[Glass] List and Fast list instances

[bruno buzzi brassesco]

Hi Martin/Norm,

Thank you !!!

Excellent information :)

regards,
bruno

On 10/12/2021 18:17, Norm Green via Glass wrote:
> Hi Bruno,
>
> The object table is first scanned to generate the list of data pages. 
> Then those data pages are scanned in order as Martin says. Each data 
> page is only processed once. Dead objects (found by a previous MFC) 
> are excluded from the results, however disconnected objects (not yet 
> been collected by an MFC) may appear in the result set.
>
> Both of these phases are multi-threaded. More threads == more 
> aggressive where the scans will use more disk IO and CPU to complete 
> in less time. "Fast" list instances uses many threads to complete the 
> scans as quickly as possible.
>
> Norm Green
>
>
>
> On 12/10/2021 12:47 PM, Martin McClure via Glass wrote:
>> On 12/10/21 11:50, bruno buzzi brassesco wrote:
>>> Martin,
>>>
>>> Thank very much  for the answer.
>> You're welcome!
>>>
>>> So fastListInstances and a like iterate over the Object Table in 
>>> order to build the result set ?
>>
>> Now we're getting into some of the areas where I don't recall all of 
>> the details. :-)
>>
>> IIRC the repository scan is done in page order to get good I/O 
>> performance, rather than in object table order.
>> But yes, the object table is used, and the dead not reclaimed set, 
>> and other data structures, so that just "an object was found on a 
>> page" is not enough for it to be included in the result set. It must 
>> be in the OT (and therefore not shadowed) and not dead. And there are 
>> a number of optimizations whose details I don't know.
>>
>> Regards,
>> -Martin
>> _______________________________________________
>> Glass mailing list
>> Glass at lists.gemtalksystems.com
>> https://lists.gemtalksystems.com/mailman/listinfo/glass
>
> _______________________________________________
> Glass mailing list
> Glass at lists.gemtalksystems.com
> https://lists.gemtalksystems.com/mailman/listinfo/glass