max dbms cache

Although the maximum advertised cache memory for Navision’s proprietary Database Server is 1 GB, there is a known bug which sometimes prevents values above 800 MB from being used. Im currently experiencing this. Im experiencing slower running reports,etc. Below sone specs, Thanks, FAIZEL 3.10a win2003 server 4gb ram 60gb database 164 users

Hi! What do you mean with “prevents values above 800MB”? You cannot start the Server? You get an error message? Or does the server just allocate 800MB? If it’s the last option, be aware that if you use COMMIT cache this will use up to 30% of the assigned DBMS cache … And of course you have assured that there are no other major services run on that machine (a Domain Controller or Printserver or something like this [xx(]) that would grab some RAM … !?

Hi, yes I cannot start the service. The max dbms cache I can assign is 796343. Im not able to get you the actual error message right now, (database is currently in use). Nope the machine is not a domain controller,etc, just being used for navision. Yes - the commit cache option has been assigned to YES. The task manager report 2.6gb available physical memory Thx, FAIZEL

Hmm … strange … I never faced this problem before … sorry

Probably an issue with 3.10a Speak to your reseller about getting 3.70B or 4.00. It should be free to upgrade the executables. This means functionality will still be 3.10a. Unfortunately, you will need to do a backup(3.10a) and restore(into newer database) which take the Navision off line for about 3-4hrs.

Forgot to mention that DBMS settings may not resolve your issue of slow running reports. A 60GB database should be split into as many slices as possible with each slice on a separate drive array on the server. What are details in that area?

Hi, the slices are: 3 x 10Gb files 6 x 5Gb files 1 x 2Gb file 1 x 1Gb file these are placed on a megaraid controller - 6 x 146gb disks on raid 10. Do you reckon the 10gb slices are too big ? Thx FAIZEL

Could there be any issue with 3.10a and server2003 ? May there are patches,etc, one needs to apply, this installlation does have server2003 sp1, but no patches/hotfixes applied to 3.10a. Faizel

quote:

Hi, the slices are: 3 x 10Gb files 6 x 5Gb files 1 x 2Gb file 1 x 1Gb file these are placed on a megaraid controller - 6 x 146gb disks on raid 10. Do you reckon the 10gb slices are too big ? Thx FAIZEL
Originally posted by faizelt - 2005 Aug 03 : 07:17:35

When Using RAID-10 the OS only get one drive and therefore Navision only work with one cache (slave.exe). If You convert to RAID-1 (so Your OS will see several disks) and create one db-file for each pair of disks You will see a huge perfomance increase. When using C/SIDE DBMS You should NEVER NEVER use anyting else then RAID-1. I don’t know how many times this has been discussed over the years now… Hope it helpes.

quote:

Hi, the slices are: 3 x 10Gb files 6 x 5Gb files 1 x 2Gb file 1 x 1Gb file FAIZEL
Originally posted by faizelt - 2005 Aug 03 : 01:17:35

Lars, I know you already said it, but… Faizel, what are you doing. This server need to be rebuilt. Please search this site for how to configure dirves. First of all every slice must be exactly the same size. The performance of that system will be a mess.

Yes, I would have to agree with David. The server should be rebuilt!!Each database slice should be evenly spread over the available drive arrays. Another trap is when you initially expand the database into these drive arrays, you must close the database after the expand has been completed and before you start the restore, otherwise the data will simply be restored in a series style, rather than in parallel. If the database has to be expanded more in the future, then expand every slice equally, not just the one slice. If a new drive is to be added, then repeat the build of the database again. BTW: I think RAID10 (1+0) gives the same performance as RAID1 anyway. http://www.acnc.com/04_01_10.html

Hi Jason, not sre if I agree wth you. [8D] I can’t see how RAID 10 can be as fast as RAID 1. ( I am not saying there should be a great deal of difference), but it only makes sense that Navision Server will be better at the logic of striping its data, than the contrroller, and it must be able to better use Commit Cache when spread across multiple elements of a strip, rahter than having just one cache for the whole drive array. Has there actually been in write test comparisons between RAID 10 and RAID 1 ? I would be very interested in the results if there are. PS the above comments relate to Navision C/SIDE server, I am not talking about SQL or any thing else. SQL can obviously make better use of hardware stripping than doing it in software. Especially when you compare CPU usage in Navision and SQL.