Subject: | |
From: | |
Reply To: | |
Date: | Wed, 14 Dec 2005 20:22:20 +0100 |
Content-Type: | text/plain |
Parts/Attachments: |
|
|
Troy Dawson wrote:
> Brett Viren wrote:
>
>> John Rowe <[log in to unmask]> writes:
>>
>>
>>> The options would seem to be:
>>>
>>> * AFS
>>> *GFS
>>
>
> GFS is really geared towards all of the machines having access to the
> disk by fiber chanel, or some similar way, and all being able to write
> to the disk.
> Not for one machine having access to the files and sharing it to all
> the others (which is more of what I picture a NFS replacement being.)
GFS also provides a 2 tier access method by utilizing the GNBD
(Generalized Network Block Device) method of serving up the data storage
area. In this model a number of nodes can be designated as I/O serving
nodes and they serve up a SAN based storage area to the other nodes. The
only stipulation is that all the I/O nodes must see (or be connected to)
the same storage area. So for instance, if you had a 10TB SAN which
allowed 4 hosts to connect to it then you could have up to 4 I/O serving
nodes with FC connection to the SAN. Each of these nodes would then
export a quarter of the file storage space as a network block device.
Each GFS client system (typically the compute nodes) imports the 4
network block devices and constructs a single filesystem using the GFS
pooling mechanism (this is basically a logical volume manager built into
GFS). Thus you get file block striping over the 4 I/O servers with the
obvious gains in I/O throughput. (Actually you have to configure
striping as by default it does not do this).
It is relatively painless to set up once you have tried it a few times.
The manuals are not exactly perfect but it is certainly one way to share
a single file storage area across a number of nodes. Please note that it
is not the most scalable filesystem if you are using a large cluster
(ala more than 400 nodes). In this case Lustre is a better bet as it has
a more scalable architecture for large cluster usage. I think Lustre
locking is more advanced than GFS's as well. Sometimes there are long
delays in GFS when the filesystem is under heavy usage. (like 5 minutes
to do an "ls" command).
Of course both Lustre and GFS can be used freely without support.
Lustre's latest versions have become available to the community since
SC05 when CFS announced simultaneous release for paying customers and
community users.
Hope this helps.
Regards,
John
Fujitsu Systems Europe.
>
> I really think the G, standing for Global, is really misleading.
>
>>> * Lustre
>>
>>
>>
>> One more:
>>
>> http://www.fs.net/sfswww/
>>
>> -Brett.
>
>
>
|
|
|