Just spent a couple of days in an aftermath of setting up the first large IMDB installation in a replicated environment. First of all - this is a pure fun. I love technology and IMDB to me, is a pure gem. The ease of implementation of this feature is fantastic. Without exaggeration, all it takes is to get hold of an appropriate hardware. All the rest - is kids playground. You need to get familiarized with a set of configuration commands. You have to understand the basics of relaxed durability and the consequences of it. But apart from that - the benefits by large out-pay the efforts. Combined with the force of replication server, relaxed durability too is not SUCH a but issue. The efforts it take to have an IMDB instance re-synchronized in a replicated environment is so insignificant that you forget about the relaxed durability pretty fast. Is relaxed durability really an issue? Depending on the type of implementation, replicate IMDB setting may give you more freedom to play with than fears to loose your data. The only nuisance I have found with IMDB so far is the start up time for large IMDB instances. Parallel start-up is still in FR lists (pleeeeez do speed up the release of this feature - "a kingdom for a horse"). On a solid storage, it may take up to an hour to start up an ASE with 1 TB IMDB instance (by way of comparison, loading/generating a dump into/from the equally sized DRDB instance will take about two hours). An hour of downtime for some systems is pretty bad news. If that does not stop you - I really recommend this feature. Solves quite a lot of performance problems, opens opportunities that were often eschewed due to lengthy disk-based operations, installation - pretty simple, very little if at all code changes required to start working with it "out-of-the-box." Lovely product. Worth every penny.
↧