Operating System availabiltiy

[Shivkumar]

Dear Sirs,

I have read somewhere that Unix OS has 90% availability, Windows 40% and IBM mainframe has 99.99%
availability. The reason was given that in Unix all applications share the same address space in memory while in IBM mainframe each applications runs in different memory address.

Does it mean that legacy mainframe had superior design than modern unix kernel ?

Thanks,
Shiv

[Steven E. Protter · Accepted solution]

Shalom Shiv,

To adequately answer your question, you'd have to provide a source.

What does 90% availability mean. I ran a webhosting company in the US. It consisted of four Linux machines and provided an uptime ratio of above 99%.

When I first started with HP-UX we were in a one machine shop and yet provided an uptime ratio of over 99%.

So I contend, based on this and many experiences posted here that what you heard was wrong.

A proprely run mainframe farm did provide good reliability, though it came at a high price.

Today Unix and Linux provide this reliability at a better price.

Even Windows probably does better than 40% uptime.

What is availability in the terms of your question and I'll answer more fully, though my bedtime approaches quickly.

SEP

[Chan 007]

Shiv,

Good one ..:-)

Mainframes are made using the CICS, which is very rough and tough to design the system's architecture. The more complex you have to pay more. The biggest limitation is the you can't develop/get a COTS based application for CICS based systems. They are more memory and processor dependedt. Also You may know the famous MIPS (Millions Instructions per Second) was more used in this architecture.

While Unix are based on RICS, which not that complex to build and design, also very easy to code/integrate with the modern technology. You name a product that can't be made compatible with Unix.

Also in the olden days there were no Clusters (HA), DR, RAC, Datagaurd sought of things, while the modern day/second you have more technology and more every second.

Also, ingerating a Mainframe with frontend is more difficult, while a free downloadable tomcat/appache will make any database to be webbased...!!!

Superiority doesn't come with OS, in the fastmoving world, we have to adopt for the latest technology for tomorrow.

Mainframe was polular in olden days where we have COBOL based applications runing.

Chan

[Shivkumar]

Sirs, What i read was that the mainframe operating system was designed for 99.99% availabilty. I believe mainframes are usually not rebooted for years and runs continuously without any downtime. It seems mainframes are more suited for dataprocessing application rather than user interface programs. Also, one of my friend had said the mainframe was used for mission critical application like moon landing.

regards,
shiv

[James R. Ferguson]

Hi Shiv:

What marketing hype have you been reading!? I'd bet it's from IBM itself.

These are senseless questions in my opinion. It's like asking "Is a plant or is an animal better suited to life?"

Is "availablity" simply a measure of time between reboots (or abends)? How do you measure "availability"? Do you measure how long an idle operating system remains running before a hardware failure? How about a before a software failure? What applications are available to run? Has the same application been ported to a Unix server that runs on a mainframe, and is it valid to compare their meantime between failures?

Mainframes (as represented not only by IBM but also the old "BUNCH" -- Burroughs, Univac, NCR, Control Data, and Honeywell) are/were proprietary hardware-based each with its own operating system and machine language.

Mainframes were closed environments. For the most part, hardware came from the manufacturer at a very high monetary cost.

Software existed in a closed realm too. Mainframe configuration files were generally binary files to conserve space. Vendor-supplied, proprietary tools were the principal way to manipulate the file's contents. The availability of documentation (or even source code) was scarce and you paid ($$$) for it.

Superior design? No, different with a whole different set of fundamental assumptions. Wrong? No, just different.

Regards!

...JRF...

[A. Clay Stephenson]

Well, I can tell you that I went 5.8 years with zero unplanned downtime on HP-UX and when the system finally failed it had nothing to do with HP-UX but was deliberately (if inadvertantly sabotaged). I've never had numbers like that under any other OS. I've also had zero unplanned downtime since that incident.

[Arunvijai_4]

Hi Shiv,

Mainframes are used to data processing and it has got proven ability for running mission critical applications. Neverthless, HPUX also proven to its own capabilities beating several other OS in the market. I would say only Windows has lesser % of availability.

-Arun

[Yogeeraj_1]

hi,

You should verify the sources. Even that you should recognise that every environment is unique and different. The HUMAN FACTOR should not be neglected also.

you should also not forget the 5 pillars of ISO Network Management Model: Configuration, Accounting, Performance, Security and Fault.

hope this helps too!
kind regards
yogeeraj

[Arunvijai_4]

Hi Shiv,

Most modern Unix kernel are outperforming mainframes including HP-UX when it comes to enterprise applications. Also, TCO is very less in Unix than of Mainframes.

A good link about Mainframes and other related stuffs, http://en.wikipedia.org/wiki/Mainframe

-Arun

[Jan van den Ende]

Shiv,

a real-live story on System Availability can be found at

http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=855602

and yes, as of today it ia still going strong.

You may notice, this is NOT a mainframe, but the system with the lowest 5-year TCO:

http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=855602

Proost.

Have one on me.

jpe

[Steven E. Protter]

Shiv,

I would suggest that what you have been reading is the opposite of helpful.

Remember marketers only care about selling things. They don't care about whether they work or not long term.

The world's largest network, the Internet is based on Unix/Linux servers. Yes there is some windows out there but this is one arena where Microsoft does not dominate.

Based on that criteria, and the fact that nobody wants their website down, its safe to say what the most reliable OS is.

SEP

[MarkSyder]

Shiv,

I hate Windows and do not understand how it has achieved the dominance it has, but one thing is certain: if it only had 40% availability it would not have achieved that dominance.

Mark Syder (like the drink but spelt different)

[Arunvijai_4]

Hi Shiv,

Another good link, http://www.cert.org/tech_tips/choose_operating_sys.html

-Arun

[Frank de Vries]

I work already 7 years in outsourcing
HP Unix servers , usually in cluster configuration,
and we have contract with our customers
for 4 nine's 99.99 or 5 nines 99.999%
unplanned downtime.
And sofar I have very few occurances
did we did not meet our targets.

Of course it is not just the hardware,
it is also your fault-tolerant configuration,
identifying single point of failure, and improving that, and of course the alert-monitoring.

So it is certainly not as black and white
that Mainframe was designed for this and
Unix was not.

If your organisation has the proper ITIL
philosophy and understand how to set up
a High Availability system, the type of
hardware is of lesser important.
(The same applies to windows servers,
you only need to make precautions)

So everything can be configured to be
High availability.

[Eric Antunes]

Hi Shiv,

I'm also running 2 HP-UX servers from 1999 and had only one unscheduled downtime since then due to hardware fault.

PS: you should always compare different products by the profit/cost ratio...

Best Regards,

Eric Antunes

[Arunvijai_4]

Hi Shiv,

Some good links about OS high availability,

http://www.networkworld.com/topics/operating-systems.html

http://whitepapers.zdnet.co.uk/0,39025945,60085897p-39000482q,00.htm

-Arun

[Geoff Wild]

The mainframe is not an OS - it's an architecture - so you can't comapare it to an Unix kernel.

As far as OS'es go - what runs on a mainframe is just different - just like Real Time OS'es - for example - QNX - which, among other benefits - allowed you to upgrade the kernel without a reboot!

Some Wiki links for you:

http://en.wikipedia.org/wiki/Mainframe

http://en.wikipedia.org/wiki/QNX

http://en.wikipedia.org/wiki/Unix

http://en.wikipedia.org/wiki/Availability

Rgds...Geoff

[Tor-Arne Nostdal]

We'll I've lived through disk crashes on both IBM mainframes as well as on UX (as well as several other OS's).

Availability is a matter of perception for the users - "am I able to use my application".

For this we're talking about "fault tolerance", redundancy and failover.

What's for sure is that an error WILL occure. No matter which system you're running. The question will be "How do you handle this".

The OS is just a brick in the total puzzle that brings service for an end-user.

If you configure your systems smart, the end-users should get 99.99% availability, no matter which OS you run.
The question will be if your organisation want to pay for it.

Some hints:
* Power supply: UPS, Diesel aggregates
* Coolers/ventilation
* Dual computerroom's
* Alarm detections - 24x7 standby personnel
* External Network: several external providers, satellite
* Internal Network: duplicated routers/switches
* Machines: multiple CPU's, power supplies on dedicated circuits, HW-monitoring
* SAN's with mirrored disk systems
* cluster configured systems

And finally but not least, SysAdmins that use their brains when managing the systems.

/2r-arne

[Dave La Mar]

Shiv -
Just my 2 cents worth as well. Having worked in "Big Blue" shops more than 30 years between two companies, I have found myself biased at times.
As others have stated, "mainframe" can refer to different hardware and different O/S, thus a direct comparison in many respects is impossible.
System availability is a relative term more tyed to "Application" availability than hardware or O/S uptime. A common user does not access the O/S directly, but rather an application regardless of the O/S hardware or platform.
The best O/S and hardware uptime means nothing if the "Application" uptime does not support the user's need.
Six years ago, HP-UX got their foot in the door where I am currently employed. I can vouch for a stable O/S as well as hardware. Response time, on the other hand, for the "Applications" running under the O/S is well below the 'mainframe' CICS "Applications". Now is that because HP-UX and the associated hardware are inferior? To that I say a resounding NO! It is due to poorly written "Applications" that run under the O/S. In the changes in computing over the years, less time has been spent in efficiency and more time in functionality. Greater functionality is directly proportional to efficiency. I firmly believe this whether the "Application" is running on HP-UX, or OS390 (IBM mainframe), or VM/VSE (IBM mainframe).
I can state, that doing the same type of business, we now have twice the staff and twice the hardware to accomodate the "Applications" required to survive in today's technical demands.
Is that good or bad? ==>Well ask yourself, 'Is my business static and we are happy with our market share?' or 'Are we in a competative business that needs to stay up to date with the rest of the business world?'
Sorry for the long rhetoric but I could literally discuss this in length for hours.

Best regards,

-dl