For someone complaining about the "lack of reality" of the current WO pricing, I find
it surprising that you would also state that "you don't consider NT a deployment
platform". :-) (empahsis on :-) I don't either-- but the client is always right
(though they usually change their mind after we
mention scalability issues the 1.5 to 2.5x cost of administration and that they will
have to pay for our travel every time we have to go to the physical server machine to
adjust something because of the lack of decent remote administration!).
If it runs Solaris 2.5 or 2.6, WO can be deployed on it... we have a client running wo
4.X (and 3.x, previously) quite successfully on some random sun clone. As far as I
know, nobody makes a clone of the sparc chipset-- just the Sun machines.... as such,
all the clones run genuine sparc
processors and, as such, tend to be very compatible with Sun machines.
As far as price/performance is concerned-- I have never seen an Intel based system of
any kind (Linux comes close... so did OpenStep 3.3pl1-- but only after some very
serious and often costly initial findling with drivers and hardware to come up with
the appropriate brew of components) that could
compare to a Solaris box in terms of stability, scalability and reliability.
Raw price/performance numbers are kinda like the whole licensing thing that led to
this entire conversation-- in and of themselves they aren't very useful because the
real costs are:
- cost of downtime
- cost of lack of reliability
- cost of scaling; intel scales well to a certain point, then you hit a wall
- cost of administration
Buying a solaris based environment is kinda like buying an insurance policy against
downtime or lack of reliability. As well, it can be scaled a hell of a lot further
than an Intel box (which is why Oracle bet $1,000,000 that Oracle is 70x faster than
Microsoft SQL Server.... did you see the
specs of the Sun box they ran their Oracle tests on???). In terms of administration,
the initial setup of a serious Solaris box can be costly-- but once it is running,
most common hardware failures are pretty cheap to address-- hot-swap a drive or power
supply and you are done. Because of the
relatively simple nature of the hardware (i.e. one or two kinds of this or that-- none
of that IRQ/IO address business), failred hardware is generally fairly easy to
identify and fix. The available back up solutions are amazing and incredibly
scalable, etc....
Too bad we don't have a webobjects-talk mailing list-- this is moving away from
devleopment issues, but is still a very pertinent discussion for those making
deployment and, even, development envitronment decisions.
[obviously, since NT is the only intel based choice for deploying WO, the cost of
administration of NT is "through the roof" compared to Mac OS X Server or Solaris--
especially if you have to send your admin to the stupid hosting site to flip the power
switch on a regular basis]
b.bum
David Fleskes wrote:
> >Deployment on NT (Intel), Solaris and HP-UX are available and fully supported under
>WO 4.0. The only platform >that has been lost between 3.5 and 4.0 has been Mach
>based Intel (yes, that is annoying). We have a number of >customers that are quite
>successfully deploying on Solaris right now.
>
> I wouldn't consider NT a deployment platform (development platform either).
>Deployment of WO on Solaris is available, but only on Sun's hardware (last I heard
>anyway). I did a price/performance comparison between sparc and intel... big
>difference.
>
> ------------------
> David Fleskes
