[OpenAFS] newbie questions

[Rich Sudlow]

Papp Tam=E1s wrote:
> Jason Edgecombe wrote:
>>
>> please give us more information about what you would like to accomplis=
h.
>=20
>=20
> Well, OK, but I'm sure, AFS is not for me.
>=20
> There are a relatively small cluster, it's aroung ~10TB right now. It=20
> has many type of directories, Some of them are static, they have only a=
=20
> few changes, and also they are smalls one. And there are a Projects=20
> directory with more directorries. These are the big ones and these can =

> grow up to the size 4-5-6 TB. Unfortunately I cannot split them to=20
> different partitions (volumes) because I don't know, which part of it=20
> will grow up.

You might also want to look at the number of files being stored in
a directory as there are some limitations there - with OpenAFS
the limit is roughly 30 - 32 K.

>=20
> So AFS would be great, but at least it should handle bigger volumes tha=
n=20
> 2 TB.

This can be worked around fairly easily.

>=20
> The advantage over Lustre will be the native Windows client (although I=
=20
> still cannot disable crypt to speed up it) and kernel doesn't need any =

> 3rd party patch.


>=20
>=20
> Lustre is fast, but for us it's too fast and not so flexible as we woul=
d=20
> like to see it.

I think you might want to look at your typical data
file sizes too - Lustre is great for large files but
doesn't seem to do well on small files. I've even seen
cases where OpenAFS outperforms Lustre depending on the
file mix.

Rich



>=20
>=20
> Thanks for all,
>=20
> tamas
> _______________________________________________
> OpenAFS-info mailing list
> OpenAFS-info@openafs.org
> https://lists.openafs.org/mailman/listinfo/openafs-info


--=20
Rich Sudlow
University of Notre Dame
Center for Research Computing
128 Information Technology Center
PO Box 539
Notre Dame, IN 46556-0539

(574) 631-7258 office phone
(574) 631-9283 office fax