hardware raid solutions?

[Eric S. Johansson]

ubuntu at rio.vg wrote:
> 
> Is it just me, or does ZFS just sound too good to be true?
> 
>>From the opensolaris site:
> 
> "What is ZFS?
> 
> ZFS is a new kind of filesystem that provides simple administration,
> transactional semantics, end-to-end data integrity, and immense
> scalability. ZFS is not an incremental improvement to existing
> technology; it is a fundamentally new approach to data management. We've
> blown away 20 years of obsolete assumptions, eliminated complexity at
> the source, and created a storage system that's actually a pleasure to use."
> 
> Doesn't that just sound like a scam?  I keep waiting to hear what the
> downside is...  or maybe I've just been in this business too long and
> have gotten burned too many times...
> 

yes, on one hand it does sound like a scam.  At the same time, many of 
the things we have in our lives would sound like scams to people 20 
years ago.  Hell, the fact that I am talking to the computer and it is 
recognizing accurately 19 words of the 20 would sound like a scam even 
15 years ago.

as for all the code size, speed, etc. etc. claims, ZFS fits with what I 
know is possible.  I've worked on filesystems and some of the things 
they have solved were questions I was asking 10+ years ago.  Such as 
what happens if you consider your disk storage a heap and start 
aggregating physical disks under that heap.  I never quite got the 
naming problem but it looks like they have come up with a good solution. 
  The copy on right tricks make a lot of sense and are very simple to 
implement.

Do they really get better performance?  We don't know but unlike most 
snake oil, we can test it.  That's the joy of open source.  You don't 
have to question, you can try it out.

I will admit to a bit of sensitivity to random doubting claims coming 
over mailing lists.  I've developed some anti-spam techniques[1] in the 
past which got shot down, not because it wasn't any good but because the 
people making comments were too lazy to do the math.  It wasn't until 
much later that I built the model myself and showed had they were off by 
at least a couple orders of magnitude on the effects of the system.

there was even one paper on this technique showing how the core 
assumptions were "wrong" but when I presented a counter model, the paper 
originators were amazingly silent.

so I ask you to be careful and measured in your doubt without 
examination of the code or system.  Real or not real, without trying and 
modeling, you can't tell.  If you have some facts, then it's OK to throw 
stones.  :-)

--- eric

[1] python knowledgeable help wanted.  Job descriptions available on 
request.