PDA

View Full Version : Speed of Various Operations


David G. Kanter
25 June 2005, 01:45 AM
. . .Another reason this solution is less-than-ideal is that switching indexes (Soundex->Last name, and back) takes *far* too long. Some thought needs to be given in future version towards optimizing algorithms for large files - anything involving indexing, including match & merge, takes much longer than is necessary!I appreciate the computer's processor speed and RAM, along with the size of the Family File, would have a direct impact on the time for those operations, but I'll offer the following times for my Ti PowerBook—with a 1 GHz processor and 1 GB of RAM—on my Family File with 22,482 persons—under OS 10.4.1:

• A re-sort of the Index from any of the 10 options to another option takes less than 1 second.
• A Find Relatives takes about 7 seconds.
• A rebuild of the whole Index takes about 14 seconds. (This is an operation that happens very infrequently for me—usually only after I've chosen to save a Compact Copy and then open that copy.)
(I'm not addressing "match & merge" as if you're using it for a large number of persons against a large number of persons it is, understandably, a very intensive operation; but for me, that's an infrequently used feature—and then it's usually a "few against many" operation.)

I'm not in a position to know what trade-offs are appropriate within Reunion that might further accelerate the various operations for whatever is Dennis' computer, but for the overwhelming majority of the operations I use, Reunion is waiting for me, not the other way around—and that's on a computer which has become relatively slow compared to every Mac that is being sold now. (Who would have believed that, say, 5 years ago?)

Dennis J. Cunniff
25 June 2005, 02:16 AM
I’m not in a position to know what trade-offs are appropriate within Reunion that might further accelerate the various operations for whatever is Dennis’ computer, but for the overwhelming majority of the operations I use, Reunion is waiting for me, not the other way around—and that’s on a computer which has become relatively slow compared to every Mac that is being sold now. (Who would have believed that, say, 5 years ago?)

Of the variables affecting speed, I suspect the most important are size of file and speed of disc. A re-build of the Index in a file of about 200,000 takes about an hour, regardless of machine (mine is an iMac G5), and a "match and merge" of exactly two marked individuals takes minutes - far longer than it should.

I suspect Reunion is built on the probably correct assumption that most files will be small, and is not coded to use optimal algorithms that would make no appreciable difference in small files (but which would relieve considerable tedium on large files). These operations also go much more quickly if one can arrange for the file to be on a RAM disk...which I haven't done under the new OS. But since a 100MG file should easily fit in 2GB of memory, an option to keep the file memory resident might be [1] practical and [2] speed things up.

David G. Kanter
25 June 2005, 02:40 PM
Of the variables affecting speed, I suspect the most important are size of file and speed of disc.My disk is a 60GB 4200-rpm Ultra ATA/66 hard drive and my Family File is 10.8 MB in size—with 22,482 persons.

. . .a "match and merge" of exactly two marked individuals takes minutes - far longer than it should.FWIW, the same operation in my case—using Criteria of just "Soundex, Similar First name"—is about 1 second. A 2-marked compared to Everybody (i.e., 22,482 persons), same criteria, takes about 3 seconds. (If I add Birth Date & Death Date within 3 years to the criteria, it's about 4 seconds for that "to-Everybody" operation.)

These operations also go much more quickly if one can arrange for the file to be on a RAM diskI haven't set up a RAM disk, but even considering the factor of 9 between the number of persons in our files, I seem to be getting far better speed than you are for those simple match-and-merge operations and, I gather, also the Index-re-sort and Index-rebuild operations.

So while I'll defer on the problem for very-large files such as yours—and I thought my file with 20,000+ persons was pretty big :)—I suspect the vast majority of Reunion users should not experience any really significant delays within Reunion.

But, then, that's just my expectation. Perhaps a few others with different computers and file sizes will chime in with what delays, if any, they are experiencing on the operations we've been discussing—providing us, also, with whatever are their computer's specifications, operating system, and the size of their Family File. That might put a broader perspective on to what extent others are running into significant delays while Reunion is doing these operations. Note: I said a few others so as to comply with the ReunionTalk Guidelines in that "me-too-only" posts aren't appropriate.