Results 1 to 7 of 7
  1. #1
    Join Date
    Feb 2014

    Unanswered: Migrating SAP from Mainframe to Blades - I ask for your experiences

    Greetings, dear fellow dbadmins and dbadmi(n)stresses ,

    please let me introduce myself, so that you can get a complete overview about my current situation:

    I am dbadmin for DB2/zOS on an IBM z114 in the IT-Department of a german City administration. I have inherited the DB2-admin position from a collegue who retired. He had about 20 years experience with Mainframes and DB2 in general, and 10 years with administering SAP-DB2 on z/OS Databases, so he left me with a very clean system. Before I took over this work, I worked closely with him, because I did SAP-Basis (without the DB-part).

    Half of the z belongs to DB2/zOS, which runs about 20 SAP-Databases on 1 CP-Processor with 73 MSU and one ZIIP-Processor. Sheer data volume is about 5TB, with the biggest Prod-Database about 500GB. The other half belongs to zVM, where the corresponding 20 SAP-Dialog-Instances run on 4 IFL-processors under zLinux with RAM ranging from 1 – 25GB. We abolished Legacy about 8 years ago.

    We have a Backup-/Reorg-/Recovery-/DB-Copy-Suite from a small firm in Hamburg, which is controlled with UC4 (now Automic) Job Scheduling. The Suite uses only DB2 utilities and runs very smoothly.

    We have usually about 600 – 800 concurrent users on our 4 main SAP-Prod-Systems. Dialog response on the main P-System is about 800ms (here I have to mention that this system runs 3rd-party-SAP applications, which are in part poorly and inefficiently coded). Dialog response on P-HR is about 400-500 ms. Dialog response time on a 3rd P-System, which we run for a customer who uses only SAP-Financial standard components is 150 – 200ms.

    We have completed SAP-Upgrades from SAP R3 4.7 ExtensionSet 110 (where we started) to ExtensionSet200 to ERP toERP EHP5/6 to Unicode in the last 10 years (I started in 2003 as SAP-Basis Consultant, before I took over my current position in 2009/10). We have never missed the deadline for the downtime.

    Speaking of downtime: We had NO unplanned downtime, neither hardware nor software, in the last ten years! Neither did we ever have to recover a broken database. And this is an absolute necessity since our „customer“, the city administration, relies on full 100% availability during business hours when we have heavy „customer“ traffic by which I mean the citizens doing their business. Also the police relies on full availability on the weekend or evening when the popular local soccer team plays at home.

    Before the z114 we had a z10, before that an z990 and before that a z800. On that machine we started with SAP on zLinux with DB2 on z/OS.

    We administer this system with 1 1/3 persons, myself for DB2/dfhsm/dfsms, the Backup/Recovery/Copy, UC4 etc., and a part time colleague, who builds in the PTFs, APARS etc.

    We do IPL („reboot“, for the non-Mainframers) about 2 times a year.

    In short: the whole system works as smoothly and efficiently as a swiss clock…

    …an EXPENSIVE swiss clock.

    Now you might guess why I post this here: there are people in our IT-department who think, they can get the SAME track record of the past 10 years, which I laid out above, also for the future…just cheaper, because we use another platform.

    VMWare on x86_64 Xeon E5-12C was mentioned. OS and DB are still to discuss.

    Now, dear Intel crowd: please do not misunderstand me. Though I love my mainframe I am none of the „haughty“ MF-Type, and I see, that there have been great developments in the „open“ world.

    But if you have followed my long post to this point, you may understand, that from my current point of view, any migration from the status quo can lead only ONE way – downhill, for a variety of reasons.

    So any sharing of experiences with migration of SAP-Databases from Mainframe to other hardware/software platforms would be appreciated. Also any experiences of running SAP-databases on Blades under whatever OS or whatever DB.

    I e.g. already have gotten some seemingly valid information that database size on DB2 LUW would be triple the size of db-size on DB2/zOS…can somebody confirm this? How many full backups do you have to keep on disk on other database systems, before you migrate it to tape? How reliant is your OS/DB combination running SAP? How is the availability? Unplanned downtime? Corrupt Databases? Restore time?

