Storage news

[GullLars]

I thought it would be usefull to have a thread where people can link to interresting storage news. As a team effort it will make sure interresting stuff don't pass us by.
I'll edit this post later with links to usefull past news articles, like Anandtechs SSD anthology and follow ups.

I'll start off by linking to a site dedicated to this stuff, though a bit more towards the enterprise:
Storagesearch.com
Storagesearch.com/SSD

[GullLars]

Collection of news for June 2010
(i will gather and list links to all news stories this month in this post for easier navigation)

June 1.
OCZ shows off HSDL (High Speed Data Link). A 20Gbps cable interface using a PCIe 2.0 x4 host card. They also have an x16 version with 4 HSDL ports.

June 3.
Anandtech: 2010 Value SSD (~$100) Roundup: Kingston and OCZ take on Intel.

[theystolemyname]

Sandforce SF-1200 SSD firmware comparison - http://benchmarkreviews.com/index.ph...sk=view&id=531

I'm still waiting on the dust to settle on these new drives before I jump in. I'm leaning toward the corsair force series, but they don't seem to be available anywhere yet.

[Computurd]

ADATA drives hitting 400 mb/s sequential, out in fall.

http://www.tweaktown.com/articles/33...iew/index.html

[mgon154]

Wow interesting news already. The only challenge with a thread like this though is managing it, but the advantage is has is slow traffic around here. Once the thread slows down it get piled in the list. I like it though and will bump if needed! Thanks for the news guys!

[Computurd]

well here is even better, check this out a new device due out in the fall as well, 500mb/s seq reads!

http://www.tweaktown.com/news/15584/...ast/index.html

[onex]

http://www.pr.com/press-release/238527
foremay again, mostly just for looking,
250K/250K R/W IOPs 2.5GBps/2.3GBps @seq - enterprise PCI-e SSD with capacities up to 8TB.

[lowfat]

http://www.youtube.com/watch?v=6nzJx...layer_embedded

OCZ consumer grade PCIe SSD.

[onex]

some more info on this:
http://www.engadget.com/2010/06/02/o...ing-fast-hsdl/

a PCIe card with between 128GB and 480GB of SSD onboard. The unit we saw here in Taipei was near-final, with a pair of SandForce SF-1200 controllers, a daughter board expansion slot (for possibly combining two in a RAID scenario) and a promised starting price of around $400 to $600 (not to mention read / write times of nearly 540MB/sec). Obviously, this hits well below the multi-thousand dollar PCIe SSD options from Fusion-io and company

.
this link was here somewhere (yet the PCI-e SSD was totally ignored),
should probably be boot-able & TRIMed too as the new Z drives are.

[GullLars]

Sorry i've been inactive the last couple of days, and likely will be mostly the rest of the week. I just moved, and have some stuff to take care of. I'll pop in about an hour a day and stay roughly up to date, but i won't organize this thread or make any more graphs before the weekend or early next week. Hope you understand. Till then, keep up posting news here, i see you do well without me

[Computurd]

yeah, real life sucks! Last time i moved it took me two weeks before i even got back online, had to get the house setup! and i knew once i got online nothing else would get done so i went into self imposed exile, the best two weeks on the internet in years....

[tunggad]

http://www.computerbase.de/news/hard...s_sata_6_gbit/

[onex]

http://www.computerbase.de/news/hard...s_sata_6_gbit/

hmfp, you missed this one.

some more at this:
http://pc.nikkeibp.co.jp/article/news/20100603/1025325/
few products form that company using marvel's and indilinx next gen controllers and seemingly some more companies for different speeds.

Edit:
and the full line of photofast SSD's with various controllers:
here

[Computurd]

jesus those are fast onex. 500 mbs read! wonder what randoms are like....

is it just me or are all of the newer drives showing bad latency numbers? what is up with that??? I was reading this inferno review, and yet again ANOTHER ssd with latency of over .10
what gives? are we gaining higher speeds at teh expense of latency? look at this page and how horrible these newer drives (sandforce / C300 included) stack up against intels and ocz's old school vertex. (well, the vertex isnt on there but it is .08 for the record)

http://www.tweaktown.com/reviews/334...ve/index7.html

[alfaunits]

Maybe they want to target the consumer system - which doesn't give a f*** about latency to that extent but does for sequentials
I would prefer an SSD which does >10K randoms, gives me 1GB/s sequentials in return, but provides excellent durability.

[Computurd]

well yea that the gist...faster sequential speed and terrible latency. what gives?

