This week's editorial is a guest post from Rodney Landrum.
We don't see much snow where I live, in Florida, but a short while ago on a weekend trip to Cleveland, Ohio, my wife and a few DBA friends found ourselves gazing in wonder over beautiful Lake Erie, iced over as far as the eye could see. Two identical stone piers broke the vast, frozen expanse of the lake. "Who knew they had pier-to-pier replication at Lake Erie?" I quipped. It was the sort of joke only another DBA would understand or tolerate, and it raised a brief snicker.
As we continued our walk, I found myself reflecting on replication and how it was that I knew so relatively little about a technology that had been native to SQL Server since 6.0, a technology that was a measured skill in the Microsoft certification exams, along with installation, configuration and maintenance. Of course, I know how each breed of replication works and, in my capacity as Solutions Architect, I know when to suggest Bi-directional replication over Peer-to-Peer, especially when cost is a factor for licensing, as the latter in an Enterprise only feature. However, in my long career as a DBA, I've never once had to set up peer to peer replication for SQL Server, or even proposed replication as a solution. It felt like an anomaly.
Partly the reason is that my requirements, as a DBA, have been mainly for a high availability or disaster recovery solutions, with automatic failover, and so we've always opted for a solution such as log shipping or, more recently, Availability Groups. Nevertheless, in my current place of work, I mix with DBAs who have done almost everything there is to do with replication. They have faced issue after issue, fixed failing jobs, re-synchronized subscriptions, passed tracer tokens and generally tablediffed and BCPed the Hekaton out of databases.
And yet, as I listen to their tales of woe, I sense a grudging respect for the technology and a feeling that their skills will have longevity. In 2025, when the skeletal remains of database mirroring, Availability Groups and log shipping lie decaying in the frozen tundra of obsolescence, I'd be willing to bet that replication will still be alive and well and being implemented in some similar form, albeit probably in the Cloud.
What do you think? Have I been lucky to avoid replication? Or unlucky in missing out on opportunities to implement a solid, mature technology in which all DBAs should be expert?