The Naked Scientists

The Naked Scientists Forum

Author Topic: Infinite disc memory (Theory of JxTquery memory servers)  (Read 3611 times)

Offline demadone

  • Sr. Member
  • ****
  • Posts: 112
    • View Profile
In a computer hard disc, the information is stored as zeros and ones. The ons and offs. Now if memory is the permutations and combination of these zeros and ones and not matter, how is it that memory volume has to increase? In other words, why not increase the possible permutations and not the actual number of zeros and ones. In fact, there are basically 2 numbers (if I'm not mistaken) that form memory.

My theory therefore is that we should be able to do more with the numbers. Much like the difference between a static and a dynamic web page, if we had a server that generates the possible combinations and a host that could query for combinations, then memory could be more efficient. We could even create a server that queries a server that queries yet another server. In effect making small memory space almost infinite.

Doesn't your flash disc not feel like a waste of space now?
« Last Edit: 21/03/2009 04:06:03 by demadone »


 

Offline LeeE

  • Neilep Level Member
  • ******
  • Posts: 3382
    • View Profile
    • Spatial
Infinite disc memory (Theory of JxTquery memory servers)
« Reply #1 on: 21/03/2009 18:22:13 »
Trinary based hardware, as opposed to binary based hardware, might be possible by using +1, 0 & -1 states instead of just +1 & 0 states, but for more states than that you'd need to start using the state levels as well as the states.  This would mean that the hardware would have to be able to not only differentiate between +1, 0 & -1 states, but also the analogue level of each state.  Considering the complexity of digital hardware - typically billions of transistors - it would be very difficult to make this work accurately in practice.
 

Offline demadone

  • Sr. Member
  • ****
  • Posts: 112
    • View Profile
Infinite disc memory (Theory of JxTquery memory servers)
« Reply #2 on: 21/03/2009 23:06:20 »
I'm of the impression that the technology you are talking about Lee is slightly different. A simple way of saying it is this. the zero-one combinations can be defined in pairs of 3. If we called 100 '2', 101 '3', 111 '4' and so on, a complex series such as 100101111 could be reduced to say 234 and if we more pairs than 3 we could use these combinations queries to retrieve the combinations they represent from the query server.
 

lyner

  • Guest
Infinite disc memory (Theory of JxTquery memory servers)
« Reply #3 on: 22/03/2009 20:54:13 »
There's nothing magic about binary storage or transmission of information. It happens to be very convenient to make machines which use binary processing.
On the subject of total capacity of memory, the standard memory system used for storage is designed on the basis that you need to recover the information quickly and conveniently. This means that it is often very wastefully stored.
However, when you consider how little storage space is needed for certain types of data (spelling dictionaries, for instance, use a fraction of the memory that you would need to store every word, perfectly spelled) you can get some idea of how much more efficient it is possible to get. Then there are the familiar systems for lossless data reduction used for music and pictures - but they are far less impressive in what they can achieve.
 

lyner

  • Guest
Infinite disc memory (Theory of JxTquery memory servers)
« Reply #4 on: 25/03/2009 00:30:15 »
One of the main problems about storing information is actually retrieving it. Our brains make a very good job of it but simple computer file systems make a lousy job because they need a filing system and need to be told which file to search in -or they would take for ever. They are more accurate and have all sorts of advantages but the brain has what we call 'content addressable memory'. This means we ask for the information by getting as close as possible to defining what we want rather than where it is stored.
As I understand it, we store things as a set of associations, which are formed by lots of interconnections in our brains. When we want to recall, say, someone's name, we feed in as many of the interconnections as we can - face, account number, claim to fame and, if we're lucky, those interconnections will cause what we want to come flooding out - in the form that it went in originally.
This is highly efficient both in time and capacity - when it works.
Unfortunately, I find that my memory for names is terrible - and getting worse. I can get a serious block about someone's name which I can only solve by inventing a whole new set of connections, artificially, which bypass the block and 'allow' me to recall what I wanted. It's a strategy which works quite often but it's sodding annoying to have to use it.
Same thing happens when I get to the top of the stairs and wonder why I'm there. I need to re run the day up till that time and sometimes I 'get it'.

Our memories have got a lot worse these days, in some respects. We no longer need to (or can) remember long passages of verse or journeys, since writing and maps have been available.
Some people still remember the entire Koran - even in a foreign language - well impressive, I think.
I wonder whether we are actually making alternative use of this capacity or is is just being wasted through under use.
 

Offline LeeE

  • Neilep Level Member
  • ******
  • Posts: 3382
    • View Profile
    • Spatial
Infinite disc memory (Theory of JxTquery memory servers)
« Reply #5 on: 25/03/2009 14:18:32 »
DBMSs (Data Base Management Systems) provide content addressable storage, but they generally sit upon a conventional filesystem, so there's an overhead as queries still have to go through the filesystem layer.  The overhead isn't as great as you'd imagine because the DB engine does all the tricky stuff and only uses the filesystem layer for low-level read/writes, but it's still there.

