Hello: We will be implementing an Axapta 3.0 installation soon and I was wondering if anyone would be available to discuss system design and architecture? Basically we are looking at a 3 tier environment with 3 instances: Development, Test, and Production. Are most deployments a full physical separation with 3 physical sets of machines or just a logical separation with 3 instances spread across the same set of hardware? I would like to open some discussion with some other system administrators on how they have laid out or looking at layout out their system infrastructure for an Axapta deployment. Thanks Daniel Pumphrey
Hello Daniel, we r running on Axapta 2.5, but i guess it’s similar too 3.0. We have 1 “Big PROD” system DB-server + 3 AOS server and 1 Dev system (SQL + Axapta on one machine) and 1 TEST system (SQL + Axapta on one machine). The dev and test are just normal desktops with 512 Meg RAM. Runs pretty fine. You can install multiple Systems (instances) on 1 Server but i preffered to have dev and test seperated, so that i can realy play with one system without any interference to the others. When you put more then 1 system on one machine … you can’t realy test SQL Service packs, Windows updates and service packs and so on. And the hardware requirements for a full installation and a “handfull of users” r so low… (and cheap). If it is only 1 or 2 persons to develop… maybe think of installing the stuff completely on their machine(s) and do not create a dev-system. Hope it helps Pirmin
Pirmin, Thank you for a quick response. How many users do you have accessing your production system? We will have about 200 users (probably 150 concurrent) in production. Are you using Axapta clustering on your 3 AOS Servers? We plan to have 1 DB Server (Sun Enterprise 6000 running Oracle - existing equipment), 1 Application Server (2 x 2.0Ghz Processors, 2 GB RAM - new equipment), 3 AOS Servers (1 x 2.0Ghz Processors, 2GB RAM - new equipment) - 2 of these will share the user load and the 3rd will be dedicated to running reports during business hours. We will also have 2 additional servers (1 x 2.0 Ghz, 2GB RAM) for Commerce Gateway and OLAP. This comes at recommendation from the VAR that is assisting us with our Axapta implementation. I’m wondering if it is overkill - could the Commerce Gateway and OLAP be consolidated to 1 server to save on hardware costs. How many users do you have in your test environment? We plan to have about 50 users (probably 25-30 concurrent). Our VAR recommended duplicating the production environment as much as possible, but we are thinking of deploying 1 Application Server (1 x 2.0 Ghz, 1GB RAM - new equipment) and 1 AOS Server (1 x 2.0 Ghz, 1GB RAM - new equipment). Our development environment will consist of 1 Application Server (1 X 1.0 Ghz, 512MB RAM - existing equipment) and 1 AOS Server (1 x 1.0 Ghz, 512MB RAM – existing equipment) for 6-8 developers. I’m not too worried about this environment as we have been running on it with a handful of users for demos during implementation and the hardware appears to handle that load fine. Mostly I’m curious if we are way off in the level of hardware we are throwing at our install. New equipment cost will be $40-50K USD. I agree and understand that in a shared physical environment that you will not be able to truly test service packs, windows updates, etc. Since our DB Server will be shared across all environments we will be unable to fully test any Oracle patches as it will affect all database instances when installed. I just wanted to get a sanity check from other System Admins… Thanks! Daniel
hi Daniel, I am really very happy & excited to read about ur Axapta3.0 implementation. we have implemented 2.5 Axapta , and we had intially used only one system for Dev. & test & Production [}:)] & guess what [:o)]we knew we were fools to that, but had to initially to save hardware cost. but later learnt that they should be separate. But we had one system where we used Production & test on it, test was for end user training. One system for Dev. & Testing by our developers. As India is very cost sensitive market we need to go out of the way. I wish u all the good luck & please be in touch so we can share knowledge. my email address is pednekar1407@yahoo.com & Mr Primin please u too be in touch & provide ur email address so i can get in touch with u.[:)] thanks with best regards to all
Hello, Daniel, we have ~ 250 users of whom are normally ~ 50 online. The DB-server is a 4 CPU / 4 GIG RAM SQL-server… Runs pretty good. The 3 AOS ( 1 CPU, 1 GIG RAM) are “Navision Clustered”, in 2.5 stupid counting of uses connected to AOS … and AOS with fewest users gets the next login. In 2.5 we found some issues with (caching) between those AOS. (Data is not always actual on all AOS !)There are also some strange issues with the Databaselog we use (feature of 2.5) And since a couple of days we also have a batch server… We don’t have heavy workload. We use most Projecs/Finance. U wrote 30 to 50 testing users… I can’t realy belelive that you will have them concurrently online if you have a total of 250 Users… This would mean 10 to 20 % would be involed in testing…??? Maybe this is on a couple of stress test days, but most of the implementation time your surely will only have some test users ( and performance won’t be the biggest problem during this time 
). Our internal project team was ~ 8 users and during the end of the implementation phaseis they were most of the time online. Now the r only online when we (IT) give them new functionality to test. 6-8 internal developers? Cool… wish is had ressources like this… A good thing for the dev-guys may be: Put SQL-server on their PC, install Axapta … and form time to time refresh it from PROD. Thats what i did … works great and i can screw up anything i like with real data !! (and you dont have to worry if any of your colleauges screwd up test…) Vaibhav… you can send me mail with clicking on the mail symbol on the top of this message. Hope i helped you Pirmin
Hi Pirmin, i did send u previously one, but the send mail function does not works , cuz i tried sending one to me & never recieved it. Any way i can get in touch with u here, but mail will be more fast. wish u all the best thanks & regards Vaibhav
Here are some comments on some of the stuff you’ve been discussing: 1, 2 or 3 instances: Absolutely at least 2 totally seperate environments for production and test/dev. In our implementations, we usually run 2 seperate environments for test and prod. In larger more complex installations where a lot of development is needed, we have 3 instances running, but usually run the dev and test on the same hardware (AOS and DB). PBE wrote: “we have ~ 250 users of whom are normally ~ 50 online. The DB-server is a 4 CPU / 4 GIG RAM SQL-server… Runs pretty good”. My comments: I’m not surprised it is running pretty good. With average 50 concurrent users, you have some serious iron working for you. But usually it is the disks and raid settings that are most important on DB. PBE wrote: “There are also some strange issues with the Databaselog we use” My comments: The first user that logs in after the AOS is started initilizes the database log. If that user does not have the appropriate rights, the database log does not initialize. Workaround is to make sure that user with admin rights logs on to Axapta immediatly after AOS shutdown/startup Daniel wrote: “1 Application Server (2 x 2.0Ghz Processors, 2 GB RAM - new equipment), 3 AOS Servers (1 x 2.0Ghz Processors, 2GB RAM - new equipment) - 2 of these will share the user load and the 3rd will be dedicated to running reports during business hours. We will also have 2 additional servers (1 x 2.0 Ghz, 2GB RAM) for Commerce Gateway and OLAP.” My comments: a) I guess you mean file server = application server. You really do not need that much power for that server. One CPU is enough for that and you are way ahead in memory. It is a basic fileservice. b) You don’t have to have the batch server running as AOS. You can just configure it as a Fat client (save on licencing another AOS). c) I would seriously consider running only one AOS machine on a bigger iron with more CPU’s. If you are thinking of AOS cluster for “fail over”, you can get that by having the AOS service “stand by” on your file server so if the AOS machine breaks down, you just change the config and you are ready (although with less performace) in couple of minutes after breakdown. Remember to check out the sizers. For IBM: http://www.developer.ibm.com/welcome/myvc.pl?mvcid=overview&packageid=999navaxap&language=en For HP/Compaq: http://www.hpproposal.com/mbs/axapta/index.html Thats it for now Birgir