[Info-vax] The best VMS features, was: Re: openvms renaming file

[Arne Vajhøj]

On 5/29/2018 7:20 PM, Johnny Billquist wrote:
> On 2018-05-29 01:44, Arne Vajhøj wrote:
>> On 5/28/2018 4:08 PM, Johnny Billquist wrote:
>>> If you ever take time to learn how Google's bigtable works, you'll 
>>> see that it is almost identical to how RMS works. Quite funny, actually.
>>
>> I would have considered RMS index-sequential files more like
>> a key value store with secondary indexes than a wide column store.
>>
>> Or?
> 
> bigtable is simply just a storage where you have rows with multiple 
> columns that you access by name instead of by byte positions. The 
> additional benefit being that it more efficiently handles if you have 
> non-existant columns in your rows, so you don't waste so much space.

I am in no way an expert in NoSQL technology.

But my understanding is that:
* a wide column store is an (usually distributed) on disk structure
   implementing what is conceptually a 3D sparse matrix with row, column
   and timestamp is dimensions.
* a key value store with secondary indexes is an on disk structure
   implementing what is conceptually an associative array - and
   secondary indexes provide additional associations pointing to
   same data

And even though they share some similarities, then I do think
RMS index-sequential files match the second much better than
the first.

> You can certainly look at bigtable as a key value kind of storage. There 
> is nothing more clever about it.

Maybe not.

But in NoSQL circles they do seem to make the distinction above.

Arne