The State of CAS

"Malcolm Weir" wrote in message
...

....

Symmetrix didn't, but Harmonix (EMC's successful AS/400 offering,
perpetually overshadowed by Symm) did.

My bad, then. I was involved with the early stages of the design of the
precursor to what eventually I think became Harmonix (how's that for a
tenuous inheritance path?) and we were planning it then - I thought based on
technology that already existed in Symmetrix.

On reflection, I'm not too sure about my Unisys comment, either: I'm pretty
sure that Unisys used some very strange sector sizes on its disks, but not
necessarily to implement redundancy checks.

But other systems have at least claimed to have had them, including Sun
storage and NetApp, IIRC.

- bill

Hi Bill-

Thanks for the info. I am somewhat new to the storage industry, so it
sounds as though there is a lot that I could learn from you.

First off, I am wondering why you claim that CAS is a marketing
gimmick? The term seems to be coming more and more embraced by the
industry as a whole. For instance Data Center Technologies just
released a new white paper on CAS at:

http://www.dacentec.com/Collateral/D...20on%20CAS.pdf

Also, you are correct that CAS is dependent on immutable data. That
is the whole point of CAS, providing optimal storage for fixed content
data (audio, video, email, attachments, etc). In this case optimal is
in terms of cost, performance, and utilization of storage capacity.
Also, as you suggested, the exploding compliance market can not get
enough of this technology.

To clarify on the location independence of CAS objects, this really
applies to writing a piece of content once, and then having that
object replicated/migrated "as is" in its immutable form to multiple
clusters for backup or distributed access strategies. It also applies
to taking the content (as is) and migrating it through it's lifecycle
to newer and newer drive technology. All of this migration is
completly transparent to the application when the read request is
made.

Finally, the last point on authenticity. Centera is not just a WORM
device, it actually guards against corrupt data that results from
physical errors to the storage media. Because the data is stored
based on the content address, the content address is used to
authenticate the data at the point that it is read. Thus if the
physical data was corrupted because of errors to the storage media,
say after 8 years of residing on tape, etc with no access during that
time, the data will still be authenticated based on the content
address. Is this something that the other products/technologies that
you mentioned also support?

I appreciate your feedback and value your opinions.

Thanks-
Ace

http://groups.yahoo.com/group/CenteraTechGroup/

Paul wrote in message ...

BTW, Did you ever get that porn site login you were looking for?

Yes Paul, thank you very much for providing the login information. We
especially enjoyed the secret home video that you provided of you and
your cousin. Did she have any idea that you were filming her?

You 'da man lil' Pauly!!!

Talk to you soon...
Ace

In article ,
says...


You 'da man lil' Pauly!!!

Just keep posting not-so-thinly veiled advertisements for EMC storage,
and watch your (and your company's) credibility rise, bro...

"acebgur" wrote in message
om...
Hi Bill-

Thanks for the info. I am somewhat new to the storage industry, so it
sounds as though there is a lot that I could learn from you.

Indeed there is.


First off, I am wondering why you claim that CAS is a marketing
gimmick? The term seems to be coming more and more embraced by the
industry as a whole. For instance Data Center Technologies just
released a new white paper on CAS at:

http://www.dacentec.com/Collateral/D...20on%20CAS.pdf

One of the things new entrants to the storage industy (and likely many other
industries) need to learn is to distinguish between marketing and
technology.

If you are a salesperson, or a marketdroid, you'll likely use 'white papers'
a lot: they're written for, and by, salespeople and marketdroids, they are
written specifically to sell a particular product rather than to provide
objective evaluations, and their technical value can be estimated
accordingly.

The white paper you cite above is an excellent example of this. The only
*actual* CAS-specific advantage among all its verbiage is CAS's ability to
collapse redundant copies of the same material (usually at the whole-file
level) into a single copy - precisely the advantage I noted in my first
response, and an ability which, as I noted, a file system could choose to
offer as well. Every other aspect of the technology has had equivalents in
file-system technology for decades (unique identifiers = inode addresses,
backup overhead reduction = archive bits, distributed storage = distributed
file system, content validation = additional redundancy check, etc.). And
CAS can accomplish these things only after some minimal CAS awareness has
been added to whatever conventional storage organization (typically, a file
system) intermediates between the CAS storage and the application.

