[OpenAFS] Commercial AFS backups

seth vidal skvidal@linux.duke.edu
Mon, 09 Oct 2006 13:57:30 -0400

> Teradactyl generally avoids chiming in on subjects such as these but si=
nce we=20
> were named specifically on this thread it seems justified to comment.  =
> true, Teradactyl is a small company with a customer base that is virtua=
lly all=20
> large government and educational entities.  We fully support the OpenAF=
> community.  We also need to pay our staff.  Therefore, the True increme=
> Backup System (TiBS) software does cost money.
> A recent post by skvidal@linux.duke.edu states =E2=80=9CTeradactyl worr=
ies the hell out=20
> of me. They're a small company with about 12 customers from what I can =
tell AND=20
> their closed source. So if they go away so does my ability to deal with=
> backups. Not a good combination. At least with TSM you had only to worr=
y about=20
> IBM changing their mind on you - but not outright vanishing as a result=
 of a few=20
> bad trades.=E2=80=9D  Teradactyl is a small company.  We don=E2=80=99t =
advertise all of our=20
> customers because many have privacy policies and we respect them.  We a=
re closed=20
> source just like TSM.  We have not changed our mind and dropped support=
 for AFS.=20
>   Conversely, we have been improving the product for OpenAFS and are wo=
rking on=20
> significant solutions for large scale OpenAFS cells.  Most importantly,=
 we have=20
> paid for all of our growth organically and don=E2=80=99t have any exter=
nal shareholders=20
> or public debt to pull us away from what the =E2=80=9Cmarket=E2=80=9D f=
eels is not a profitable=20
> platform to support.
> TiBS doesn=E2=80=99t require a license key for restores to ensure that =
our customers=20
> have access to their data at all times.  Multiple restore servers can b=
e setup=20
> without charge for massive parallel restores in the event of catastroph=
> failure.  We only license the backup function and our model is built on=
> processing power to scale with your needs.

Great. But that doesn't matter to what I wrote.

> The real issue at this AFS site and others is that they need to pay for=
> software.  Presumably this site purchased TSM and/or paid large mainten=
ance fees=20
> to support the application but now the same site seems unwilling to buy=
> solution from a company that actively maintains support for OpenAFS and=
> invests precious resources in the OpenAFS community.  It stuns us, as o=
wners of=20
> Teradactyl, that entities will devote considerable amounts of Universit=
y and=20
> Government resources, in the form of salaries and associated computing =
> to find a way to produce a solution which lacks scale and many of the b=
> our patent brings to the community to solve the problems associated wit=
h OpenAFS=20
> backups.  We have built this company with our own money, time, and a gr=
eat deal=20
> of effort.  We don=E2=80=99t have shareholders pushing us away from AFS=
 because it lacks=20
> the revenue required to meet break even points that the larger corporat=
ions have=20
> with their size and structure.  If you want commercial support of OpenA=
> you want quality customer service, then it will cost some money.  When =
you try=20
> and talk a private company into basically giving their software away to=
> because of your name and then publicly criticize them when they refuse,=
 you will=20
> find little commercial interest.

When did I say I wanted you to give away your software? I didn't. I said
I was reticent to use or encourage its use b/c of the role it plays in
an infrastructure and it being closed source. I never expect you to give
anything away, nor did I ask that.

> Perhaps, backing a small firm that uses its money to grow and improve w=
> produce better results than trying to reinvent the wheel or publicly at=
> them.  The more OpenAFS sites that become our client the more time we c=
an spend=20
> in pushing advancements in synthetic consolidation for OpenAFS and back=
ups on=20
> other platforms.  With a 100% customer retention rate, approaching 8 ye=
ars in=20
> business, and more than 11 years of work on our technology, we have dem=
> a solid, viable, and on-going solution.  We even take some of that hard=
> money and put it back into the OpenAFS community.  Some solutions are f=
> available for smaller sites and we do not try to compete in this space.=
> Teradactyl is focused on the larger OpenAFS cells that need advancement=
s in=20
> synthetic consolidation to scale their cells while maintaining a viable=
> and disaster recovery solution.  We simply ask you to avoid tarnishing =
our name=20
> when you lack the facts about us and our technology.  Not very cool dud=

What facts did I make up? You are a small company. Small companies have
a hard time weathering difficult economic conditions. I need backups to
be restore-able for years - sometimes during economic downturns. I don't
want guessing games about the viability of the backup mechanism.
Moreover I don't have any attachment or interest in TSM - it is what we
were saddled with in the group I work with at duke. I'd prefer if we
were not using it, either. I think backup solutions should be like
document storage solutions - not given to the whims of a single company
but based on vendor-independent standards and in the best of all worlds
an open-source developed and maintained solution.

openafs is open source - it's the only only functional implementation of
an afs server left that I know of. It would make sense that it would
have a backup mechanism coupled to it that was also open source.

I never said your product was not good or functional and I will not say
such things. It looks like TIBS would do a great job my concern is with
the other realities that come with relying on closed-source-software.
When IBM announced dropping AFS support in TSM I said the exact same
thing about IBM to our folks reviewing our contract.

I worry about the impact a company's decisions or the markets discretion
will have on the viability of me to maintain my backups and therefore my
users ability to get to their data.

What's 'Not cool' here is you deciding that no one is allowed to present
doubts and concerns about the longtime functionality of relying on
closed source software solutions especially those from small companies
and especially when our economy is taking a swing into a downturn.

So let me make sure my point is abundantly clear:
 "In my opinion closed-source software that does not rely on a public
standard with multiple implementations is a bad choice for backup
solutions. More so from small companies or 'unique' implementations."

The only reason Teradactyl was mentioned at all was b/c of it being an
example of the above.

I have a whole other set of criteria I use when evaluating the viability
of open-source backup solutions. They have to do with the viability of
the project and the commonality of the implementation.