back in the 2007-2010 timeframe when the
>> >> system was designed, it does not mean that those components will still
>> >> easily commercially available for the long term. So again, they may want
>> >> to have a whole bunch of power supplies etc purchased.
>> >
>>
>> > Interesting. I hadn't considered this angle of it. But everything you say
>> > makes sense.
>>
>> To illustrate that kind of problem. in mid-2010 I bought a PC which came
>> with 2GB RAM occupying one of the 2 memory slots.  I could add another 2GB
>> for ~$50  but the 4GB units came in at $180 apiece.  It was a no brainer
>> to settle for 4GB maximum.
>>
>>
>>
>> However 18 months later I really wanted more RAM to run virtual machine
>> instances, so the choice was between taking the hit on those 4GB units
>> (and of course hoping their price had dropped) or buying a second system.
>>
>> By that time I couldn't find any of those 4GB units, let alone get a price
>> on them.  I imagine anyone who really needed those units had already got
>> them or was at the mercy of suppliers who knew how hard they were to find.
>
> Yes, but :)
>
> This was presumably a standard desktop PC, not very many slots,
> non-ECC RAM, usual low end stuff?

Tower system but otherwise true.

> Server boxes come with more RAM slots so the "throw it away" question
> is less likely to arise. Server RAM (ECC, registered?) comes more
> expensive (more profitable?) up front than desktop RAM. Server kit
> should be expected to have longer lifecycles than desktop kit.
>
> Add all those together and the obsolesecence problem doesn't go away
> but it might have more extended timescales before component availability
> becomes a major problem.

I cannot argue with that, but you do put yourself at the mercy of suppliers
who can charge a premium for such stuff.

> I'm not sure extending the life of IA64 hardware availability is a *good*
> idea, but I am saying that availability and manufacturing issues may not
> be as big an issue for servers (even IA64 servers) as they are in the
> volume commodity PC market.

OK.

--
Paul Sture