[Info-vax] IBM nearing deal to acquire Red Hat

[Stephen Hoffman]

On 2019-04-30 11:52:54 +0000, IanD said:

> I never worked on machines with Galaxy enabled but didn't they give you 
> the ability to expand memory in fight?

Galaxy instances (guests) could acquire memory from a pool of 
unallocated memory, but there was no way to release that memory short 
of shutting down the instance.

> Hmmm, now that I think more about VMS and memory parameters that sounds 
> like a silly question but it's been a very long time since I looked at 
> Galaxy stuff so I'll let the silly question stay
> 
> I don't think VMware allows for dynamic memory alteration on a running 
> instance does it?
> 
> If the above question for Galaxy is no, then Galaxy probably offers no 
> advantage

It's possible to point to the lack of a hypervisor—Galaxy used SRM as 
its coordination—but that's a fairly arcane difference.

SRM posted a subset hardware configuration to the instances (guests), 
and the instances then cooperated.

Galaxy lacked enforcement against rogue or malicious accesses.  Keeping 
guests from stomping on each other is hard.  See efforts such as SGX 
and TrustZone, and see Qubes.  
http://www.cs.wayne.edu/fengwei/paper/sgxsev-hasp18.pdf  
https://www.qubes-os.org Etc.

Galaxy was never able to virtualize processors or memory or other 
resources.  IBM VM and other hypervisors have been offering that for 
many years.

The current equivalent would involve presenting subset ACPI data.  
Which is what hypervisors provide to the guests.


-- 
Pure Personal Opinion | HoffmanLabs LLC