Karl R. Balsmeier wrote:
>> The ability of iSCSI to support high throughput depends on:
>>
>> 1) How the back-end storage being served up by iSCSI is configured
>> 2) How the network interconnects between the iSCSI targets and
>> initiators are configured
>> 3) How well the FTP software does at reading the data from disk and
>> pumping it out the network
>>
>> 1Gbps ethernet can handle up to 115MB/s per interface. Using MPIO
>> round-robin over several interfaces you can continue to add throughput
>> if the application can scale well across these multiple paths.
>>
>> -Ross
>>
>>
> hey this part is fascinating, -so how would one practically deploy this,
> -say 4 GB NICs and some supported hardware? for traffic 100 - 200 megs
> daily perhaps this is too much?
There's no such thing as 'too fast', but do you really need to complete
you daily transfer in less than a second? On the practical side the
underlying disks aren't going to be that fast anyway.
> On the storage side, not sure if MPIO
> will auto-detect the device, but maybe it'll see it. -wonder if Vsftpd
> would play well with all of this.
If you put a filesystem on an iscsi target and mount it, vsftpd won't
know/care about the actual device type.
--
Les Mikesell
lesmikesell at gmail.com