8 Apr
2013
8 Apr
'13
10:42 p.m.
I do hate answering the obvious trolling, but I am weak willed so some
answers in lin
------ Original Message ------
From: "MG" <marcogb at xs4all.nl>
To: cctalk at classiccmp.org
Sent: 08/04/2013 21:14:48
Subject: Re: Tech is the biggest problem facing archiving ? The Register
I can't possibly comment...
... but others say VMWare is getting expensive compared to the Microsoft
Hyper-V.
No one outside IBM knows for sure. Getting figures for how many zSeries
boxes in use is pretty challenging..
I don't remember it being hard to obtain. I think the "overpriced"
argument only appeared when they produced the cheaper "windows only"
servers.
Poor software library was also an issue, so whilst you could get the
Microsoft Exchange server on Alpha, you need an Intel box to run the
Lotus Notes and Profs/SNADS connectors.
They match with the survey we did before we went to VMWare.
Do you remember where it may be. If so, could you please point me
to the results of that survey?
Sorry it was a survey of our internal sever estate, so its not available
on the public domain.
I feel Oracle (Sun as was) is going the same way as IBM zSeries. Not
only has our Mainframe gone but also most of our Oracle (SUN) servers.
VMWare lets us deliver Linux or Windows VMs as a service. Ring me up in
the morning and I'll usually have a new Red Hat or Windows server up and
running by lunch time ready for you to install your apps on. (it could
be made quicker but I don't do enough to make it worth while) I guess I
could also deliver Solaris X86 the same way. We don't have enough Oracle
(nee SUN) boxes to be able to deliver them at that rate, but in theory
Oracle Solaris can do the same.
VMware works just like IBM used to with 360 mainframes. It has such a
really low entry point (S/360 model 30 comes to mind), and generally
once you are in you are in and addicted. The smallest physical VMWare
server my employer has around $3,000 dollars and was used to consolidate
virtualize a small pile of desktop PCs used to manage a thin client
environment, and run a time management system. Total cost around $3000
dollars. I am also looking at some big boxes for a CPU intensive
application. These may well turn out at $25,000 or more each. If you are
really big you can pay twice that...
However provided the physical boxes are on the same VLAN and SAN I can
transparently move the RUNNING servers from the smallest box to the
largest with no impact to the user, no re-boot, and the OS won't even
notice the move.
http://www.vmware.com/uk/products/datacenter-virtualization/vsphere/vmotion…
Microsoft have a similar facility for Hyper-V called "live migration".
http://www.microsoft.com/en-gb/download/details.aspx?id=12601
Watching VMWare VMotion a VM between two ESX hosts is still something I
find difficult to believe, but the system does it all the time, to load
balance. It will also do dynamic power management where it automatically
consolidates the work load during times of low capacity demand. (I
haven't tried that yet)
Last time I looked IBM was promising this for a future release of zVM
but it may be here now. Of course IBM boxes are big enough that they
don't need it for performance and there is enough redundancy in a big
box for most needs..
On 8-apr-2013 0:13, Dave wrote:
They are big and expensive and difficult to justify for a small
business? Even with "Capacity on Demand" the entry level prices are such
that an SME (google for what percentage of the market are SME) can't
afford to get on the ladder. There is a lot more to VMWare (and Hyper-V)
than just Virtualization...
IBM sell and support VMWare on their Intel and AMD
servers.
Why don't they sell more z/VM on their "z" things instead of VMware
on "x" things? I mean, if mainframes are so future-proof, desirable
and perfect.
Because of the
"suits". By keeping zSeries as a niche product IBM
can charge premium prices and keep the profit margins high.
But wouldn't that mean that they're trying to line their pockets,
fill their wallets and freeload, milking out in so far things can
(still?) be milked out?
When we discarded our Mainframe, a small
Multiprise box IBM showed
no interest in selling us an upgrade. They are quite happy to keep
selling us xSeries servers though. By selling xSeries and iSeries
boxes "relatively" cheaply they can sell zSeries as a high
reliability,
high margin box.
But why would they if more andm ore are going to "i" and "x"?
I personally think this is what allows zSeries to
survive.
Do you know for sure, though? Some articles I've read seem to
suggest otherwise.
In my opinion
the biggest mistake DEC made was to try and take the
Alpha chip down market. It spoilt the brand image.
That's the first time I heard anyone say this before... Most people
complain that it was overpriced and (still) to hard to obtain, along
with (of course) a poor software library (e.g. under AXP versions of
Windows).
Another
interesting statistic is average capacity utilization:
Wintel-based servers: 8-15%
Unix/RISC: 28-45%
Mainframes: 65-75%
Where did you get these figures from?
IBM typically
sell you a box with more CPUs in that you need, and
charge
by CPU power so it makes sense to run those that are enabled pretty
near
flat out. If you need a boost then they sell you a code to allow you
to
enable more CPU speed.
HP seems to be offering that as well (also "pay-per-use" and other
setups).
But that's what I mean, what is IBM doing to make sure that they
still 'stand out' and are 'ahead of the curve'? To me, as a
spectator, it seems like it's going down hill.
It's slowly becoming a bi-polar, x86 and ARM, world. In my view,
IBM, HP, Oracle, etc. aren't really fighting it very hard either.
Maybe IBM a bit more than the latter two, but maybe it's even too
much for them (not even so much that they're allowing to let it
happen).
- MG
Hope thats all clear. Its past my early
bed time over here in East
Pondia...
Dave
G4UGM