[Info-vax] Databases versus RMS

[Johnny Billquist]

On 2012-04-18 22:36, Jan-Erik Soderholm wrote:
> Johnny Billquist wrote 2012-04-18 22:17:
>> On 2012-04-18 20:53, Jan-Erik Soderholm wrote:
>>> Johnny Billquist wrote 2012-04-18 19:53:
>>>> On 2012-04-18 15:43, Bob Koehler wrote:
>>>>> In article<jmm94e$atj$1 at Iltempo.Update.UU.SE>, Johnny
>>>>> Billquist<bqt at softjar.se> writes:
>>>>>>
>>>>>> Even beyond any language issue, or RMS details, the OS can cache and
>>>>>> defer actual writes to the disk without you ever knowing about it.
>>>>>> Not
>>>>>> to mention that disks also cache things...
>>>>>>
>>>>>
>>>>> OS cache? Are you dreaming UNIX? VMS has never cahced any data I
>>>>> cared about without my explicitly telling it to do so.
>>>>
>>>> Really? I find it hard to believe. RSX have done caching of things
>>>> from the
>>>> disk at that low level since day one. F11ACP do some caching for the
>>>> system, keeping the most recently referenced directories around in a
>>>> cache
>>>> in the ACP.
>>>>
>>>> Newer versions of RSX also do caching of individual blocks fully
>>>> transparent at the OS level. It's selectable if you want writes to be
>>>> write-back or write-through.
>>>>
>>>> That VMS would have none of this sounds unlikely, as well as very
>>>> bad for
>>>> performance.
>>>>
>>>>> Disks that cache should use a non-volatile memory to do so, or
>>>>> contain internal power storage sufficient to flush the cache.
>>>>
>>>> Right. But that is still a cache that can cause surprising results,
>>>> although I freely admit that the caches in the disks are normally very
>>>> reliable.
>>>>
>>>> Johnny
>>>>
>>>
>>> You are mixing read and write cache. Make sure you are talking
>>> about the same thing before arguing... :-)
>>
>> Huh? They are not separate! If they were, you'd get some really bad
>> results.
>>
>> But even disregarding that, I do not understand what you were commenting.
>> Please explain.
>>
>
> I read a lot of discussing about "caching" but do not understand in
> every case if it is read or write caching that is ment.

Ok...

> As an example (the part that triggered my comment) :
>
>  >>>> OS cache? Are you dreaming UNIX? VMS has never cahced any data I
>  >>>> cared about without my explicitly telling it to do so.
>  >>>
>
> This comment is obvisouly about *write* caching.
> (Of course VMS caches *reads* of different things without
> explicitly telling it to do to).

To me, it is not obvious that it's only about write, but anyway I found 
the distinction irrelevant,

>  >>> Really? I find it hard to believe. RSX have done caching of things
>  >>> from the disk at that low level since day one. F11ACP do some
>  >>> caching for the system, keeping the most recently referenced
>  >>> directories around in a cache in the ACP.
>
> And this (as a reply to the statement above) is abviously about
> *read* caching.

No. It's both for read and write.
If you read, it's obvious that it uses the cache. If you want to update 
the data, of course the cached data in the ACP also are updated, and not 
just thrown away. The cache cannot ignore and pretend writes don't exist.

But my comment was also a general response to the claim that VMS don't 
do any caching, since I found that extremely hard to believe.

> Since they comment on two different things (read vs. write cache)
> the whole discussion becomes a bit weird... :-)

I don't know how you look at it. There is no such thing as read vs. 
write cache. You might be thinking of write-through as opposed to 
write-back cache. The latter might be what you think of as a write cache...?

	Johnny

-- 
Johnny Billquist                  || "I'm on a bus
                                   ||  on a psychedelic trip
email: bqt at softjar.se             ||  Reading murder books
pdp is alive!                     ||  tryin' to stay hip" - B. Idol