Multicast is a separate and distinct delivery method and uses certain core components. The first one is usually used for staging a large package before the deployment. So the way I approach it, is I look at every distribution job separately. I evaluate the requirements for the distribution task:
- How large is the package.
- Is there more than one package.
- Does it require a reboot.
- How fast does the deployment need to take place.
- How many devices.
- Where are the devices located, and bandwidth restrictions.
- How long does the package take to install.
I find that in my shop we have basically Policy Supported Pushes and Optional policies. Urgency dictates PSP and optional are for on demand tasks. I think it is important to understand why the distribution fails and at what %. Researching failures sometimes is more art than science.
Multicast can't be used with Policies as a multicast involves distributing to target machines simultaneously, whereas a policy could be initiated at any time by the device so no way to know when a multicast session would work. In other words, you can use Multicast only if your delivery method contains a Push method.
David is right as to how to make a decision on what to use. Think what you are trying to send and consider what 'cost' you are willing to pay in relation to how quickly you want it to get there. In this case, 'cost' equals bandwidth. For example, installing MS Office to one machine using run from source is faster than using multicast but puts higher load on your source. Deploying MS Office to 1,000 machines using run from source will probably kill your server and WAN links and therefore be much slower than using Multicast. Swings and roundabouts, so the scenario and the need influance the correct delivery method significantly.
LANDesk Silver ESP
The One-Stop Shop for LANDesk Enhancements
Update - WoW & State Notifier now integrate for even more functionality
Update - State Notifier now detects machine and user Idle states