VMS to HP-UX conversion - capacity planning requirements?

[Mott Given]

Are there any white papers, manuals, technical notes, or rules of thumb for estimating the capacity planning requirements in a VMS to HP-UX conversion?

What factors in the VMS enviroment would be important to consider in evaluating the capacity needs?

Mott Given

[Jan van den Ende]

Mott,

What kind of answer would you expect here?
I guess you ARE looking for horror stories!
As a start,
http://h71000.www7.hp.com/openvms/whitepapers/index.html
( courtesy Sue Skonetski, just a few entries down in this forum. I am sure Sue will not object passing this on!)



Well, some 'rule of thumb' from me:

Triple your support personnel (probably at half price, so say up cost 50%).
(at least) double your unplanned downtime.
(at least) multiply planned downtime by 10.

Increase security risks by (not specifiable large factor).

Re-read Kevin Mitnik's testimony:
"any M$ system is mine within 5 minutes. Hardly any *IX can withstand me over one hour. The only system I consistently cannot get into without some form of inside help is VMS"



but on the down-side: quite a lot of software is not available, or no longer supported.
Blame DEC/CPQ/HP management & marketing for that!

Several more stories at

http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=555089

hth

Proost.

Have one on me.

Jan

[Hein van den Heuvel]

YOu don't give us much to go on here do you!?
(not that it would help, we still wouldn't know :-).
Is this for example an Oracle DB application, a home grown number cruncher? Is it a more or less standard application for which hpux sizing information exists perhpas?
Are you coming of a 2P or a 32 P box?
Intending to make a stop at PA-Risc or straight to IPF?
Or the othe rway around... move to OpenVMS on IPF first, then consider moving to HPUX if and when you really have the need for that (are forced to do so but the all knowing management perhaps. Speaking of which... if the proverbial 'they' know you got to migrate, then 'they' surely know how to size already right? So just ask 'them'!

Ok all kiding aside... an IO bound job is still going to be IO bound no? So configure the disk susystem similrly for similar perforance. That normally would be erring on the safe side, because unix often does write caching/buffering, and vms generally does not (as it plays it safe!).
Is the application CPU bound? Well I would start out with assuming that a good (Ev69 2 Ghz or EV7 Alpha) is rouglhy equivalent to a single 1.5 Ghz IPF. For older Alpha's, say Ev6 @525 I would take 2 IPFs for every 3 Alpha chips. LArge SMP? Numa? The Superdome has better latency than the Wildfire (GS320), but worse than the Marvel (GS1280)

Just some thoughts! Be sure to involve sales, to help you further.
Also, our Tru64 friends have spend more time analyzing and preparing for an Alpha -> Hpux move and have material in for example:

http://h30097.www3.hp.com/transition/modules.html?jumpid=go/transition-modules

Specially if you have an Oracle database to transition you want to check that out.

hth,
Hein.

[Mott Given]

The systems being migrated from VMS to HP-UX are:
- 4 Compaq DS25 2-CPU servers running production at approximately 100% CPU busy (2 of the 4 servers are at one location, while the other 2 are at another location)
- 2 Compaq DS10 test/staging machines
- 2 COmpaq Alpha 1200 development machines

The machines that they will be migrated to are HP-UX Itanium machines. I am looking for any rules of thumb I can use for a capacity sizing or anybody's previous experince with this migration (i.e. the before and after servers used). I don't need to know anything about how to do the migration.

Mott

[Wim Van den Wyngaert]

MG (sounds good),

What are you running ? DBS/user programs/webservers/terminal users/packages ?

I would say that Unix with equivallent cpu power (take rules of Hein) would perform better but without safety belts and unix is bad for dividing the cpu over many concurrent processes (e.g. Sybase servers will eat what they). So, extra power may be needed.

Wim

[Mott Given]

I'm in the process of finding out more about the VMS applications/databases running on the machine.

Mott

[Antoniov.]

MG,
you can find more lucky solutions here
http://forums1.itrc.hp.com/service/forums/familyhome.do?familyId=117
where people love HP-UX.
I'm with other vms gurus about migration from vms to unix.

Good Luck :-|
Antonio Vigliotti

[Doug Phillips]

I'm in the process of finding out more about the VMS applications/databases running on the machine.


That's all that matters, isn't it? The platform you're running on now is irrelevant.

First, you determine what load the current system is carrying (applications, users, transactions, records in database, etc) and how that might change. Then you buy the hardware & software that fills the requirement.

There shouldn't be any "rules-of-thumb" for converting from one system to another, but if you find one, you would be wise to ignore it.

Doug

[Jan van den Ende]

Doug,


First, you determine what load the current system is carrying (applications, users, transactions, records in database, etc)


Well, CERTAINLY for production environments, I would add to that list (or start it with)
Availability, Stability, & Security.

And then these factors are to be given relative weights (which will differ for each case). After that, at least a decision will stand on somewhat more solid arguments.

hth

Proost.

Have one on me.

Jan

[Antoniov.]

MG,
here there the HP web page about multi os and migration
http://docs.hp.com/en/multios.html
I hope you can find what you are searching.

Antonio Vigliotti

[Jan van den Ende]

Mott,

I found some other reading that might be interesting.

http://h71000.www7.hp.com/openvms/whitepapers/Illuminata.pdf

Proost.

Have one on me.

Jan

[Daniel Fernandez Illan]

Mott,

Other link to migrate to Tru64 (it can help you):

http://h30097.www3.hp.com/docs/porting/HTML/openvms/TITLE.HTM


See you.
Daniel.

[Doug Phillips]

Jan,


I would add to that list (or start it with)
Availability, Stability, & Security.


Yes. Those important factors *must* be considered when choosing any platform.

Mott said that he was only looking for capacity planning, though, and wasn't concerned with the actual conversion. I take that to mean that the decision is probably already made. Most likely, his applications are available on HP-UX and the data conversion and such will not pose a problem.

Mott,

I guess you can tell we're all rather biased, so could you share with us why this move is happening?

-Doug

[Mott Given]

Thanks for all your help. The applications being ported were in-house applications.
To answer the last question, I'm not sure why they were being ported.

Mott Given