Velocity Reviews - Computer Hardware Reviews

Velocity Reviews > Newsgroups > Computing > NZ Computing > Anti-Virtualization

Reply
Thread Tools

Anti-Virtualization

 
 
Lawrence D'Oliveiro
Guest
Posts: n/a
 
      09-22-2009
With sufficiently good performance per watt, could you simply abandon the
virtualization trend and pack lots of cheap, cool-running physical CPUs
into the box?

<http://arstechnica.com/business/news/2009/09/arm-attacks-atom-with-2ghz-a9-can-servers-be-far-behind.ars>
 
Reply With Quote
 
 
 
 
thingy
Guest
Posts: n/a
 
      09-22-2009
On Sep 22, 12:44*pm, Lawrence D'Oliveiro <l...@geek-
central.gen.new_zealand> wrote:
> With sufficiently good performance per watt, could you simply abandon the
> virtualization trend and pack lots of cheap, cool-running physical CPUs
> into the box?
>
> <http://arstechnica.com/business/news/2009/09/arm-attacks-atom-with-2g...>


No....again you have a tendency to only look at the micro
level...there is more to a server in terms of heat output than just
the CPU. Then there are other aspects like high availability that
virtualisation can offer, hardware independence.....easy migration,
standard "hardware" platforms for the guests....snap shots....adequate
peak load safety margins....

Early blades were really meant to be what is discussed here, small,
low wattage boxes in quantity, Virtualisation achieves the same thing
only cheaper still....As an example we deploy a virtual guest for
about $1500~$2000, and deploy 40~50 of these in a 2 x 3U form factor
(3u for the server and 3u for its disk array...ie 1 tray of EMC SAN
disk. So about $90k for 50 guests....or about 30watts per guest
total...



 
Reply With Quote
 
 
 
 
Lawrence D'Oliveiro
Guest
Posts: n/a
 
      09-22-2009
In message <0c54938a-
http://www.velocityreviews.com/forums/(E-Mail Removed)>, thingy wrote:

> Then there are other aspects like high availability that
> virtualisation can offer ...


Virtualization does not offer high availability. If the hardware goes down,
ALL the virtual machines running on that hardware go down.

> Early blades were really meant to be what is discussed here, small,
> low wattage boxes in quantity, Virtualisation achieves the same thing
> only cheaper still....


Maybe not. That was one of the points in the article, in case you didn't
notice.

> ....or about 30watts per guest total...


ARM CPUs use a fraction of that. Which is the point.
 
Reply With Quote
 
Simon
Guest
Posts: n/a
 
      09-22-2009
On Sep 22, 4:58*pm, Lawrence D'Oliveiro <l...@geek-
central.gen.new_zealand> wrote:
> In message <0c54938a-
>
> (E-Mail Removed)>, thingy wrote:
> > Then there are other aspects like high availability that
> > virtualisation can offer ...

> Virtualization does not offer high availability. If the hardware goes down,
> ALL the virtual machines running on that hardware go down.


In a high availability configuration we run, it can. We have multiple
servers deployed in a HA cluster. Normally, the virtual machines run
distributed across the servers, however should one server physically
fail, then the VM's become operational on one of the remaining
operational servers.

 
Reply With Quote
 
Enkidu
Guest
Posts: n/a
 
      09-22-2009
Lawrence D'Oliveiro wrote:
> In message <0c54938a-
> (E-Mail Removed)>, thingy
> wrote:
>
>> Then there are other aspects like high availability that
>> virtualisation can offer ...

>
> Virtualization does not offer high availability. If the hardware goes
> down, ALL the virtual machines running on that hardware go down.
>

....and can be brought up seconds later on another host server.

Cheers,

Cliff

--

The Internet is interesting in that although the nicknames may change,
the same old personalities show through.
 
Reply With Quote
 
Lawrence D'Oliveiro
Guest
Posts: n/a
 
      09-22-2009
In message <(E-Mail Removed)>, Simon wrote:

> On Sep 22, 4:58 pm, Lawrence D'Oliveiro <(E-Mail Removed)_zealand> wrote:
>> In message <0c54938a-
>>
>> (E-Mail Removed)>, thingy wrote:
>>
>>> Then there are other aspects like high availability that
>>> virtualisation can offer ...

>>
>> Virtualization does not offer high availability. If the hardware goes
>> down, ALL the virtual machines running on that hardware go down.

>
> In a high availability configuration we run, it can. We have multiple
> servers deployed in a HA cluster. Normally, the virtual machines run
> distributed across the servers, however should one server physically
> fail, then the VM's become operational on one of the remaining
> operational servers.


But there is still a few seconds' downtime, is there not, for ALL of the VMs
affected by the outage? As opposed to only a single physical machine being
affected.

The fact remains, virtualization is inherently in conflict with high
availability.
 
Reply With Quote
 
Simon
Guest
Posts: n/a
 
      09-23-2009
On Sep 22, 8:19*pm, Lawrence D'Oliveiro <l...@geek-
central.gen.new_zealand> wrote:

> But there is still a few seconds' downtime, is there not, for ALL of the VMs
> affected by the outage? As opposed to only a single physical machine being
> affected.


Again, that depends upon the exact configuration. Our business has
mandated three nine's (99.9%) of reliability, so we actually have
around 10 minutes per week of allowable downtime. Should a failure
occur, this is more than sufficient for the fail-over to occur and the
inactive VM's to initialise. There are also a number of alternative
configurations that would allow near instantaneous initialisation, and
we are currently investigating these.

> The fact remains, virtualization is inherently in conflict with high
> availability.


No, not at all.
 
Reply With Quote
 
 
 
Reply

Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are On
Pingbacks are On
Refbacks are Off




Advertisments