In my last article, “Mail Flow in Exchange Server 2007”, I introduced the concept of bifurcation. Bifurcation is a fun word to say, but it’s important in reducing bandwidth consumption and in making disk storage usage more efficient. In fact, support for delayed bifurcation is what leads to Single Instance Storage in an Exchange mailbox database.
In
simple terms, bifurcation means that you take a single message and split
it into multiple copies, preferably only when you are required to make
the split. The process of delaying as long as you can in making that
split is called delayed fan out and it is also commonly referred to as
an Exchange 2007 feature.
All of these are related. Let’s discuss them and see how they apply.
Single Instance Storage (SIS) applies to special messages. The
requirements for a message to use SIS are:
1] Destination recipients are all in the same Active Directory site
2] Destination recipients are all on the same mailbox server
3] Destination recipients are all in the same mailbox database
If all of these three requirements are true, and there are multiple
recipients for a message, then only a single copy of the message’s
attachments is stored into the destination mailbox database.
Yes, I said a single copy of the attachments. In releases of Exchange
prior to Exchange 2007, only a single copy of the entire message was
stored (message body plus message attachment). However, that had a high
disk cost (that is, it required more input-output operations than
storing multiple copies of the message); and one of the major goals of
Exchange 2007 was to reduce disk cost. Many of the tradeoffs involved in
reaching that goal required using more disk space in order to reduce
various I/O operations.
Attachments are accessed less often than message bodies, so they retain
SIS. Also, note that Microsoft recommends against using SIS when
planning for your disk usage requirement in Exchange Server.
Next, fan out describes when bifurcation occurs. In releases of Exchange
prior to Exchange 2007, if the transport engine detected that the
destination recipients were on separate Exchange servers, then
bifurcation happened immediately, even if those recipient Exchange
servers were side-by-side four routing groups away (that is, multiple
copies of a message would cross the network). In Exchange 2007,
bifurcation is delayed as long as possible, which leads to the feature
known as delayed fan out.
If the Exchange servers for two recipients are in the same Active
Directory site, but two separate servers, the bifurcation of the message
does not occur until the message reaches a hub transport server located
within the destination AD site. If two recipients are on the same
Exchange server, but two separate mailbox databases, bifurcation does
not occur until the message reaches the information store service on the
destination mailbox server. And finally, if two recipients are on the
same mailbox database, bifurcation of the message body happens when the
message is stored into the database, bifurcation of any attachments does
not occur.
Delayed fan out can result in significant bandwidth savings if many of
your destination recipients are not in your local Active Directory site.
SIS can reduce the storage used by attachments in your Exchange mailbox
databases.
Bifurcation is a basic process of all mail transport engines. In this
article, you have learned how it is used within Exchange.
Exchange Server 2007 Active Directory Site and Connector Selection
Algorithms
http://msexchangeteam.com/archive/2006/09/14/428920.aspx
Single Instance Storage in Exchange 2007
http://msexchangeteam.com/archive/2008/02/08/448095.aspx