10 Replies Latest reply on Apr 16, 2015 7:26 AM by Kenyon

    Help me understand Multicasting


      One of these days maybe I wont have to say this . . . . I'm pretty new to Landesk.  I have deployed new agent on a domain of about 1000 computers. I have also started software distriribution and begun managing patch management.


      Multicast still confises me a bit.  I remeber a few years back and I set up a computer and somone ELSE remoted in and set up IIS to turn that computer into a peer repository (still unfamiliar with all the terms - yes I am devouring the manaual.).  Fast forward two years and a different netowrk with Landesk 9.5.


      Without anything but a core server and many clients that have no additional settings other than current agent installed and running normally,spread across many subnets, how does multicast work. If I deply a msi software package acriss the entire domain, will the first few machines on any subnet that receive the distribution then begin to act as repositories for peers on the same submet?  This happens without any additional configuration?


      Now, what about TARGETED multicasting?  Is this where I "seed" the network with the package on a few machines on every subnet and then, once in place, push out the MSI to everyone?  Is the different between this targeted approach and just setting multicasting on a package without any TARGETING simply that I am controling the placement and WAN to LAN switch over so to speak and it is much more controlled whereas multitasking wthout any such seeding is more of a shotgun blast and let LANDesk sort it out from there?


      Amd what's the differenct between this and what this person was attempting to do a couple years ago.  I suspect this was trying to set up a DISTRIBUTION server.   I understand the concept of distribution servers and that I can have different agents look to different servers. But i need to understand multicast strategy better.



        • 1. Re: Help me understand Multicasting

          T think I figured this out finally.

          • 2. Re: Help me understand Multicasting

            What did you figure out, because its still quite merky to me...

            • 3. Re: Help me understand Multicasting

              You need to specify machines as Multicast Domain Reps (MDRS).  The, when you push something out and use muticasting it deploys to the MDR and then local machines get it from there.  So sorry I never saw this sooner, you probably got your answer already, huh? 


              MDR can be any machine with Landesk Installed.  Naturally, you would want boxes that would be on. The pnly thing soecific about an MDR machine is that it is designated to keep patch/install package in it's cache for x amount of days. If you have som HUGE item getting pushed, you can also pre-deploy to the MDRs caches and then, when you start task, there isnt even that inital wan traffic.


              We user servers across our WAN as MDRS bit they could just as easily be a machine or two on any subnet.  WAN pushes to LAN, Peers pick up form peers or and MDR.  If you have MRDS defined and enable a multicast push, you see in the activity of the task that MDRS are contacted.   From that point on, you know the rest to that segment is happening across the local LAN.

              • 4. Re: Help me understand Multicasting

                How does this differ from setting up local file servers/desktops in remote branches as preferred servers?  Is there a benefit to using Multicast instead?

                • 5. Re: Help me understand Multicasting

                  I have not done too much with preferred servers, So take my response in stride.

                  Preferred servers are basically mirrors of the core. So, they allow for distribution to be parceled on the WAN. Maybe think of a preferred server as being extensions of core.


                  MDRs can be anything on your network that has an agent installed. We have one or two MDRs on each subnet on a  WAN. When you are using MDRs for multicast, you are using their cache to spread out distributions.  Once that rep. machine gets the distribution, others get it from that MDR first, and then each other. They don't fetch from the WAN unless MDR is not available.  That's why you can set two different bandwidth percentages for Multicast. One for how much of the WAN you want to tie up, other for the LAN.


                  If you have all your machines in one location, MDRs become less necessary. But our WAN is across the state, so it makes real good sense. WAN Traffic is limited to Core (or all preferred servers) to the MDRs. Ones it gets to the local network, traffic is limited to the LAN.


                  All this is in ref. to patch deployment.  That's all I really do. I have no doubt that if you want to push out computer images and such you would not that happening across the wan if you could help it. Then, Preferred servers make much more sense. You need IIS installed on all preferred servers. All you need on a MDR (Server or workstation) is 1. Agent installed 2. a machine that you know will be on when distribution goes out.


                  Hope that helps.

                  • 6. Re: Help me understand Multicasting

                    We have 80-something branches all over the country, which is why I started deploying preferred servers on local file servers and desktops a long time ago.  Originally deployed for packages when performing upgrades (Office 2013, which is about 1.5GB), I then expanded it to include images, drivers and patches afterwards.  I just don't understand the difference well enough to know whether using local severs/desktops in branches as preferred servers is better than Multicast or not.  I've had a great deal of success with the current setup, but I am not opposed to changing it if there's a better way.

                    • 7. Re: Help me understand Multicasting
                      Kenyon Expert

                      Multicast does have a similarity to preferred server in the fact that the distribution source files will be distributed at a local level but the primary reason for utilizing TMC is bandwidth savings. Using a Preferred Server allows you to distribute from the local subnet but the clients still individually copy the files locally. That means that a 100MB distribution to 100 machines on a single subnet will take 1GB of data to cross the infrastructure. In a TMC scenario a 100MB package to 100 machines will take 100MB of data to cross the infrastructure. This means that you will minimize the impact on the local network as well as hopefully deploying a large package more quickly. The multicast is a joint conversation between all targeted clients. It would be like the difference of being on a conference call verses calling each person individually.


                      As with any good thing there are some drawbacks. The deployment will only be as fast as the slowest link and the source files have to originate from the host share. Where as with a preferred Server scenario you can stage the source file locally to the target subnet. That's not to say you cant pre-stage the files on a MDR first.

                      • 8. Re: Help me understand Multicasting
                        Kenyon Expert

                        You are not required to designate MDR machines as the multicast tool will chose a MDR automatically if one is not designated. If you do designate one you can at least manage the device and even pre-stage distribution source files.

                        • 9. Re: Help me understand Multicasting
                          technobabble Apprentice

                          Are you still able to designate an MDR for a subnet in LDMS 9.6 as you did in LDMS 9.5? My understanding is that designating an MDR is only for machines with LDMS 9.5 or below agents.

                          • 10. Re: Help me understand Multicasting
                            Kenyon Expert

                            You are correct. 9.6 agents can not be configured as a MDR. 9.6 now has what is called Self-Organizing Multicast.