>> -----Original Message-----
>> From: [email protected] [mailto:[email protected]] On Behalf
>> Of Brian McKee
>> 
>> I think that's hopelessly naive - If you think they will have the same
>> ram on the shelf a year from now that shipped with the system, I've
>> got a bridge for you.

Allow me to rephrase:

Yes, if you have a 2yr old dell server, and you want to buy a memory or hard
drive upgrade kit from Dell, you can rest assured that it's available for
purchase.  They need to keep these things in stock anyway, for warranty
replacements, so they'll happily keep some number of them for sale too.  I
have done this many times, and I am doing it again now.

By the time the server is 3-5 years old, the parts might be more difficult
to identify or locate, or unavailable, because now you're approaching what
they would have called the end of life of the product.

The same is not true for commodity parts - drives from newegg etc - Because
the consumers who buy from these outlets are overwhelmingly looking for
what's available *now* and not looking for what was available 6 or 18 months
ago, if there's something "better" available now.  It's simply not worth
while for these mfgrs to keep their products available in this type of
market, several months or years after they were introduced.

A case can be made, if the system-branded products (Dell etc) cost over 2x
higher than the commodity products, that perhaps there's something to gain
by buying 1.5x or 2x as many commodity products and just keeping the extras
onhand as spares.  But there's one thing you're never going to escape:  Even
when commodity parts are standards compliant and supposedly interchangeable,
different implementations of some standard are sometimes still not
compatible (or buggy.)  The supposedly standard commodity parts were never
tested on this system, with this chipset, as thoroughly as the
system-branded product.

Take the wifi consortium for example.  Before the consortium existed,
802.11a,b, and g had already been around for years.  Many companies had
already produced implementations, and products on the market ... But
generally speaking, you needed a single brand of access point and clients,
because the different brands weren't compatible with each other, even though
they were all complying to the same standard.  Since they created the
consortium, if you want to put the "WiFi" logo on your product, you must
bring your product to the twice-annual gathering, and undergo compatibility
testing with just about every known chipset or product out there.  You're
allowed some level of failure, but too much, and you can't call your product
"WiFi."

The reason to pay for all one brand of components is because they're all one
brand of components.

If you stray from this, there is ground to be gained.  If you're lucky, you
may save money.  If you're slightly unlucky, such as Ski with his Dell &
Crucial memory, the 3rd party parts might simply not work at all.  That's
not too bad; the extent of damage is just time and money getting up to the
point of failure.  If you're really unlucky, you might get a system that you
think is working, but is actually destroying data.  This is a huge topic of
discussion in the ZFS discussion list - SSD's that ignore the cache flush
command, if used for ZIL, cause risk to your entire data pool.  The official
answer is to only buy Sun hardware, with Sun firmware that's certified.  But
the unofficial answer is a bunch of people with random websites, who post
various homebrew techniques of testing the cache flush command on drives,
and results from their tests, stating make, model, firmware level of the
drives tested.  You have to trust their results, or trust the method to
retest it yourself, or whatever.  

Or just buy the Sun hardware at 5x the price and rest assured you won't be
fired.

_______________________________________________
Tech mailing list
[email protected]
http://lopsa.org/cgi-bin/mailman/listinfo/tech
This list provided by the League of Professional System Administrators
 http://lopsa.org/

Reply via email to