Implementing the DB container directly on the hardware, without the OS filesystem layer could speed things up quite a bit.  For example, compare linux swap handling with normal filesystem access (this doesn't apply so much with windows as the swap space sits on top of the filesystem whereas with linux the swap space is in it's own partition and isn't accessed via the standard filesystem layer).

Even then though, you'd need DB datatypes and functions to handle all the possible different types of data, which isn't really feasible, so a lot of work would still need to be done by the OS/Application to handle binary objects for which there was no specific datatype.

I believe that this was the idea behind the WinFS filesystem, which was originally planned to be part of what eventually became Windows Vista when it was still known as 'Longhorn'.
 

Offline Alan McDougall

  • Neilep Level Member
  • ******
  • Posts: 1285
  • Thanked: 14 times
    • View Profile
Infinite disc memory (Theory of JxTquery memory servers)
« Reply #6 on: 01/04/2009 16:04:48 »
As far as I know the quantum computer, if it is ever realised , with solve any storage and data filing problem. The problem the present and near future divices are heat and the limitation of the spped of light

Quantum , nonlocalty and entanglement will overcome that difficulty. In a quantum computer it would not matter how or where or even when you stored the information.

Access should be almost instantaneous
 

lyner

  • Guest
Infinite disc memory (Theory of JxTquery memory servers)
« Reply #7 on: 01/04/2009 23:09:26 »
I'm not sure that the quantum processor speed is relevant to the memory access problem. Why should it be?
 

lyner

  • Guest
Infinite disc memory (Theory of JxTquery memory servers)
« Reply #8 on: 04/04/2009 00:37:15 »
Lee e
Quote
DBMSs (Data Base Management Systems) provide content addressable storage,
Do you really mean that the storage medium is not addressed conventionally or do mean there is some software control which gives a 'virtual' content addressing?
I think that true CAM needs to have a neural network kind of structure in which the data is both impressed on the network and accessed from it in the same way. It goes along with almost limitless parallel processing, too.
 

Offline LeeE

  • Neilep Level Member
  • ******
  • Posts: 3382
    • View Profile
    • Spatial
Infinite disc memory (Theory of JxTquery memory servers)
« Reply #9 on: 04/04/2009 19:55:10 »
Lee e
Quote
DBMSs (Data Base Management Systems) provide content addressable storage,
Do you really mean that the storage medium is not addressed conventionally or do mean there is some software control which gives a 'virtual' content addressing?
I think that true CAM needs to have a neural network kind of structure in which the data is both impressed on the network and accessed from it in the same way. It goes along with almost limitless parallel processing, too.

I think it depends what you mean by the storage medium.

If you're thinking in terms of the physical media, then I don't think 'true' CAM (Content Addressable Media) is possible.  The physical media itself doesn't offer data to the system because it's completely passive.  Instead, the filesystem keeps track of where files are located on the media; when a file is requested the location on the media is read, the file is extracted from the block of data held at that location by the filesystem, and then passed on to the requesting process/application.

If you're thinking in terms of what the filesystem can present though, then Content Addressable Filestore/Filesystems are possible because there is an active component in the chain.  It is this respect that DBMS's act like CAM; When you request data from a DB all you do is identify the data that you want and only that data is returned.

I guess it comes down to how data is read from the physical media i.e. by physical location on the media.

However, if you could come up with a scheme where the data was also the location then you could say that the media was content addressable.  For example, if you want to store a series of integers, you could just use the blocks as the data instead of storing the data in the block; the contents of the blocks would be irrelevant, so block 1 = data value 1.  This wouldn't be very useful though because you have already had to specify '1' to retrieve '1'.

This actually highlights another issue with CAM; to address the contents, you have to specify the contents for the contents to be identified.  Of course, you can get partial matches, but while it's easy to formulate a partial match on alphanumeric data it's a lot harder to formulate a partial match for a visual image or a piece of music, especially if you want to include different resolutions and colour balances for images, and tempos, instruments and performances with music.

So not only is it difficult to formulate a query just on the data (without relying upon meta-tags), it's even harder to figure out a way for the data to become it's location on the media.

As long as you've got an active layer somewhere between the physical media and the requesting process though, you can present a CAM abstract layer.  This is where your idea about neural networks might come in handy, especially for for those partial media data matches.
 

lyner

  • Guest
Infinite disc memory (Theory of JxTquery memory servers)
« Reply #10 on: 05/04/2009 20:09:22 »
Yes. You need to put stuff in the right place so that you can retrieve it. It's got to be like my shed.
 

The Naked Scientists Forum

Infinite disc memory (Theory of JxTquery memory servers)
« Reply #10 on: 05/04/2009 20:09:22 »

 

SMF 2.0.10 | SMF © 2015, Simple Machines
SMFAds for Free Forums