| Home |
| Articles |
| Backup |
| Books |
| Certification |
| FAQ |
| Products |
| Replication |
| Scripts |
| Seminars |
| Training |
| TSQL |
|
| MSDN Fourms |
| Fort Worth SSUG |
| Oklahoma City SSDG |
|
| Resume |
|
| Champion Valley Pens |
|
|
SQL Server 2000 Preview 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
Replication
- Replication can run over MSMQ
- Data replication – transform with DTS before replication. Can to datatype
mapping/changes, column renaming etc. Especially useful for heterogeneous OLE-DB
subscribers. Can be used to do subscriber specific horizontal/vertical partitions which
takes the load off the log reader from doing this function. Transform can be done either
at the distributor or the subscriber node – uses COM+ to run remotely as necessary.
Transformable subscriptions can not be validated
Queued (disconnected) replication for both transactional and merge. Utilizes either a
local SQL Server table or an MSMQ 2.0 queue. A new replication agent (QueueReader) submits
these to the publisher asynchronously.
Can have an immediate updating subscriber that just falls back to queued if the
connection isn’t there. When all queued transactions are dequeued/committed then it
will start doing immediate updates again.
Improved snapshot concurrency/throughput. Takes a shared-lock briefly and marks a point
in the log. Then grabs all the rows with no locks. Then goes back to the log and gets any
new/updated rows that are relevant. With SQL 7.0 it would take a shared-lock for the
duration of the snapshot which would make the published table read-only for the duration
of the snapshot which could be a long time for large tables.
Add/drop column for schema replication. Changes must be done through replication
procedures instead of a direct alter table.
Can also put adhoc SQL scripts into the distribution database so they are run by all
subscribers. E.g. create new SP’s, tables etc.
Attach and go replication – attach a database and get the subscription
transparently. Makes it much easier to deploy via CD.
Snapshots can now be sent as CAB files for reduced network bandwidth.
FTP snapshots can use a different folder on the FTP server per subscriber.
Can synchronize with any available hub with merge, not just the primary one.
Many new merge conflict resolvers – can still write any custom ones yourself. New
ones include interactive (pop-up), priority value by column, min/max value, first/last
change, additive/average, append text.
Host Integration Server allows for DB/2 and Oracle publishing.
Public interface for any other app/dbms to be a publisher – i.e. stick entries into
the distribution database.
Improved error handling. Ability to continue replication on user-definable errors –
e.g. primary key violation.
Validation on vertical partitioning for transactional replication. Also exposed in the
UI now.
Support for materialized views. Can replicate either the underlying table or the
materialized (summarized) view. Useful for subscribers who may only need summarized
information.
Merge agent is now multi-threaded – improves performance on multi-threaded
machines.
New optimizations for determining data changes pertinent to a merge subscriber.
Most merge operations 50%-300% faster. Some 500% or more.
Identity ranges can be assigned to each merge subscriber to prevent conflicts. When a
subscriber range nearly runs out it can request a new one automatically from the
publisher.
Timestamp columns now supported for merge.
More powerful dynamic filters – can use user defined TSQL functions.
New COM interfaces for heterogeneous merge publishers. MS using this for Host
Integration (DB/2, Oracle), Windows CE, Exchange and Jet. May make it public for other
ISV’s to use.
Can upgrade replication in place
New snapshot explorer
Before and after scripts allow you to write scripts to run before and after a snapshot
is applied
Remote agents can offload the replication agents to another machine
SQL Server 2000 Preview 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
|
