Pros/Cons of Local Workplace Install

[steve_lambermont2]

I currently have local developers running their own Netweaver Workplace instances (J2EE + MAXDB + NWDS). I can see supporting these local environments to be a headache. For example it takes 2 hours just to apply SP STACK11 to one pc assuming their pc configuration is consistent.

What is the true impact of users having to connect to the development server to develop their apps?

Is their an easier way to patch these local environments?

Any insightful comments are appreciated.

[detlev_beutner · Accepted solution]

Hi Steve,

Personally I definitely prefer local dev environments for different reasons:

1.) To be able to develop everywhere; this only holds for dev environments on notebooks, but this in fact is our situation.

2.) To have the possibility to test something without impact on a central dev system; so without the danger of disturbing someone.

3.) No problems of parallel deployment processes.

4.) No problems with debugging modes.

The only serious disadvantage I see is the parallel patching etc. of these environments. But in the end, this also can bee seen somewhat as an advantage - everyone learns how to do this (if every developer has to do it by his/her own).

We have a central dev instance deployed, but before this local dev instances. The central instance, where many people are one, is all but stable. My own dev instance is stable as a rock, installed Jan 04, patches regulary. I love it

Hope it helps

Detlev

PS: Please consider rewarding points for helpful answers on SDN. Thanks in advance!

[Former Member · Accepted solution]

Hi,

Could you please specify the hardware you use for local installtion?

We encounter problems with performance on desktop machines after local dev installation.

Is there some performace tunig for this sort of installation?

Best Regards.

Victor.

[detlev_beutner · Accepted solution]

Hi Victor,

> Could you please specify the hardware

> you use for local installtion?

For EP6 NW04 2GB RAM are absolutely recommended; for SP2 on 6.20 1GB did work (nevertheless, 1.5 I would recommend). I'm sitting in front of a 1.6 GHz laptop (Pentium M), that's ok. So memory is more important than speed

> Is there some performace tunig

> for this sort of installation?

One definitely should run the analyzer tools and maybe play with the Java VM memory settings.

Hope it helps

Detlev

PS: I forgot some alternative to patch each decentral dev installation: Use a VMWare, which is centrally patched and can be copied locally after patching (loosing all things configured in the old versions). I personally do not like this for the VMWare overhead (and for loosing my settings, deployments etc.), also I like to know what happens in detail with my installation. But if decentralized patching should be considered as a problem, this can be considered as a solution.

[Former Member · Accepted solution]

Hi,

Thank you Detlev, that's all I can do since this tread is not open by me.

In addition, the other option for cenral patching would be to use NWDI. It has a transport and delivery system in a old ABAP style and this system can be used to patch all the systems in a landscepe centrally

[detlev_beutner · Accepted solution]

Hi Victor,

> Thank you Detlev, that's all I can do

You're welcome

> the other option for cenral patching

> would be to use NWDI

OK, I'm still not very familiar with it, but time will do... But from my restricted understanding, wouldn't you normally have the central (dev/qa/prod) (always online) systems as destination systems instead of decentralized and "wildly distributed" (not always online) systems (and these as supplying, not as "being supplied" systems)?!

Best regards

Detlev

[Former Member · Accepted solution]

Hi,

You can define different tracks. One for you central (always online) systems and others for your destibuted systems. Then, for example, you can connenct you "central system" track to the track of distubuted systems to deliver SPs automatically or manage the "distirbuted" track separately.

Acctulay what I will prefer is to define track for each one of "distributed" systems. Each track will contain only DEV system and then connect all this tracks and this way manage them separatlly from "central" systems.

Victor.

Message was edited by: Victor Shafran