[Pvfs2-developers] bstream bucket spreading

[Rob Ross]

thanks. -- rob

On Jan 21, 2008, at 3:00 PM, Pete Wyckoff wrote:

> rross at mcs.anl.gov wrote on Mon, 21 Jan 2008 14:24 -0600:
>> New versions will still be able to find bstream files for files  
>> created by
>> old servers? -- Rob
>
> Bstreams didn't move by this patch.  Else we would have needed a
> major version bump.
>
> 		-- Pete
>
>> On Jan 18, 2008, at 12:42 PM, Pete Wyckoff wrote:
>>
>>> slang at mcs.anl.gov wrote on Fri, 18 Jan 2008 11:51 -0600:
>>>> On Jan 17, 2008, at 3:26 PM, Pete Wyckoff wrote:
>>>>
>>>>> I just noticed this bit in trove-dbpf/dbpf.h.  The code carefully
>>>>> builds ((<collid> << 24) | <handleid>) then does a modulo against
>>>>> a power of 2, effectively stripping off all but the low bits of
>>>>> the handleid.
>>>>>
>>>>> The collid is not used in the hash at all.  Just in case anybody
>>>>> thought this was important.  I've ever only used one collection  
>>>>> per
>>>>> server.
>>>>>
>>>>
>>>> KEYVAL_GET_BUCKET is left over from the days when we did a keyval  
>>>> DB per
>>>> file.  With just one keyval DB now, we don't use that macro --  
>>>> there's no
>>>> where in the code it gets called -- so you could safely remove it  
>>>> if you
>>>> want.
>>>>
>>>> As for BSTREAM_GET_BUCKET, that's funny that the coll-id isn't  
>>>> used in
>>>> the
>>>> hash, but it turns out not to matter.  The buckets end up as  
>>>> directories
>>>> that allow us to keep the bstream directories smaller, but here  
>>>> too, all
>>>> those directories are within a directory for the entire  
>>>> collection.  Go
>>>> ahead and pull out the coll_id bits of the hash if you want.
>>>
>>> I killed the extranneous stuff.  Maybe it'll save somebody else some
>>> head scratching in the future.
>