Thursday, June 28, 2012

Innovator-C: 446 Days of Continuous Uptime in Jeopardy


This morning Hannes Visagie posted to the IIUG SIGs about a change to the Innovator-C limitations listed on the IBM Informix Innovator-C page that states Innovator-C will be limited to run on 1 core, down from the previous limitation of 4 cores. I sure hope this refers to CPUVPs and not hardware cores in the server, because who has a single core machine handy these days?

Is this the end of the line for our Innovator-C instance? Time will only tell. If the above limitation is real and not just a typo (edit: confirmed not a typo) we will have to reconfigure the instance down to 1 CPUVP and bounce the engine to see if a single CPUVP Innovator-C instance can meet our requirements and the uptime clock will reset to 0. Oh well, I've still got plenty of Growth and Ultimate engines with over a year of uptime.

Even if we do have to adhere to this new limitation, I'm hopeful that we will be just fine. Unlike our Growth and Ultimate engines, the Innovator-C engine doesn't run very hot. top shows the single socket, 4 core CPU running pretty idle and onstat -g glo shows our CPUVPs are running at a high efficiency.

So, how do I feel about IBM making this change? I guess I have mixed feelings. On one hand, I hate to have to give up 3 CPUVPs worth of performance, but on the other hand I still can't beat the price. We opt to pay for support, which is VERY affordable, but if we didn't IBM would still let us run Innovator-C in production for FREE. Think about that for a second before getting too terribly upset about this change and equating this to Darth Vader altering the deal he had with Lando.

Some of you may be thinking "DB2 Express-C, the free version of DB2, is allowed to run on 2 cores and use 4GB of memory. Shouldn't Innovator-C at least have the same limitations?" That would be nice, but I don't think the have to be the same on paper. I don't think this move by IBM to change the Innovator-C requirements is an attempt to play a favorite to DB2 and not promote Informix (they have other ways of doing that, ZING!!!) If anything this is saying that Free Informix runs so well that we need to limit the hardware it runs on to 1/2 of what Free DB2 runs on to level the playing field. Edit: I see people saying similar things in the IIUG SIGs so I'm not the only one thinking this.

Despite the recent changes to the Innovator-C limitations I still think Innovator-C is a viable option for a lot of production environments. I would be really interested to see the Fastest Informix DBA contest or Eric's TPC-C benchmark run against a 4 CPUVP and 1 CPUVP Innovator-C installation. On second thought, maybe I don't. Maybe I should just keep my big mouth shut about Innovator-C before the word gets out about how powerful it is and it is limited further. This post will self destruct in 5...4...3...2...1...

7 comments:

  1. Some notes:

    1- If you downloaded one software that comes with a license that says 4 cores I don't think you'd need to reduce. While you can stick with that version, you should be fine
    2- You can remove CPU-VPs dynamically... so I don't think you'd need to restart..

    Regards

    ReplyDelete
  2. I agree with #1 and this was confirmed today. I'll be sticking with the version I initially downloaded until it goes out of support if I need the 4 CPUVPs and don't need any features or fixes released in later versions.

    On #2 you can't always dynamically remove CPUVPs. I attempted via 'onmode -p -3 cpu' on an idle test instance that had 4 CPUVPs configured and received

    onmode: error when trying to change the number of cpu virtual processors
    by -3. Check the online log for any additional cpu VP messages.

    and the following in the online.log

    12:59:34 Dynamically dropped 0 cpu VPs

    ReplyDelete
  3. Hmmmm.... Strange... Do you have the listeners in NET VPs?
    Did you try to remove them one by 1?
    I did a quick test and it worked as I'd expect.... But maybe there's something in your configuration that prevents it...

    Regards

    ReplyDelete
  4. TCP poll threads (soctcppoll) are running on NETVPs, but listeners (soctcplst) are running on CPUVPs, attempting to drop only 1 CPUVP had the same result (error).

    KAIO is enabled, maybe this is the cause.

    ReplyDelete
  5. According to the read me file for IDS1170FC5IE

    Innovator-C is now available on the production site:
    https://www14.software.ibm.com/webapp/iwm/web/reg/pick.do?source=swg-ifxnce&S_TACT=101G506W&lang=en_US

    Replication is no longer offered for ER and HDR with 11.70.FC5.

    ReplyDelete
  6. It's Mega Maid. She's gone from suck to blow.

    ReplyDelete
  7. I think IBM made a typo error on their web site concerning number of cores.

    From the new readme file.

    Will test next week

    H Visagie
    ---
    Changes to Replication in Informix 11.70.xC5 Innovator-C edition

    For 11.70.xC5, Informix Innovator-C edition no longer supports Enterprise Replication (ER) or High-availability data replication (HDR). If you were using replication in an older version of Innovator-C, you must disable replication after upgrading to 11.70.xC5.

    If you want to continue using replication in Informix 11.70.xC5, you can purchase the Choice, Growth or Ultimate edition.

    Removing Enterprise Replication

    If Enterprise Replication was defined for the previous version of the server, remove replication by running the following command:

    cdr delete server --force

    Removing HDR

    If High-availability data replication was defined for the previous version of the server, convert the server to a standard server by running the following command:

    onmode -d standard

    ReplyDelete