    Any help would be appreciated!

    Yours sincerely

  2. #2
    Join Date
    Jun 2003
    Toronto, Canada
    Provided Answers: 1
    I don't have much experience with SAP, but generally a set of hardware comparable with System z on performance and reliability usually is also comparable on price. If there are no other reasons to migrate except perceived cost savings I'd probably choose not to. People tend to ignore the costs of migrating a smoothly performing system, which are often very significant.
    "It does not work" is not a valid problem statement.

  3. #3
    Join Date
    Feb 2014

    thank you. Exactly what I am telling my "pointy haired boss"es for the past 2 weeks. Actually, I am feeling like my screen opened and sucked me into some weird Dilbert cartoon.

  4. #4
    Join Date
    Apr 2012
    I have supported DB2 on the Z/OS, AIX, Linux and Windows. Z/OS has the best availability without a doubt. AIX would come in second in my point of view, but does still require upgrade outages (yes you can do rolling but not as clean as on Z/OS). Linux upgrade down time issues as well, more outages in my experience. Windows...least reliable. If you have the infrastructure already for Z/OS and require 99.9999, then it is a nobrainer, stay on Z/OS. Oh, and if you have to migrate off of Z/OS you will need to export (or use federated db's) all of the data since it is a different characterset, can be done but a pain..

  5. #5
    Join Date
    Jan 2009
    Zoetermeer, Holland
    Quote Originally Posted by Stadtoberamtsrat View Post
    ... sucked me into some weird Dilbert cartoon.
    So, the management want to cut cost and downgrade from mainframe to dinky-toy. As a simple engineer you are not able to stop that process so do not try to and yes: they WILL pay in the form of weaker uptimes... But when that becomes clear to them in some excel-cell it is too late anyhow.

    What you can do: have a lot of fun in this next step in your career, learn a lot of new stuff and enjoy an interesting project which will keep you occupied for .... months? years? No boring routine work for you

    (yep: always look at the bright side of life)
    Somewhere between " too small" and " too large" lies the size that is just right.
    - Scott Hayes

  6. #6
    Join Date
    Dec 2007
    Richmond, VA
    Provided Answers: 5
    Don't forget the very extended downtime to swap over, unless you do some creative copying and using q-rep/cdc(if this can be added to the budget) something along those lines that can capture the changes from the time you make your copies until you get all loaded on new and ready to start receiving the changes. If budget doesn't support the purchase of some tooling to help you out, then you would be left with a homegrown solution to capture info, get your thinking cap on, or be stuck with the extneded down time that I first mentioned.
    Also, has the staffing been taken into consideration? I know of a few places that went this route and had to spread the applications out on so many servers to get same type of response time, that they had to hire additional DBA staff.

  7. #7
    Join Date
    Feb 2014
    @azready, dr_te_z, dav1mo

    Thank you very much for your information and opinions.

    @azready - we rely heavily on 99.9999, but only during business hours, i.e. more or less 6 am to 6 pm, and on some weekends when the local soccer team plays at home, then the police has to have access to some data. So normally we can have planned downtime easily almost every evening during the night, or on Saturday/Sunday. Unplanned downtime is anathema - they had an unplanned downtime in a neighbouring city about twice the size of my town just last week...the citizen's centers, where the people do business for e.g. drivers licenses, passports and the like, could not operate for 1 1/2 days, and this made regional news, both in print and radio. Municipal politicians positively hate this kind of publicity.

    @dr_te_z - well, actually I have found something I can try to put an end to this...I have convinced my department head to let IBM perform an EAGLE TCO Analysis ( There IBM will - as I understand it - analyze our present costs and compare it to another platform which would be able to handle our workload with the same reliability and response times. Maybe this might help...does anyone have experience with that?

    @dav1mo - of course, staffing has not been considerated. As I mentioned, we manage to administer the equivalent of 40+ Servers with 1 1/3 persons due to highly efficient automatizacion and the reliability of the platform... well, the department head has asked my 1/3 coleague, if she would like to increase hours...

    Again, thank you!

Tags for this Thread

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts