TMN/Gateway distribution 1

[Glory2G0D]

Hello,

Currently, my environment has one TMR with three managed node/gateways. Of the 500 or so endpoints, over 75% of them are pointing to the TMR Server. Shouldn't this be more evenly distributed or does it not matter?

Two gateways are in different locations than the TMR server, one is in the same building.

Thanks for any and all assistance.

Glory2GOD

 

[Stiddy]

Well, 500 endpoints certaining will not overwhelm a TMR. A few years ago it was said that a TMR can support 10K or so endpoints with no problem. Today this number has grown and it is not heard of for some companies to have upwards of 20K endpoints hanging of a TMR. What is more important is your design and consideration of placing your endpoint/MN relationships accordingly to your network design and applications.

I would say the answer to your question is yes, your endpoints could be more evenly distributed but I would be contridicting my earlier statement. You mentioned that 2 of the MN/GW's are in remote locations. I would say that if you are running applications like Software distribution or Tivoli Inventory then it would make much more since to have the endpoints in the remote locations use the MN/GW's which are closer to them not only geograhically, but with faster response time and better connectivity. Think about it, if you where to distribute a software package to the 500 or so endpoints, would you want your TMR taking the time to push that software package to the 75% of your endpoints. I would obviously appreciate my GW's contributing to the effort as well. It is difficult to suggest to rechart your endpoint/MN configuration without more knowledge of your infustructure, but just from what you have wriiten earlier I would say sure. Just consider your the applications which depend on Tivoli's framework for operability. I hope I have helped you a little.

HTH-Stiddy

 

[Glory2G0D]

Stiddy,

Thanks for your input, that was very helpful.

Since you brought up the subject of software distribution, I'd like to question you on it. I also notice that packages have been created on each MN and distributed from there. This seems confusing to me. If I create one package and send it out to an endpoint on another MN, doesn't it temporarily place the package there anyway? Or should I be creating a different package for each location? Thanks.

Glory2G0D

 

[Stiddy]

Basically you do the following:

1) create a software package for your machine types:1 for windows, 1 for AIX, 1 for UNIX

2) Assign your endpoints to 'DataLess Profile managers'.

3) Distribute your packages to the profile managers with house your endpoints.


Tivoli takes care of distributing the packages. Tivoli will distribute the package to your gateways and your gateways will distribute the packages to your endpoints. Now of coure that is a very vague explaination and a lot of configuration would need to be previously setup, but you get the just of it.

I also want to clarify a statement I made earlier because when I read it after posting I didn't feel I was so clear. When I said "A few years ago it was said that a TMR can support 10K or so endpoints with no problem. Today this number has grown and it is not heard of for some companies to have upwards of 20K endpoints hanging of a TMR", I was not suggesting that a Tivoli TMR Server could handle 10-20K endpoints directly hanging off of it. What I was trying to suggest is that today's Tivoli environments can handle 10-20K endpoints in a sinle Tivoli Management Region in a hole. As a general rule, gateways can support up to 500, give or take a few, endpoints. That really again depends on the types of applications you run. I certainly would not expect a MN/GW during software distribution to deploy say a 250-500meg package to 500 endpoints in the most efficent manner. Yet, it is possible, but it would take a lot of tweaking of those gateways to see how many multiple connections it could handle at a time.

HTH-Stiddy

 

[dcoate]

For maximum efficiency, and to minimize the burden on the TMR server, you want the endpoints to connect to the gateways.

In general, in order to minimize network traffic and response time, you want the endpoints to connect to the nearest gateway (i.e., the gateway in the same location or on the same network segment). (You might deviate from this somewhat if it resulted in too uneven a distribution of endpoints among the gateways.)

You can migrate endpoints in order to correct the situation.

 

[Glory2G0D]

Thanks, dcoate.

Since I've changed the preferred gateway on my endpoints and migrated them to the gateway I want them to point to, I've noticed some of them keep changing back to the TMR. Why would an endpoint's perferred-gateway change from what I set it to be? Thanks.

Glory@G0D

 

[Stiddy]

do some research on your select gateway policy

Stiddy