1. AD DS: Read-Only Domain Controllers
Updated: January 13, 2011
Applies To: Windows Server 2008
A read-only domain controller (RODC) is a new type of domain controller in the Windows Server® 2008 operating
system. With an RODC, organizations can easily deploy a domain controller in locations where physical security
cannot be guaranteed. An RODC hosts read-only partitions of the Active Directory® Domain Services (AD DS)
database.
Before the release of Windows Server 2008, if users had to authenticate with a domain controller over a wide area
network (WAN), there was no real alternative. In many cases, this was not an efficient solution. Branch offices often
cannot provide the adequate physical security that is required for a writable domain controller. Furthermore, branch
offices often have poor network bandwidth when they are connected to a hub site. This can increase the amount of
time that is required to log on. It can also hamper access to network resources.
Beginning with Windows Server 2008, an organization can deploy an RODC to address these problems. As a result,
users in this situation can receive the following benefits:
Improved security
Faster logon times
More efficient access to resources on the network
For more information about RODCs, see the Read-Only Domain Controller (RODC) Planning and Deployment
Guide (http://go.microsoft.com/fwlink/?LinkID=135993).
What does an RODC do?
Inadequate physical security is the most common reason to consider deploying an RODC. An RODC provides a way
to deploy a domain controller more securely in locations that require fast and reliable authentication services but
cannot ensure physical security for a writable domain controller.
However, your organization may also choose to deploy an RODC for special administrative requirements. For
example, a line-of-business (LOB) application may run successfully only if it is installed on a domain controller. Or,
the domain controller might be the only server in the branch office, and it may have to host server applications.
In such cases, the LOB application owner must often log on to the domain controller interactively or use Terminal
Services to configure and manage the application. This situation creates a security risk that may be unacceptable on
a writable domain controller.
An RODC provides a more secure mechanism for deploying a domain controller in this scenario. You can grant a
nonadministrative domain user the right to log on to an RODC while minimizing the security risk to the
Active Directory forest.
You might also deploy an RODC in other scenarios where local storage of all domain user passwords is a primary
threat, for example, in an extranet or application-facing role.
Who will be interested in this feature?

2. RODC is designed primarily to be deployed in remote or branch office environments. Branch offices typically have
the following characteristics:
Relatively few users
Poor physical security
Relatively poor network bandwidth to a hub site
Little knowledge of information technology (IT)
You should review this section, and the additional supporting documentation about RODC, if you are in any of the
following groups:
IT planners and analysts who are technically evaluating the product
Enterprise IT planners and designers for organizations
Those responsible for IT security
AD DS administrators who deal with small branch offices
Are there any special considerations?
To deploy an RODC, at least one writable domain controller in the domain must be running Windows Server 2008.
In addition, the functional level for the domain and forest must be Windows Server 2003 or higher.
For more information about prerequisites for deploying an RODC, see How should I prepare to deploy this feature?
What new functionality does this feature provide?
RODC addresses some of the problems that are commonly found in branch offices. These locations might not have a
domain controller. Or, they might have a writable domain controller but not the physical security, network
bandwidth, or local expertise to support it. The following RODC functionality mitigates these problems:
Read-only AD DS database
Unidirectional replication
Credential caching
Administrator role separation
Read-only Domain Name System (DNS)
Read-only AD DS database
Except for account passwords, an RODC holds all the Active Directory objects and attributes that a writable domain
controller holds. However, changes cannot be made to the database that is stored on the RODC. Changes must be
made on a writable domain controller and then replicated back to the RODC.

3. Local applications that request Read access to the directory can obtain access. Lightweight Directory Application
Protocol (LDAP) applications that request Write access receive an LDAP referral response. This response directs
them to a writable domain controller, normally in a hub site.
RODC filtered attribute set
Some applications that use AD DS as a data store might have credential-like data (such as passwords, credentials, or
encryption keys) that you do not want to be stored on an RODC in case the RODC is compromised.
For these types of applications, you can dynamically configure a set of attributes in the schema for domain objects
that will not replicate to an RODC. This set of attributes is called the RODC filtered attribute set. Attributes that are
defined in the RODC filtered attribute set are not allowed to replicate to any RODCs in the forest.
A malicious user who compromises an RODC can attempt to configure it in such a way that it tries to replicate
attributes that are defined in the RODC filtered attribute set. If the RODC tries to replicate those attributes from a
domain controller that is running Windows Server 2008, the replication request is denied. However, if the RODC
tries to replicate those attributes from a domain controller that is running Windows Server 2003, the replication
request can succeed.
Therefore, as a security precaution, ensure that forest functional level is Windows Server 2008 if you plan to
configure the RODC filtered attribute set. When the forest functional level is Windows Server 2008, an RODC that
is compromised cannot be exploited in this manner because domain controllers that are running
Windows Server 2003 are not allowed in the forest.
You cannot add system-critical attributes to the RODC filtered attribute set. An attribute is system-critical if it is
required for AD DS; Local Security Authority (LSA); Security Accounts Manager (SAM; and Microsoft-specific
Security Service Provider Interfaces (SSPIs), such as Kerberos; to function properly. A system-critical attribute has
a schemaFlagsEx attribute value equal to 1 (schemaFlagsEx attribute value & 0x1 = TRUE).
The RODC filtered attribute set is configured on the server that holds the schema operations master role. If you try
to add a system-critical attribute to the RODC filtered set while the schema master is running Windows Server 2008,
the server returns an "unwillingToPerform" LDAP error. If you try to add a system-critical attribute to the RODC
filtered attribute set on a Windows Server 2003 schema master, the operation appears to succeed but the attribute is
not actually added. Therefore, it is recommended that the schema master be a Windows Server 2008 domain
controller when you add attributes to RODC filtered attribute set. This ensures that system-critical attributes are not
included in the RODC filtered attribute set.
Unidirectional replication
Because no changes are written directly to the RODC, no changes originate at the RODC. Accordingly, writable
domain controllers that are replication partners do not have to pull changes from the RODC. This means that any
changes or corruption that a malicious user might make at branch locations cannot replicate from the RODC to the
rest of the forest. This also reduces the workload of bridgehead servers in the hub and the effort required to monitor
replication.
RODC unidirectional replication applies to both AD DS and Distributed File System (DFS) Replication of
SYSVOL. The RODC performs normal inbound replication for AD DS and SYSVOL changes.
Note
Any other shares on an RODC that you configure to replicate using DFS Replication would be bidirectional.

