Replication and High Availability

joeloleson

2 Apr 2007 3:25 PM

Comments 24

The term replication comes up quite frequently in large deployments. It means a number of things to a number of people. Here are some examples:

Geo Replication - Geographical distributed replicated farms (Not supported in the product)
SQL Replication - Online replication of tables (Not supported in the product) Most people will find that logshipping does most of what they are looking for, but it's not online.. Look at Database mirroring for online. (both SQL Log shipping (SQL 2000/2005) and Database Mirroring are supported (SQL 2005))
Multi Master Replication/Two Way Sync - a variation of Geo Replication with data and UI replicating between multiple locations. Watch for data only replications vs. data and UI. Search and Alerts do add complexity to this. (Not in the product)
Data Center Disaster Recovery Replication/Nearline "Read only" Replica - SQL Log shipping for replica used only in the case of failure. (Supported in SQL 2000/2005)
Client Offline Replication - Client has an offline copy. You'll see read only replicas, and read/write where you do the merge or save when you get back.

Fault Tolerance - Software, Hardware, Network, and Datacenter fault tolerance - Usually a request for some solution that provides a redundant solution outside of the datacenter X miles away. Load balancing within the farm or Clustering on the SQL side are examples of this within the data center, and Database Mirroring is an example of this outside the datacenter.

The high availability technologies listed in some of the fault tolerance and high availability I'll list here for quick reference:

NLB - Windows (Network) Load Balancing - hardware, software, and disk failure on the Web Front Ends. Requires multiple servers. (See NLB post for troubleshooting tips)
SQL Clustering - SQL Server 2000 or 2005 Clustering an instance of SQL with common shared disks such as with a SAN or hardware device for shared disks. Provides for hardware or software fault tolerance. Doesn't provide for disk subsystem failure. RAID on the disks can provide fault tolerance for disk failure. Requires multiple servers and shared disk system.
Database Mirroring - SQL 2005 only, supports mirroring the databases for failover (see disaster recovery post for more detail) You can mirror between 2 different datacenters. Bandwidth is a consideration. There is a doc on TechNet for SQL 2005 and for SharePoint tech. Bill Baer from MS IT has deployed this a few times and shares his experience.
SQL Log Shipping - Read only and Read write failovers are possible and supported. SharePoint 2007 Unleashed chapter on high availability has a good overview of these. Single or multiple instances clusters are fine. You backup/restore and setup a logship relationship. The database receiving the transaction logs is essentially offline until you break the relationship and bring it online either read-only or read/write. You then either failback in the read case, or logship back in the read/write case. (Note this does require managing the namespaces and complexities around the config database and server naming.)
Content Publishing - Content Publishing in MOSS 2007 Standard and Enterprise Edition (Not in WSS) can provide for a read only copy of a WCM site such as a read-only publishing site. Although this truly is one way publishing, you could publish to two or more production farms on either side of the world and do either IP ranges per farm or round robin DNS and pull out the affected farm manually or with a heart beat (I wouldn't recommend NLB in this case). Hardware loadbalancing devices would as well load balance the farms until one was down.

Offline Replica Client Solutions part of Microsoft Office 2007.

Groove 2007 - (Client Solution) Groove is one way of getting an offline replica of your workspaces including data in your client with the ability to resync with other peers while not on the corporate network.
Outlook 2007 - Outlook has offline capability for lists with one way replication of connected lists to the client.

There are partners that offer high availability "replication" and offline solutions:

Syntergy Replicator - offers data replication, for lists and libraries including check in and check out, including bi-directional replication. Support for WSS 2.0 & SPS 2003 and WSS 3.0 and MOSS 2007. Including limited functionality for replication between the 2003 and 2007 products lists!
WinApp Technologies Echo - offers the ability to "replicate" changes across site collections, and sites, such as web parts configuration, list settings, permissions, etc... across hundreds to thousands of sites. 2007 support coming soon.
CorasWorks Design Migrator - was designed to help migrate web parts to unghosted pages, but assists with managing customizations and rolling out, deploying, or maintaining consistency across large numbers of sites or farms.
Neverfail - Offers nondisruptive fail over for SharePoint systems. Automatic or manual failover with auto resync with primary and secondary servers. Agnostic to 2003/2007 support.
DoubleTake - Continuous Byte level replication across the WAN. Lower level, considered agnostic to 2003/2007 support.
Colligo Reader and Contributor - Offline client for users including documents, lists, and metadata small client footprint. Reader free for individual use. "Contributor" product for 2 way (saving back to the server) requires licenses (30 day trial). (Written for WSS 3.0 and MOSS 2007). The WAN implications here should be considered. The client performance is quite amazing. The save back happens in the background while the commits are local with drag and drop functionality. Note: Does not support blogs and wikis for offline.
IOra Accelerator - offers offline data in the client including web UI, and delta replication branch office solution (file share or IIS) including two way shipment of package. Note: Be sure to look at the performance to your production farm and the amount of data you need to replicate. Works with WSS 3.0 and MOSS 2007, but there will be an updated version with enhanced APIs calls in the future. They do have customers using their current product with MOSS 2007.

Other:

I'm purposely going to skip WAN accelerators, and cache devices, although they may help in some scenarios. I'm skipping them since they don't offer replication, HA or DR (at least not in SharePoint, WAFS is another story), they are more about consolidation and optimization across the WAN. If network and WANs are your concern, then look at these related posts: