On Oct 11, 2007, at 10:45 PM, John C. Welch wrote:
On 10/08/2007 16:11 PM, "Dave Pooser"
<email@hidden> wrote:
I agree, but to play devil's advocate, a lot of the win in Apple's
mind comes from using Mac OS X Server *with* Xserve, in essentially
one supported configuration,
Except that it's also supported on Mac Pro, iMac G4/G5/Intel, Mac
Mini,
eMac, and Xserve G4/G5/G4 Cluster Node/G5 Cluster Node. So adding
a single
supported VM configuration isn't that much additional overhead.
Define "not that much"?
You still need sufficient "disk" to run the OS and whatever service
you
want. If your VM has to be 40GB or 80GB, that's what you're going
to need.
VM doesn't make that go away. You still need sufficient RAM, Disk I/O,
memory I/O, and network I/O to do the services the VM needs.
Not to mention that OS X is /very/ particular about I/O latency
issues. Think "spinning beachballs". Ever see one on even a lightly
loaded system?
For Windows, VMs are a no-brainer, because Windows has craptacular
memory
management, and due to the fragility of the registry and a host of
other bad
decisions, running multiple services, or even a single heavily loaded
service on a Windows box is a Really Bad Idea.
I think this is something w've both been stressing. VMs make
wonderful sense for running things where you need to partition off
servers. Under Windows this is pretty much a requirement for every
service. SOAs is nice but in that realm each service requires it's
own machine.
For Unix, the VM decision isn't as automatic. Unix-based OS's are
usually
more reliable than Windows, and definitely more repairable. They
have far
less management overhead than Windows per box, and unless you have
some
solid Active Directory and VBA/SMI skills, on a per-network basis too.
I know at my company, the VMWare for Unixen question isn't as much
of an
issue, because to virtualize even my Xserve G4 would put a real hit
on a VM
box. All my Unix servers run at 90% utilization before I even
*think* about
adding another box. Database, network, you name it. They all run
happy as
clams at 90% or higher utilization, so the problems you hit with
Windows at
these kinds of usage levels are not an issue.
And this explains, in part, why unix and other OSen have been so
popular over time. They're very capable of running processes/services/
demons effectively and if something does crap out it doesn't take out
the whole enchilada.
While *nix based systems /are/ good at this OS X still has a certain
amount of fragility. But in OS X's case that fragility is more
pointed towards issues that can be dealt with, basically the issues
revolve around resource allocation, and those are a killer to even
VM'ed instances on the same box, so the need for VM is less defined.
Of course since VM is *so hot* everyone thinks they need it. The need
really comes down to sandboxing. And most of that comes down to
wanting to have multiple boxes for development/test/production and
thse really should always be separate boxen anyway from a datacenter
mgmt perspective, security perspective, and resource allocation
perspective.
Virtualization, like everything else, solves a specific set of
problems. It
is NOT a spell of "never needing hardware again", nor is it a spell of
"magically running n OS instances for near-zero resource costs".
And there are several good reasons for wanting to have virtualized OS
X instances. However very few of these equate to the same reasons you
need them in Windows environments.
Really folks, this is a rehash of the concepts that brought about VM
30 years ago, which I know we all remember. And the reasons for not
using them were learned 25 years ago too.
It is no more the "perfect" solution for all needs than any other
"perfect"
solution has been. But it is getting the "Virtualization is a magic
spell"
press a lot, so I can see why people start to think it is.
I dunno John, anyone work for any press ;)
Seriously, an article about why VM isn't a panacea is a good topic,
if not only from the trolling perspective.
Heck, you could just repost an article from the past an who'd be the
wiser? Seems like today's IT ppl want to re-invent the past and its
problems. Soemtime's there's good reasons and sometimes it's just the
bandwagon.