4. RODCs also perform automatic load balancing of inbound replication connection objects across a set of bridgehead
servers in a hub site. For more information, see Bridgehead Server Selection
(http://go.microsoft.com/fwlink/?LinkID=208721).
Credential caching
Credential caching is the storage of user or computer credentials. Credentials consist of a small set of approximately
10 passwords that are associated with security principals. By default, an RODC does not store user or computer
credentials. The exceptions are the computer account of the RODC and a special krbtgt account that each RODC
has. You must explicitly allow any other credential caching on an RODC.
The RODC is advertised as the Key Distribution Center (KDC) for the branch office. The RODC uses a different
krbtgt account and password than the KDC on a writable domain controller uses when it signs or encrypts ticket-
granting ticket (TGT) requests.
After an account is successfully authenticated, the RODC attempts to contact a writable domain controller at the hub
site and requests a copy of the appropriate credentials. The writable domain controller recognizes that the request is
coming from an RODC and consults the Password Replication Policy in effect for that RODC.
The Password Replication Policy determines if a user's credentials or a computer's credentials can be replicated from
the writable domain controller to the RODC. If the Password Replication Policy allows it, the writable domain
controller replicates the credentials to the RODC, and the RODC caches them.
After the credentials are cached on the RODC, the RODC can directly service that user's logon requests until the
credentials change. (When a TGT is signed with the krbtgt account of the RODC, the RODC recognizes that it has a
cached copy of the credentials. If another domain controller signs the TGT, the RODC forwards requests to a
writable domain controller.)
By limiting credential caching only to users who have authenticated to the RODC, the potential exposure of
credentials by a compromise of the RODC is also limited. Typically, only a small subset of domain users has
credentials cached on any given RODC. Therefore, in the event that the RODC is stolen, only those credentials that
are cached can potentially be cracked.
Leaving credential caching disabled might further limit exposure, but it results in all authentication requests being
forwarded to a writable domain controller. An administrator can modify the default Password Replication Policy to
allow users' credentials to be cached at the RODC.
Administrator role separation
You can delegate local administrative permissions for an RODC to any domain user without granting that user any
user rights for the domain or other domain controllers. This permits a local branch user to log on to an RODC and
perform maintenance work on the server, such as upgrading a driver. However, the branch user cannot log on to any
other domain controller or perform any other administrative task in the domain. In this way, the branch user can be
delegated the ability to effectively manage the RODC in the branch office without compromising the security of the
rest of the domain.
Read-only DNS
You can install the DNS Server service on an RODC. An RODC is able to replicate all application directory
partitions that DNS uses, including ForestDNSZones and DomainDNSZones. If the DNS server is installed on an
RODC, clients can query it for name resolution as they query any other DNS server.

5. However, the DNS server on an RODC is read-only and therefore does not support client updates directly. For more
information about how DNS client updates are processed by a DNS server on an RODC, see DNS updates for
clients that are located in an RODC site.
What settings have been added or changed?
To support the RODC Password Replication Policy, Windows Server 2008 AD DS includes new attributes. The
Password Replication Policy is the mechanism for determining whether a user's credentials or a computer's
credentials are allowed to replicate from a writable domain controller to an RODC. The Password Replication Policy
is always set on a writable domain controller running Windows Server 2008.
AD DS attributes that are added in the Windows Server 2008 Active Directory schema to support RODCs include
the following:
msDS-Reveal-OnDemandGroup
msDS-NeverRevealGroup
msDS-RevealedList
msDS-AuthenticatedToAccountList
For more information about these attributes, see the RODC Planning and Deployment Guide
(http://go.microsoft.com/fwlink/?LinkID=135993).
How should I prepare to deploy this feature?
The prerequisites for deploying an RODC are as follows:
The RODC must forward authentication requests to a writable domain controller running Windows
Server 2008. The Password Replication Policy is set on this domain controller to determine if credentials
are replicated to the branch location for a forwarded request from the RODC.
The domain functional level must be Windows Server 2003 or higher so that Kerberos constrained
delegation is available. Constrained delegation is used for security calls that must be impersonated under
the context of the caller.
The forest functional level must be Windows Server 2003 or higher so that linked-value replication is
available. This provides a higher level of replication consistency.
You must run adprep /rodcprep once in the forest to update the permissions on all the DNS application
directory partitions in the forest. This way, all RODCs that are also DNS servers can replicate the
permissions successfully.
What's the difference between transferring a FSMO role and
seizing one? Which one should you NOT seize? Why?
Seizing an FSMO can be a destructive process and
should only be attempted if the existing server with the

6. FSMO is no longer available.
If you perform a seizure of the FSMO roles from a DC, you
need to ensure two things:
the current holder is actually dead and offline, and that
the old DC will NEVER return to the network.
If you do an FSMO role Seize and then bring the previous
holder back online, you'll have a problem.
An FSMO role TRANSFER is the graceful movement of
the roles from a live, working DC to another live DC
During the process, the current DC holding the role(s) is
updated, so it becomes aware it is no longer the role holder