Defragmentation on SSD drives

Question

We just moved our Sybase ASE15.7 databases from standard "mechanical" HDD drives to SSD drives (IBM SAN).
We suspect that with SSD drives, ASE indexes/tables will require less maintenance against fragmentation.
1) Is reorg/recreate maintenance on tables/indexes still relevant with data on SSD drives?
After moving the db devices to SSD drives, I still see index fragmentation using derived_stat() on sysobjets/sysindexes values.
So I suspect that reorg/recreate indexes are still important but their impact on SSD won't be as important as on standard HDD.
2) Is the concept of fragmentation applicable when a table/index is sent to cache?
In other terms, once the object is loaded in cache, should we ignore the fact that the table was initially fragmented?
Thanks,
Vincent

Answer 1

Regardless of drive storage technology, ASE uses cluster ratio's and the like (among many other factors) to choose optimal query plans.  So these number very much matter.  If comparing the same query plans from mechanical drive to SSD, you obviously will see a performance improvement because the physical io's are faster with SSD (perhaps).   What you _might_ consider is changing the ratio of the two configuration parameters
        cost of a logical io = 2,        cost of a physical io = 25
which by default states that a physio is 12.5 times longer than a cache hit.  For SSD, that ratio might look something more like
         cost of a logical io = 2,        cost of a physical io = 8
maybe more maybe less,  your mileage may vary, but what that does is close the gap between these two costs so that the performance aspect of the storage is more likely taken into account by the optimizer.  BTW - this could be true for SAN based storage (for one example) with heavy caches in front of them as well.
Bottom line is that your "fragmentation" numbers play a role in optimization and query plan determinations.  Performance of the arrived at query plan is influenced by the storage tech (and cpu, network, etc).