----- Original Message -----
From: “Pashupati Kumar”
To: “File Systems Developers”
Sent: Friday, February 22, 2002 8:04 PM
Subject: [ntfsd] Re: How to share storage device on logical block acce ss
level?
> for true cluster file system, d locks would be needed.
Not only are d-locks not needed, they’re a crock - both as currently
implemented and in just about any envisionable implementation. The current
implementation partitions the data according to the (very limited) number of
d-locks available on the device, rather than allowing locking at whatever
granularity the system actually requires. The disk does not have the
intelligence required to participate in actual file system transactions,
which makes dealing with failure conditions at best awkward and in some
cases just about impossible. Disk-local locks make it difficult to use the
additional available performance of replicas effectively (i.e., the replica
disks are effectively only standbys until needed) unless the lock
information is also replicated at the replicas (which is expensive and,
again, difficult to keep synchronized across possible failures).
(? don’t want
> participating fs to communicate too much on the network).
A good DLM uses only as much communication as necessary: fewer messages
than the d-lock approach, and a far better system results.
My question is
how
> far are we away from seeing dlocks getting implemented on all scsi disks.
Infinity: they’ll never be a good solution. Even the CMU NASD/OOD people
have come to recognize this. D-locks appear to have originated with a bunch
of university researchers eager to make their fortunes plus a disk
manufacturer eager to find a way to differentiate its hardware, rather than
because of any intrinsic merit.
- bill
----- Original Message -----
From: “Geoff Clow”
To: “File Systems Developers”
Sent: Friday, February 22, 2002 9:03 PM
Subject: [ntfsd] Re: How to share storage device on logical block acce ss
level?
…
> Putting the GLM in the disk array is an interesting idea, since it would
be
> implemented in hardware and accessed at SAN bus speeds, rather than in a
> software server accessed over the LAN. The Lock, SN Check and Disk Read
> could be combined in a single call, implemented in the device, where the
> data is not transferred if the lock serial number is current. A single
> mechanism of this sort could support both file level and block level
> locking.
The idea of embedding that degree of intelligence in an array is
considerably less silly than that of embedding it in each disk, but at that
point the array starts looking considerably less like a ‘device’ than like a
full-fledged data server that cooperates both with its peers (it’s not a
single, central server but an expandible set of servers that provide
features like whole-site mirroring for disaster tolerance) and with its
clients in access coordination (which is in fact what I believe the best
eventual solutions will look like).
>
> -----Original Message-----
> From: Jamey Kirby [mailto:xxxxx@storagecraft.com]
> Sent: Friday, February 22, 2002 2:45 PM
> To: File Systems Developers
> Subject: [ntfsd] Re: How to share storage device on logical block acce ss
> level?
>
>
> I suspect it is not as much the cache manager, but how the current file
> systems use the cache manager. To support shared disks, you will need a
> whopping cache consistency and locking mechanism; similar to oplocks.
> However, oplock are controlled on the server and in a SAN the device
itself
> is a server.
As Geoff observed elsewhere, locking control can be (and I would suggest is
best) distributed among the hosts anyway. And he also noted that in the
absence of the problems that mapped files create for purging the host cache
on demand (to release a given portion of a file for modification elsewhere)
building such a system is feasible.
- bill
iSCSI can be the solution.
----- Original Message -----
From: “Arthur Kreitman”
To: “File Systems Developers”
Sent: Tuesday, April 23, 2002 2:56 AM
Subject: [ntfsd] Re: How to share storage device on logical block acce ss level?
> I’m also interested in the “solution” to this problem. I’d love to
> know the name of one commercially successful (success==real market share)
> product that “solves” this problem.
>
> -----Original Message-----
> From: Eric Lee Steadle [mailto:xxxxx@spinnakernet.com]
> Sent: Monday, April 22, 2002 6:32 PM
> To: File Systems Developers
> Subject: [ntfsd] Re: How to share storage device on logical block acce ss
> level?
>
>
> >> First, its a hard problem.
> >Hard? Well, the implementation is not trivial. But it is definitely a
> >solved problem, so the issue is not with uncharted territory.
>
> I’m interested in how this problem was “solved”.
> Theoretically, or practically – I don’t care.
> Can you provide details, or a link, maybe?
>
>
> >No, a good NAS box does something significantly different. In fact, the
> >best NAS box would likely be a box containing an internal cluster of
> nodes
> >using a clustered file system to serve the data to the rest of the NAS
> world.
>
> Yes, I agree 100% 
>
>
> >While there are certainly limits to the degree to which existing
> distributed
> >lock managers can scale, those limits are significantly higher than the
> >limits to which a ‘good NAS box’ can scale - at least when considered as an
> >alternative for intense server-style cluster applications (the kind that a
> >‘cluster file system’ is designed to support) as opposed to light access by
> >possibly very large numbers of casual clients (where a NAS-style approach
> is
> >entirely appropriate).
>
> It’s been my experience that the scalability of a NAS box is limited by
> the network and processing resources needed by the CIFS protocol. NFS needs
> very
> little of either to get good performance. CIFS seems to require gobs of
> both,
> just to piddle along. Distributed Lock Managment seems like it wouldn’t
> affect
> CIFS performance all that much. I’m interested in your opinion on the
> matter.
>
>
> ERX
>
>
>
>
>
> —
> You are currently subscribed to ntfsd as: xxxxx@congruent.com
> To unsubscribe send a blank email to %%email.unsub%%
>
> —
> You are currently subscribed to ntfsd as: xxxxx@storagecraft.com
> To unsubscribe send a blank email to %%email.unsub%%
>
----- Original Message -----
From: “Arthur Kreitman”
To: “File Systems Developers”
Sent: Monday, April 22, 2002 6:56 PM
Subject: [ntfsd] Re: How to share storage device on logical block acce ss
level?
> I’m also interested in the “solution” to this problem. I’d love to
> know the name of one commercially successful (success==real market share)
> product that “solves” this problem.
VMS. While it’s relegated to niche status today, for a good portion of the
'80s VAX/VMS dominated the minicomputer/superminicomputer market, and from
1984 onward using clusters (indeed, the first commonly-used cluster product
and arguably the first cluster product period, though Tandem and IBM might
contend otherwise) which included a fully-distributed, shared-device cluster
file system. And VMS’s later fade had nothing to do with any inadequacies
in such areas (though with $4 billion annual revenue and $800 million annual
profit right through Y2K, despite nearly a decade of neglect by its owners,
it would not be too difficult to make a case that it was ‘commercially
successful’ right up into the periods your own experience seems to
encompass).
Any other questions?
- bill