[Info-vax] VSI: "Official 8.4-1H1 Launch"

[Jan-Erik Soderholm]

johnwallace4 at yahoo.co.uk skrev den 2015-06-06 21:24:
> On Saturday, 6 June 2015 18:11:27 UTC+1, John Reagan  wrote:
>> On Saturday, June 6, 2015 at 1:00:42 PM UTC-4, Jan-Erik Soderholm wrote:
>>> Simon Clubley skrev den 2015-06-06 13:18:
>>>> On 2015-06-05, johnwallace4 <johnwallace4> wrote:
>>>>>
>>>>> Things we now know that we didn't know this time yesterday:
>>>>> * VMS will indeed (as JES assumed) be the guest under KVM. Makes sense
>>>>> to me.
>>>>
>>>> I wonder what kind of jitter/latency this will cause for real time
>>>> applications ?
>>>>
>>>> Simon.
>>>>
>>>
>>> I had the same thought, but as I understood Clairs post, they would
>>> also provide a native (non-VM) VMS port for x86 (?).
>>>
>>
>> Yes.  Both.
>>
>>> Now, it also depends on the definition of "real time". I do not
>>> expect to see x86 VMS systems of any sort having dedicated
>>> communication or DAC/ADC cards or similar. Most of that kind
>>> of things has moved "down" to PLCs or other embedded systems.
>>
>> You don't have such on Itanium or Alpha today, do you?
>>
>> NonStop on x86 doesn't use/need any special hardware.
>
> [Lots of words, no TLDR, sorry]
>
> I assume (!) Jan-Erik isn't familiar with National Instruments
> product range (hardware and software).

Assume as much as you like. :-)

> Lots of people round here
> probably aren't, which is fair enough. It'd be nice if there were
> one or two users around here...
>
> In one paragraph: any lab or instrumentation or automation-style
> IO you fancy (within reason), typically controlled by computers on
> CompactPCI cards (think PCI electricals in VME-style cards, or look
> it up).

The cards take to role of the "real time" environment.
They then talk to the LabView applications using NI's
drivers and other software.

So any "jitter/latency" on the OS or application level
is of less importance to the instrumentation control.

In the old times, cards in PDPs or VAX'es was less smart
and relied on the OS and applications to be "real time".

But again, it 100% depends on your definition of "real time".

For some, getting a screen reply in 2-3 sec is "real time",
for others not acting on a signal in 1 ms is *not* "real time".

The whole problem here is that Simons original question about
"real time applications" is impossible to answer without
definitions. Real time or not, you have to put your own
(quantifed) expections up against what you actualy have.

But yes, it would be hard to expect no extra time from the
additional software layers (host OS and KVM) at all...
Questions is, does it matter? Yes it theoretically can.
No, in most cases it will probably not matter.

Jan-Erik.