If you actually want to *learn* something about storage, first acquaint
yourself with the basic physical properties of disk operation (tape is
already less central to most storage issues), then start reading technical
journals: the annual Conference on Very Large Databases covers an
incredibly wide range of relevant material, the ACM's Transactions on
Database Systems does as well, and the Usenix Conference on File and Storage
Technologies (FaST) may be the best place to begin your education (though
the early chapters in Jim Gray's "Transaction Processing: Concepts and
Techniques" are probably the best single description of soup-to-nuts storage
technologies that I've ever encountered).


Also, you are correct that CAS is dependent on immutable data. That
is the whole point of CAS, providing optimal storage for fixed content
data (audio, video, email, attachments, etc). In this case optimal is
in terms of cost, performance, and utilization of storage capacity.
Also, as you suggested, the exploding compliance market can not get
enough of this technology.

You really do sound like a marketdroid: are you sure you aren't one
already?

Optimal? Bull****. CAS solutions don't use any less hardware resources,
are slower (because of the index look-up required to translate the file
signature to the file's actual location, a stage which is not required in a
conventional file system), and, as a non-standard proprietary solution, cost
an arm and a leg compared to dirt-simple conventional file storage (which
can easily be made robust by use of underlying RAID facilities).

The market for absolutely bomb-proof reliability, via additional redundancy
checks, has proven to be *very* small: modern disks are extremely good at
either delivering the content that was written or returning a failure
indication (allowing you to get the content from its redundant copy
elsewhere), and the various interfaces and busses between them and main
memory are also at least fairly reliable in this regard (I do happen to
think that there's a place for additional redundancy checking in a
conventional file system, at least optionally, but only because it can be
done quite inexpensively: the need certainly wouldn't normally justify
considering a CAS solution).

And pure software solutions using conventional file storage have also been
accepted for 'compliance' purposes.


To clarify on the location independence of CAS objects, this really
applies to writing a piece of content once, and then having that
object replicated/migrated "as is" in its immutable form to multiple
clusters for backup or distributed access strategies. It also applies
to taking the content (as is) and migrating it through it's lifecycle
to newer and newer drive technology. All of this migration is
completly transparent to the application when the read request is
made.

Just like competent file and storage systems do. You might want to acquaint
yourself with VMS clustering: it has existed since 1984, and via both
shared storage and distributed host-based mirror facilities is fully capable
of the functions you describe above - and supports mutable files in the
process.


Finally, the last point on authenticity. Centera is not just a WORM
device, it actually guards against corrupt data that results from
physical errors to the storage media. Because the data is stored
based on the content address, the content address is used to
authenticate the data at the point that it is read. Thus if the
physical data was corrupted because of errors to the storage media,
say after 8 years of residing on tape, etc with no access during that
time, the data will still be authenticated based on the content
address. Is this something that the other products/technologies that
you mentioned also support?

Yes: I described it in my previous response. The check is usually
implemented at the block storage level rather than the whole-file level.

Of course, the redundancy already present in disk and tape devices, and the
busses and interfaces to them, catches virtually all such errors already
without any such special checks: the worst (somewhat anecdotal) estimate
I've seen for undetected end-to-end errors is on the order of 1 in 10^13
bits, and that was for PC-quality hardware quite a few years ago.

- bill

Hi Bill-

Thanks again for the info. I am wondering why some on this board are
so adverse to CAS and CAS based discussions. What do you, and others,
consider the "sweet spot" for messages in this group?

It sounds as though you consider CAS to be nothing more than SW Voodoo
that makes use of standard drive technology. Even if this is true,
are you claiming that there is no value-add to be derived? And that
this technology will pretty much evaporate in the very near future?
To me it seems as though interest in the technology is growing at a
very rapid pace.

As I mentioned I am very new to the HW industry, and am coming from a
SW development background. Again it sounds as CAS could be considered
a SW layer that makes better use of storage space. It sounds as
though you may not agree with the term "better" in this case. The
point is that I am not an expert on storage, and as I mentioned
previously I appreciate your feedback.

You also seem to be claiming that CAS and file-systems are 2 different
things. The truth is that there are companies, such as Permabit, that
expose standard file systems on top of CAS managed storage devices.
Also, there are several EMC Centera SW solutions that do similar
things for Centera (e.g.; CAG, DX2K, StorageSwitch, etc).

Finally, is this not the right group to be discussing this topic?
Every time that I post something here I feel as though one of your
cronies is throwing me out headfirst through the swinging saloon
doors. But on the flipside 10 or 15 new members silently join the
CenteraTechGroup. It is very hard to figure out.

Thanks for the recomendations on research materials.

Thanks!
Ace

http://groups.yahoo.com/group/CenteraTechGroup/

In article , (acebgur) writes:
Hi Bill-

Thanks again for the info. I am wondering why some on this board are
so adverse to CAS and CAS based discussions.

It is a niche. It requires tying into an API. In a few years,
I suspect various solutions won't require API tie-in and so many
of these products today will be flash in the pans.

How will it be done? 3PAR is doing something similar today.
They really won't give you that much storage, just a container
that can grow that full. Similarly, in a future time those blocks you
think on a fibre based driver are actually on iSCSI or ATA and are
migrated to and fro in a QoS fashion. It appears to be a fabric
level thing to get it done. When/if Brocade and partners get it
done, it will render an API based solution awkward and obsolete.

Rob

"acebgur" wrote in message
om...
Hi Bill-

Thanks again for the info. I am wondering why some on this board are
so adverse to CAS and CAS based discussions.

I'd suggest that what they're averse (I believe that's the word you wanted)
to is hype, whether of the CAS variety or any other.

What do you, and others,
consider the "sweet spot" for messages in this group?

It sounds as though you consider CAS to be nothing more than SW Voodoo
that makes use of standard drive technology.

No, it's a bit less than that.

Even if this is true,
are you claiming that there is no value-add to be derived?

In most cases, yes. In many of the rest, the value added doesn't begin to
approach the added cost.

And that
this technology will pretty much evaporate in the very near future?

I'm not in the crystall-ball-gazing business.

....

You also seem to be claiming that CAS and file-systems are 2 different
things.

They are, though CAS duplicates one of the lower layers of a typical file
system to a significant degree.

The truth is that there are companies, such as Permabit, that
expose standard file systems on top of CAS managed storage devices.

How do they handle small updates to large files (I guess they might be able
to using block-based CAS, but that's hardly fully supported across all CAS
vendors)?

....

But on the flipside 10 or 15 new members silently join the
CenteraTechGroup. It is very hard to figure out.

The phrase "There's a sucker born every minute" comes to mind. Of course,
if these CAS customers are used to paying Symmetrix prices for storage, then
Centera may look like a bargain to them.

- bill

(Rob Young) wrote in message ...

It is a niche. It requires tying into an API. In a few years,
I suspect various solutions won't require API tie-in and so many
of these products today will be flash in the pans.


Hi Rob-

I just wanted to clarify one of your points. Not all CAS-based
storage solutions require the use of an API. For instance Centera
exposes a file system via an application gateway. There are also
HSM/ILM based solutions that expose Centera as a file system. And,
aside from Centera, Permabit offers a full software only CAS solution
that is completely file based, with no API.

It is hard to imagine that the technology will become a "flash in the
pan". Compliance regulations are only getting stricter as we move
forward. When a financial institution is required to save all of
their emails for 7+ years and be able to deliver/authenticate them on
a moments notice, then a CAS-based product really becomes the ideal
solution. Otherwise which company out there wants to save 100+
individual copies of the same 2 MB attachment for 7+ years?

Thanks-
Ace

In article ,
acebgur wrote:
forward. When a financial institution is required to save all of
their emails for 7+ years and be able to deliver/authenticate them on
a moments notice, then a CAS-based product really becomes the ideal
solution. Otherwise which company out there wants to save 100+
individual copies of the same 2 MB attachment for 7+ years?

Why would it be necessary to do that? There are mail servers that don't
keep multiple copies of identical messages and/or attachments.

--
I've seen things you people can't imagine. Chimneysweeps on fire over the roofs
of London. I've watched kite-strings glitter in the sun at Hyde Park Gate. All
these things will be lost in time, like chalk-paintings in the rain. `-_-'
Time for your nap. | Peter da Silva | Har du kramat din varg, idag? 'U`