Re: 3 parameter/onstat questions

Hi Floyd, I apologize for not getting back sooner. I'm afraid that your informix contact confused bc 1 with bc 2. The change to enable the new page structure (BC 1) is the behavior described by your Informix contact below. BC2 on the other hand marks all dbspaces and chunks to use the "Big Chunk" page format, and the size of the chunk doesn't matter. Essentially though all we do at the time the command is issued, is change the status for each chunk. We don't change the physical pages for data or index, until they are read in, and written back out. So to have every used data page switched to the new page structure immediately (BC2), you would essentially have to use the same trick that is used to force in-place alters to complete. Now the place where the Old Page/ New Page has the biggest impact is the logical logs. Since they are all circular, you don't have to do anything, once your logs wrap, all of their pages will be in the new format. Hope this Helps. -Mark ----- Original Message ---- From: Floyd Wellershaus <floyd@fwellers.com> To: Mark Jamison <majp51@yahoo.com>; informix-list@iiug.org Sent: Thursday, April 10, 2008 5:04:06 AM Subject: Re: 3 parameter/onstat questions Thanks. I just got off the phone with Informix. They told me that the change ( bc 2 ), doesn't change any of the existing chunks that are 2 gigs and less in size. It only works for new chunks created after it is applied. I guess that's why it works so quickly. ----- Original Message ----- Subject: Re: 3 parameter/onstat questions From: "Mark Jamison" <majp51@yahoo.com> Date: Wed, April 9, 2008 15:56 Floyd Wellershaus <floyd@fwellers.com> wrote: I just got off the chat-with-labs with John Miller. He left me with 3 questions I am hoping to get more info about, having failed to get anything useful from the version 10 docs. I'm on ids10.0.fc5. 1) He strongly recommended having the chunk mode set to expanded 'always'. He said that if it's not that way, I am suffering in performance because there is a translation that needs to go on with each IO between the old type of IO and the new type. ( something like that ). --So is this correct for my version ? -- If it is, then onmode -BC 2 would do it, is there any danger to that, or does it take a while , or is it instantaneous. ( yes I know I can check, but I would have to make an instance to check it on ). Hi Floyd, Answering the above. Yes, it is correct for your version. In addition to improving performance, it also means you use less memory, and can avoid certain issues that can arise if you run out of memory. In order to enable Expanded chunks, you have to first run onmode -BC 1 before trying to run onmode -BC 2. In practical usage, it takes zero time, and applies an in place style of change for the pages on disk. ------------------------------------------------------------------------------------ Praise the Lord oh my soul, let my inmost being praise His holy name. -Psalm 103:1