[Info-vax] Distributed Applications, Hashgraph, Automation

[Kerry Main]

> -----Original Message-----
> From: Info-vax [mailto:info-vax-bounces at rbnsn.com] On Behalf Of
> Stephen Hoffman via Info-vax
> Sent: February 22, 2018 5:57 PM
> To: info-vax at rbnsn.com
> Cc: Stephen Hoffman <seaohveh at hoffmanlabs.invalid>
> Subject: Re: [Info-vax] Distributed Applications, Hashgraph,
Automation
> 
> On 2018-02-22 22:00:59 +0000, DaveFroble said:
> 
> > Simon Clubley wrote:
> >> On 2018-02-21, DaveFroble <davef at tsoft-inc.com> wrote:
> >>> But, when talking about being able to "spin up" additional
resources
> on
> >>> demand, I have to ask first, what is the problem, and what type
and
> >>> amount of resources should be thrown at the problem.
> >>
> >> An example problem is when your website load has a normal base
> load but
> >> which can also vary dramatically depending on the season (eg:
> >> Christmas) or marketing days (eg: Black Friday), etc.
> >>
> >> Simon.
> >
> > Not a problem, here ..
> >
> > Frankly, for some needs, one can not purchase a system small/weak
> > enough to exactly fit the needs.  So, one might have excess
capability,
> > but if that is the smallest available, so what?
> >
> > Today's HW can be extremely capable.  Maybe not for everyone.
> >
> > It all comes back to, who needs certain things, and who doesn't.
It's
> > that breakdown that seems to be totally absent in these discussions.
> 
> What's also missing is that over-building server configurations is
> normal on OpenVMS; there's just no entry-level server available, and
> there really hasn't been one for a decade or two.   

Reality exists for Windows/Linux servers as well. Entry level ProLiant
servers end up with server utilization of less than 20% busy.

Reality is that the server HW has advanced so much, that the traditional
one bus app to one server model is really a waste.

Heck, even many Linux/Windows VM's we have migrated in the past are not
that much busier either.

> Of how capable or
> over-capable some of that hardware really is.   What's also missing is
> that not spinning up new servers for testing or for prototyping is
> normal on OpenVMS.   That even installing and configuring a new server
> is an involved task on OpenVMS.  

Using traditional "put the OS CD in and boot" that is true. Definitely
room for improvement.

Of course, the same is true for Windows and Linux environments as well.

Of course, for med to large environments, no one installs new OS's that
way anyway, so this is really only true for small environments.

In med to large environments, they use "gold" images (image saveset of
pre-built master image) or templates (VMware) to deploy new OS's.

In OpenVMS, you could also build master LD containers to image backup to
new OS, customize and reboot. Maybe 15-30 minutes start to finish?

Could also be part of cluster with common system disk which most other
platforms do not have (common start-ups etc.)

> That clustering has been an
> exceedingly expensive approach to license on OpenVMS.  

Apparently to be addressed in X86-64 license model.

> That clustering
> itself hasn't been further integrated and updated in OpenVMS.   That
> you can't gain access to a guest or a slice or a private server at a
> hosting provider fully online and within minutes, with OpenVMS.  

Cloud marketing hype.

Please tell me of any Customer who absolutely needs VM or OS's within
minutes and I will tell you a Customer who is not involved with their
business and does not know didley about capacity planning.

For those 10 or less Customers in the world that really need OS's within
minutes - go to Amazon and Azure (hopefully consistent latency is not an
issue for their App)

> Folks
> that are used to other platforms aren't used to these assumptions and
> these limits and these requirements; 

Cloud marketing hype. What you get within minutes is "IT lite" i.e. A VM
spin-up, but without any AV, no OS patching, no monitoring integrated
with your service desk, no backups, no security hardening, no firewalls.

Yes, you can add these on (extra costs), but this where cloud providers
insert "yes, additional planning is required". Duhh..

Internal IT Depts provide full service OS instances per their company
policies. Does it take longer? Sure. But it's much more integrated and
"ready" for a prod environment.

> the whole
> runs-on-commodity-hardware discussion is soon in play for everybody.
> OpenVMS and its apps are headed into a completely different world.

Agree.

> With actual entry-level hardware.  Whether any business takes
> profitable advantage of this?   I know of a number of folks running
> lightly-loaded rx2800 boxes that may well end up replacing them with
> servers that are a fraction of the size (toaster-sized or cartridges
or
> otherwise), and at a fraction of the hardware prices of the Integrity
> boxes that they'll be replacing.   Or where you can spin up and run a
> test system for tens of dollars a month, and somebody else deals with
> the hardware and the network and the rest of that.
> 

Lets not forget the SW costs are typically the biggest slice of the IT
stack by far - on every OS.

Small dual core server in a 2 node RAC cluster on Windows/Linux running
Oracle Server will likely cost about $200K for the Oracle licenses
alone.

Servers running at $3K-$10k are usually not that big of a factor.

Good news for Oracle on OpenVMS Integrity or Alpha Customers - their
Oracle server (AND Rdb) license costs should drop by 50% when they move
to X86-64.

This is because the Processor Multiplication Factor which Oracle uses to
stifle their competitive HW platforms is 1.0 for Integrity, Power and
Alpha, but is only 0.5 for X86-64. It makes no difference what the OS
is.

This alone will likely make a move from OpenVMS Integrity to OpenVMS
X86-64 a no brainer.

> These sorts of differences in costs and revenues and availabilities
> won't interest some accounting departments and some managers and
> some
> developers...
> 

You are comparing the now to the past, not the now to the future
announced plans.

As example - new license model (subscription based?), new virtualization
capabilities, new emulators etc.

> For folks with under-loaded Integrity boxes, I'd be looking at a range
> with some of the following at the low-end...
> 
> https://www.supermicro.com/products/system/Mini-ITX/SYS-E300-
> 8D.cfm
> https://www.supermicro.com/products/system/midtower/5028/SYS-
> 5028L-TN2.cfm
> https://www.supermicro.com/products/system/midtower/5028/SYS-
> 5028D-TN4T.cfm
> 
> This all also depends heavily on which processors, chipsets and I/O
> widgets gets supported by VSI, and what sorts of local and served
> storage and storage interconnects and protocols will be available and
> supported and necessary, of course.  FC, maybe SCS over DTLS, iSCSI
and
> maybe iSER, maybe FCIP, SMB, etc.
> 
> At the middle-range of OpenVMS usage, I know of data centers running
> racks of OpenVMS servers that might well end up with the whole
> environment replaced by a few 3U/6U-class boxes; blades and SSDs and
> all.
> 
> And I'd be surprised to not see a number of OpenVMS systems running
> hosted.  Some in production.  Some for testing.  Though this
definitely
> hinges on the VSI pricing and packaging practices for x86-64.
> 

No doubt existing OpenVMS hosting and remote support providers like SCI
(and others) will jump on the HCI movement which is taking over many
large data centers. It is a way to significantly reduce the
infrastructure and associated server/storage costs typically used for
hosting very high numbers of VM's. 

Note - SAN folks will not like this as HCI uses local storage with a
hypervisor on top to server the local drives to other cluster nodes. It
is not unlike OpenVMS clustering in a box.

Some HCI companies, like Nutanix (very popular now btw), support KVM
hypervisors which would be a nice fit for the planned virtualization
features planned for OpenVMS V9+ releases.

Reference:
<www.nutanix.com>

Note - while VMware is the most popular hypervisor today, many Customers
are really complaining about the high licensing costs, so KVM solutions
could become much more popular in the future. With a small company
called Red Hat developing KVM, you can be sure KVM will only get more
popular in the future. Yes, OpenVMS and VSI can hopefully ride the wave
..

Remember my earlier point about the next 10 years being all about
reducing SW costs?


Regards,

Kerry Main
Kerry dot main at starkgaming dot com