Replication Topology 103 - End to End

Tags :

Prerequisites are Replication Topology 101 (the basics) and 102 (peer to peer) before you hit this one.

From the beginning, I gave a scenario of how it looks

Basically data starts on one end, passes through multiple servers through replication and then comes right back.  Timing becomes and issue to make sure that data can make it all the way down and back before the next baton is passed.  Think of it as runners that pass the baton, and if one runner takes off too early, who knows where the baton is.

So I hopefully already broke you away from the idea of a meshed environment in class 101 due to the sheer number of connection records that are possible and messy management.

End to end offers it's own set of benefits and pitfalls, of course.  If you can imagine your science class from way back in elementary school.....where they gave you a stack of batteries and a bunch of light bulbs.  You were then told to light them all up.  The first thought is batteries, then wire to next bulb, then wire to next bulb and so on until they were all connected.  Well if one went out in that serial connection idea, then everyone behind them went out.  So the teacher taught you about parallel connectivity to get around it.  Which end to end does not do in the true form.  Any variation moves it towards circular or even tiered architecture (with a bizarre slope).

The benefit is that data passes along in a cycle, reducing replication conflicts.  Save conflicts are entirely different as people across the string could be editing the exact document on every server.  Timing, as I mentioned, also becomes and issue since it could run any amount of time to get the data back and forth.  If a server or network is down, the others will replicate as scheduled, yet that missing link in the middle brings the idea of timeliness to a screeching halt on each end.

The end result is a long line of servers, spread in the same room or geographically, that have a start and end point.  Sure, you can argue that every topology has a start and end point.  But with the proper hub cluster setup, only an individual spoke failure would affect any users.  In end to end design, there are too many holes along the way.