[FoRK] Seagate Kinetic

Tomasz Rola rtomek at ceti.pl
Fri Oct 25 12:41:33 PDT 2013


On Fri, 25 Oct 2013, Eugen Leitl wrote:

> 
> (caveat, enough marketese to give you brain cancer)
> 
> http://www.seagate.com/tech-insights/kinetic-vision-how-seagate-new-developer-tools-meets-the-needs-of-cloud-storage-platforms-master-ti/
> 
[...]
> What do they say?
> 
> - Managing key (object) ordering
> 
> I don't see what this is for. Ok, if you've got hybrid memory
> you could store index on flash, and data on spinning rust.

Hmm. So instead of having even faster index in RAM they would want me to 
operate on it over Ethernet?

> - Quality of service
> 
> Looks relevant if you've got thousands of spindles and
> a really busy network.

I admit it, whenever I hear of QoS, my stomach wants to revolt...

> - Policy-based drive-to-drive data migration
> 
> Ok, so you can detect damage and flush out stored blobs
> to a new device before swapping it out.

Would be much better to replicate before drive blows up. I wonder if they 
give much control over replication policy, how do they resolve updates in 
such case? Again, it doesn't sound like much more effective, because 
instead of sql server making decision, they want their proprietary lib to 
do the same stuff. Just speculating, based on what I understood from the 
link. So if such is the case, where is performance gain?

> - Handling of partial device failures and other management
> 
> So this is a bit like zfs
> 
> - Data-at-rest security  
> 
> Same thing, sounds like scrubbing.
> 
> Do you see anything interesting for you in there? 

>From my very subjective perspective, nothing.

They cater to the very specific user, apparently. Even if I had such a 
need as they say they could meet, there is a question of replacing 
existing stuff/solution with something new. Say, if I had twenty/fifty 
years worth of sql database, would I happily jumped into this new storage? 
I guess not. So, this may be easily adopted by some new initiative while 
so called legacy systems will probably stay conservative.

However, I am very very backwards and my needs are close to personal - for 
example, I can easily resign from hw raid and stay with sw one (or rather, 
not adopt hw raid at all, because it gives me more flexibility). I think 
their target clientele for this kind of drive is the one who actually 
builds/operates data stores - meaning, with concrete walls etc. Perhaps it 
will be possible for some small scale operation (SOHO etc) to use it 
instead of normal database. But, this sounds strange, because there are 
ready solutions for such small scale.

So, perhaps the very big data, medium size records type of business will 
be happy with this. And they will still backup to tape, because nobody 
gets fired for backuping to tapes (but don't cite me and this is not a 
business advice).

BTW, this idea seems to go against having processing close to the data, so 
I expect there will be great potential for bottlenecks.

BTW2, "read whole thing, write whole thing" - great, but if thing is a 
gigabyte?

BTW3, twenty years from now, if I get random drive and some paper 
documentation, will I be able to access data on it successfuly in case of 
this key/val storage? Right now, I expect better chance with old style, 
PATA/SATA/SCSI stuff and their block-based design.

(Like I wrote above, my perspective is nowhere close to any kind of 
business I can imagine, so my opinions may not apply to such businesses).

Regards,
Tomasz Rola

--
** A C programmer asked whether computer had Buddha's nature.      **
** As the answer, master did "rm -rif" on the programmer's home    **
** directory. And then the C programmer became enlightened...      **
**                                                                 **
** Tomasz Rola          mailto:tomasz_rola at bigfoot.com             **


More information about the FoRK mailing list