Status of iSCSI

John Von Essen john at essenz.com
Tue Sep 28 21:46:44 PDT 2004


(Forrest - Had to post to list too as my mail also bounced back)

Not sure if your budget can handle it, but if I where you....

I would stick with NFS over GigE network. But go with a very robust 
hardened NFS soluton, like a NetApp FAS960 Fileserver. Throughput is 
very good, SPOF is non-existent, the disk shelves are fiber channel so 
you can start with 0.5Tb and scale as high (30Tb) as you want on the 
fly (within nfs filesystem mount limits).

A good NetApp system with 1/2 a terabyte will run you around $70,000 
(last time I checked). Can be lower through refurb and/or second hand 
markets.

There are cheaper alternatives but these are no where near the 
performance level of the NetApp units.

-john

On Sep 28, 2004, at 11:47 PM, Forrest Aldrich wrote:

> [Scott, sorry about the bounced mail - it was an old IP block I had, 
> it's fixed now.]
>
> I'm working on a project that requires a scalable mail store, which is 
> poised to support 25k users initially, but scale to 100's of thousands 
> of users.
>
> The budget won't provide for a SAN right now; iSCSI is a little new, 
> but unfortunately it's not supported in FreeBSD at this time.
>
> I've been looking into the storage market areas specifically of Linux 
> and/vs FreeBSD.   A Linux/NAS/iSCSI model and that of a 
> FreeBSD/NAS/NFS model -- not sure we want to do direct storage at this 
> time, if we did, we'd need to plan how that purchased hardware would 
> fit into a larger plan.
>
> NAS has the advantage of being independent; some have their own OS and 
> most have redundancies in place. I'm not sure if  it's possible to 
> dual-attach Linux or FreeBSD boxes to a FC fabric; a while ago, that 
> wasn't possible.
>
> The front-end servers will be split up into scalable groups - ie: some 
> servers doing SpamAssassin, some antivirus, some MTA-in and MTA-out, 
> etc.
>
> The common denomenator, and driving factor of this design, is the 
> backend mail store.   I'd like to explore what (realistic) options 
> FreeBSD may have here - as I dread the thought of Linux-anything in 
> this scenario.
>
> All input/feedback welcomed.
>
>
>
> Forrest
>
>
>
>
>
> There was an implementation done by Lucent last year for 4.x, but it 
> has
> a sticky license and is probably out of date. I and several others see
> iSCSI as something that really needs to get done, but the 3-4 months of
> development time is more than can be done on evenings and weekends. I
> would also want to do it 'right' and implement new infrastructure in 
> CAM
> to accompany it rather than making it monolithic like the Lucent
> implementation.
>
> What kind of project do you need it for, and what kind of resources do
> you have right now?
>
> Scott
>
>
> Forrest Aldrich wrote:
> > I read the April 2004 report (on freebsd.org); presuming that's
> > up-to-date, it may be a while before we have iSCSI support in 
> FreeBSD.
> > I wanted to verify here, etc.
> >
> > I'm involved in a project that will require something of that nature.
> >
> >
> > Thanks,
> > Forrest
> >
>
>
> _______________________________________________
> freebsd-scsi at freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-scsi
> To unsubscribe, send any mail to "freebsd-scsi-unsubscribe at freebsd.org"
>
>
John Von Essen (john at essenz.com)
President, Essenz Consulting (www.essenz.com)
Phone: (800) 248-1736
Fax: (800) 852-3387



More information about the freebsd-scsi mailing list