Multicast has been around for a long time (source). If you don’t do any special configurations to Configuration Manager, all packages will be sent as Unicast. This is fine, if you are installing a few computers at a time. But for every computer you add, the network and disk burden will also be higher. If you for example work at a school and need to reinstall 50 computers at a time, it might not be a good idea to use unicast, but instead multicast.


How multicast works:
– Single datagram is sent instead of one per computer, as in unicast
– Sent to a multicast group address, a class D address. Range is: to


Addresses in the range of to ( are reserved for local subnet multicast traffic. Datagrams sent to addresses in this range are not forwarded by IP routers.

Addresses in the range of to are known as globally scoped addresses, which means they can be used for multicasting across an intranet or the Internet. Some addresses within this range are reserved by the Internet Assigned Numbers Authority (IANA) for special purposes.

Addresses in the range of to ( are reserved for administratively scoped addresses. Administratively scoped addresses as defined by RFC 2365 are used to prevent the forwarding of multicast traffic across boundaries configured for the address. Multicast boundaries are described in more detail later in this document.

Multicast will start a “wave” which clients connect to, which means that it doesn’t matter if you are installing 1 or 50 computers, the network and disk usage will be the same.
Might be a good idea to use scheduled multicast:

Scheduled multicast SCCM Configuration Manager 2012 R2


When deciding if you want to use multicast or not, take in consideration:

1. How many computers are you installing at a time
2. How important is timeliness?
3. Would your network support multicast?
4. Network bandwidth, site server hardware…

One cool thing with multicast is that you can monitor the WDS server on the Distribution Point.

Wdsutil /get-allnamespaces /details:Clients

The speed of the transfer will depend on your slowest client. It will become the “Master Client”.


When turning on multicast, make sure that the packages are actually transferred in multicast. I have seen this several times at customers, where they think everything is being transferred in multicast, but when it in fact is distributed in Unicast.

In order to get multicast to work you need to:
1. Allow multicast in your network
1.1 Routers must support the Internet Group Membership Protocol (IGMP), multicast forwarding, and multicast routing protocols.
2. Allow multicast on your Distribution Point
3. Allow multicast on the packages you want to be transferred in multicast. A recomendation is to have it on your OS image. Remember that if you enable multicast on your DP or on your packages, you will need to Update the Distribution Points.
4. Update Distribution Points for the OS image after you have enabled mutlicast for the Distribution Point. If not, you will most likely run in to a hash mismatch error.

Enable Multicast on the Operating System Image Configuration Manager 2012 R2

Multicast doesn’t speed up the OSD process, but it can minimize bandwidth and disk load when deploying a lot of machines at the same time.

You can further conserve bandwidth by configuring when the operating system image is deployed from the distribution point. This consolidates the client requests into a specific time frame, which optimizes the benefits of multicast but might require clients to wait before the operating system package is available. The multicast session starts when it replies to a service location request from a client computer. The distribution point can be configured to require a specified number of requests before it begins to multicast, or it can wait a specified number of minutes after the first request before it begins to multicast. The multicast session ends when the maximum number of client requests is reached. Clients can join a multicast session already in progress. When the multicast is complete, the client will then download missing portions of the package.

When you deploy the task sequence that multicasts the operating system, the content referenced by the task sequence must be downloaded locally from the distribution point to the destination computers. When you deploy your task sequence you can specify that the content is downloaded locally when it is needed by the task sequence or that all the content is downloaded before the task sequence is run.

What it comes down to:
– How many computers you need to deploy at the same time
– The limitations on your hardwareh