SolMan 7.0 + IDES ERP 6.0 on 1 machine for 1-man-show with MaxDB?

[Former Member]

Sorry - I know that what I've done is unusual and somewhat suspect, but anyway: I've successfully installed -just for personal interest- both (SolMan 7.0 plus IDES ERP 6.0 SR3) on 1 machine, which is a very small PC with Windows Server 2003 x64 and MaxDB. I can start SolMan OR I can start ERP. But every time the used memory climbs up to about 10 to 12 MB.

If I would start both in parallel (but I hav'nt done that yet because I had not the courage to find it out), real + virtual memory might not be sufficient (and the mainboard of the pc does not allow more than 4 MB of memory).

Has everyone ever tried to downsize the default parameters of sapinst to the lowest possible values for a 1-man-show? I wunder what the minimum values for all the parameters of such a SolMan + ERP on MaxDB configuration might be, that would perfectly work if there would be only 1 user?

Is there any help or interest in this? If I could not receive helpful answers, it might be possible to me to provide the results of my furthers studies. But I would really prefere to make use of other collegues' experiences first. Thank you for your answer.

[lbreddemann · Accepted solution]

> Sorry - I know that what I've done is unusual and somewhat suspect, but anyway: I've successfully installed -just for personal interest- both (SolMan 7.0 plus IDES ERP 6.0 SR3) on 1 machine, which is a very small PC with Windows Server 2003 x64 and MaxDB. I can start SolMan OR I can start ERP. But every time the used memory climbs up to about 10 to 12 MB.

> If I would start both in parallel (but I hav'nt done that yet because I had not the courage to find it out), real + virtual memory might not be sufficient (and the mainboard of the pc does not allow more than 4 MB of memory).

Hi there,

first of all - you're talking about Gigabytes instead of Megabytes are you?

Thought so... Anyhow, it doesn't matter how much RAM you can put into the machine, since Windows supports virtual memory, a.k.a. a pagefile.

So basically it's not a question if the whole thing would run, but rather a performance question.

The next thing is: you didn't try it. Why not? What bad thing should happen?

Move on - try it out.

MaxDB will start as well with a very small data cache.

NetWeaver will start with only a few disp+work processes and perhaps you don't need a J2EE instance.

Depending on what you want to do, you can save a lot resources.

regards,

Lars

[Former Member · Accepted solution]

Hi Lars, you're right, memory installed is 4 GB - sorry for that.

With the page files I've learned that one cannot assign more than 4096 xB per partition. With 3 physical drives (=partitions) I'm limited to only 12 xB as maximum. Is it that I'm wrong?

Performance is not an issue to me: a 1-man-show will tolerate any delay if it would work at all.

In the profile I've already reduced the number of dia processes and I've got a reduction of about 1 xB.

Unfortunately I'm well trained in Microsoft SQL Server but not in MaxDB, and therefore I havn't known whether the 2 products could have been joined into a single MaxDB data base - therefore I actually have to start MaxDB twice. This may cost a lot of memory.

I'm quiet sure, that there are ways to reduce memory requirements, but I'm not familiar with the MaxDB parameters. If there would be a recommendation for setting them more convenient for a 1-man-show, I would like to check them out. But to try it on my one I'm afraid that all the work that I've done (about 2 weeks until now) could be lost - couldn't it?

[lbreddemann · Accepted solution]

Dear Heiner,

as far as I can remember there was a kind of "trick" to have multiple pagefiles per partition by putting them into folders - I would have to google that right now, so that's your turn

Anyhow, you won't damage your installation by lowering the e.g. DATA_CACHE size of the MaxDB instances.

The database will just have to perform more I/O then.

Concerning multiple products in one MaxDB instance - that's possible via MCOD, but not like MSSQL Server does it right now.

Anyhow, the overhead per instance is usually not an issue, if you configure them correctly.

best regards,

Lars