I’m usually requested, “How can I relocate data to the cloud, and improve the databases, applications, security, governance, and dataops as the migration occurs?” Everyone is in search of a shortcut or a magical software that may migrate and mechanically enhance the state of the information. Sorry, that magic doesn’t but exist.
In the meantime, a nonmagical migration course of gives one of the best odds of success. Before we discover that course of, I’ll point out just a few issues:
First, cloud relocation doesn’t use a waterfall method. Certain duties should be accomplished to maneuver on to the subsequent duties, however not all. These dependences will probably be readily obvious, however be happy to do any of the duties under out of sequence.
Second, to get this proper the primary time, observe the method outlined under with the right mix of expertise. You’ll want material specialists for databases, safety, ops, governance, cloud-specific companies, and so forth. Those individuals are tough to search out proper now.
Finally, it is a basic method. You might want to add or take away some objects. For occasion, for those who’re a well being care firm, that you must cope with extra compliance and governance points across the use, migration, and deployment of information.
With all that mentioned, right here’s the method:
- Assess the “as is” state of the information, together with fashions (object, relational, in reminiscence, particular function, or different), metadata, utility coupling, and necessities (safety, governance, enterprise continuity/catastrophe restoration, and administration). Tagging begins right here.
- Look for alternatives to scale back redundancy and improve effectivity. This might be as impactful as shifting from one mannequin to a different (relational to object) which requires quite a lot of utility refactoring, normalization of all information schemas, defining a single supply of fact, and so forth. You want to think about safety, governance, and information ops as effectively, that are redundant to every part listed right here, simply to be clear.
- Define the “to be” state with the modifications and necessities outlined above. One of the paths I like to recommend is the event of a CDM (widespread metadata mannequin). A CDM, at its essence, gives a single supply of fact for many and typically all the information that exists in an enterprise. It’s made up of many alternative databases that will use totally different database fashions, equivalent to relational and object, and many alternative constructions or schemas. However, it seems to all who use the CDM as a single, unified, summary database…
