Bill Benson
bensonforums at gmail.com
Thu Mar 13 14:16:17 CDT 2014
As for what I would do, I would teleport the data into another dimension, FTL. Then bring it back later: it will be younger than it was, and it might come back carrying my missing socks from the dryer. But that's me. You asked. Just think it is more intuitive to move data if I need the room, and not before. Of course my plan has no practical outlet and doesn't even make logical sense. Otherwise I am writing all the time and STILL have the same problem, like trying to make all the children above average. Or if I am going make a lot of writes, I would expect, if the controller knows the write count of every bit, it would know to find the bits that have been written to the least and use them next, for official (real) writes. When it runs out of unused bits, it would grab the next least used bits. I see no value in the extra writing. It just ensures every bit suffers. I think what is being missed on my part is what DELETING DATA does. I ASSUME (ahem... caveats) that deleting means not wiping, but masking data to make it not readable. So the controller actually CAN write to all bits in a long stream of bita, not re-using deleted areas until it runs out of free area. Then I would expect the controller to do this marshaling thing with every official write. As opposed to pre-emptively. Because, all things being equal, who is to say how much space I am going to write to, until I go ahead and do it. I can't envision how pre - emptive writing moving) blocks customers down the wear and tear, unless it is to try to be AFTER at the time bona fide writes are being done. It's the difference between moving money around within the Fed just to move money around, and doing it because there is real borrowing going on. Most of the time my hdd is doing absolutely nothing (no writes) but at some point it might do a lot. Do I want this controller pushing dirt from one end of the room to another when I haven't asked for a place to lie down, just to keep the room evenly filthy? But again, that's me and my inability to visualize this topic. No irony. There is no getting through to me John. Though your class module thing did finally sink in, sorta. On Mar 13, 2014 2:37 PM, "John W Colby" <jwcolby at gmail.com> wrote: > Well Bill think of it this way. > > 1) Each area can only be written to X times (about 3000 in fact) > 2) I have a spreadsheet which has only been written to once (when it was > stored initially) That area can be written to 2999 more times. > 3) I have a database area that has been written to 2500 times. It can > still hold data but is approaching the end of it's useful life. > > What would YOU do? > > If you don't move the database area, then in 500 more writes it will fail > to correctly store and you will lose data. > > Ohhh.. wait a minute, I happen to have a block over here which was written > a year ago, exactly one time. Hmmm... if I were to swap those two blocks > then the database can continue to be written to. The spreadsheet is still > safe, and it is NEVER written to. The database is still safe and can > continue to be written to. > > SSDs have a fundamental "problem" that any given NAND block can only be > written to X times. You have to do SOMETHING to "level" the wear evenly > across the disk or you will very quickly completely destroy some areas and > leave others entirely unused. > > When I say destroy I mean unable to hold anything at all. Notice that > when I swapped the spreadsheet and the db, the spreadsheet was perfectly > safe in the new location. So we prevented "destroying" that storage > location by selecting something that wouldn't be written again and storing > that in there. > > So you are correct you ARE wearing down ALL of your locations - EVENLY. > It is in fact possible to completely destroy an SSD disk. If there are no > "static" files, files which are not updated, if EVERY file is written over > and over and over (3000 times) then EVERY location will be written until it > simply is unable to hold data. > > The articles I posted links to explain all this stuff. > > John W. Colby > > Reality is what refuses to go away > when you do not believe in it > > On 3/13/2014 2:04 PM, Bill Benson wrote: > >> To me this sounds STUPID. Not stupidly Described, but stupidly contrived. >> Therefore since people are making and spending scored of Billions on it, >> CLEARLY I DON'T UNDERSTAND IT. >> >> It sounds like this: I know disk areas are getting worn down so I will >> just wear down ALL my areas. LOL+LOL to the Google power. >> >> >> > > --- > This email is free from viruses and malware because avast! Antivirus > protection is active. > http://www.avast.com > > -- > AccessD mailing list > AccessD at databaseadvisors.com > http://databaseadvisors.com/mailman/listinfo/accessd > Website: http://www.databaseadvisors.com >