[onex]

the C300 uses dual ARM A9 cores for the marvell controller,
the C300 is slow with regards to latency and especially on write, one possibility is that the drive is caching small random write packets before sending them to the nand though this could be probably be tested with bigger packets which would saturate the cache and force it to send the data strait to the nand, this could also create a bottleneck so the drive would show up even higher latencies.
maybe it is also a good way to test it's NCQ capabilities with higher QD.
yet intel is intel, and seem like no one can take they're understanding with controller design so far.
even viewing the new sandforce controllers next to the Intels, reveals a great deal of bandwidth capabilities in favor of the sandforce yet still at random read HDTune bench, the intels latency and overall performance as for 500KB is much better (9ms vs 2) and we are talking about a controller which is about 2 years old now...

from a brief look, it seems as if crucial has brought a big bang drive to the market with it's high sequential bandwidth and SATA3 interface yet when we are looking closer, it seems as if it comes as compensate of writing latency which can imply about not a state of the art design of the controller, like a GMC engine which also brings about a heavy car.

what is seen through the drive is the obvious relation between the controller it self to the latencies, the SF drives operate the same and the Marvel and Intel's are different form them and each other.
the marvel; also could suffer from overhead due to the fact it is using dual processor and so has to have a more complex inter process communication (IPC).

the WD drive is using a Jmicron JMF-612 (the 618 only difference is allowing toshiba nand) controller and so they have some issues, the patriot (zypher) is as seen, using the same one.

as for the barefoot vs the SF, well, this is uncertain, maybe the compression with the SF has something to do with it, the Intel is (should be) using none.
as long as they produce better IOPs, it can also be a design consideration, that a bit more latency doesn't harm the drive overall performance as long as it performs better at a different level.

[onex]

good new,
the SNIA (Storage Networking Industry Association) has announced for some time the formation of SSSI (Solid State Storage Initiative) which would arrange standardization for solid state devices and better cooperation within the associated manufacturers.
the SSSI is about to release a solid state device Performance Test Suit (PTS) as to standardize the testing of SSD's by manufacturers and so deliver a transparent operation of they're products in an equal comparison with others.
now what that means is that there would be no more seeing of 10K IOPs devices where it is unsure whether this is a 4KB test, 2KB or 512B one,
as well, no specific benching by manufacturers which would allow them to levitate results and so manipulate the markets with theoretical performance capabilities, and so would require them to do they're best effort in order to deliver a state of the art hardware where the gain would obviously benefit the better designed and not the better tweaked,
as a side effect, this should also mediate costs of the hardware.

you can all view the testing methods delivered by SMART Modular Technology for they're XceedIOPS SSD line here
which states an IOMeter run over random and sequential patterns preconditioned by a full scale 512B to 1MB 100%/0% R/W to 0%/100% R/W leveled 12 hour work load.

now this is getting a bit advanced for most users, so please bear in mind that if any of you are willing to get a better understanding of SSD you will have to step forward into it, and don't forget to be patient with newcomers which might get overwhelmed with the amount of information lent, we don't want people running away by any lack of understanding rather the opposite, this is the meaning of forums.

you can all read about the news here, and have a view of the SSSI targets here.
good luck.

though if people are not going to respond to it, i'm out of here,
this kind of in-serious behavior has gone too far already.

be well.

you simply don't give a F$#$ about anything, people just turn they're back at you and go on, typical behavior here, spineless and selfish,
freaking egoists.

[Computurd]

Wow, that is a very vague rant there. It would help if you aim your frustrations at individuals, as I do. Gets you farther.
It is strange though, onex, that you flip out unprovoked, and this isn't the first time.
If you have an issue with certain individuals just let it out, bro. Let it fly!

[mbreslin]

Hey guys good news, there is a new standardized ssd testing platform being created.

P.S. Use it or I'm leaving.

What a horrible way to start a dialogue. Some kind of universal ssd performance metric is an awesome idea and I'd certainly welcome it. I don't think any threats or such needed be included.

[onex]

i never meant threatening, it is just people go over important stuff that can take us further in behalf of Fermi threads...
how can we move on if people are still occupied with being enthusiasts over this piece of hardware or the other?
of course this is nice, though we have to step forward, research, investigate.
taking things seriously meaning figuring out how things operates, and not being test rabbits of manufacturers, bound to marketing hypes and technics,
each has to be capable of understanding the hardware from bottom to top and not just doing what he feels or comfortable with.
we have to get to a point where we can design and bring new idea's, manufacturer engineers are not always aware at what is essentially good for customers and are often influenced by market demand, all sort of strains by competition between the groups, and even they're own way of thinking.
there is no us and them here, there is us as people who use and manufacture this equipment, we are all part of this equation,
it is our responsibility to be a full part of the process, not just wait for things to come at us, and it means, study, and study hard, understanding everything there is,
it also means of course, what people are not really willing to, growing up with it.
it happens though that one expect after 4 months of being in a forum to find people more collaborating and not in a state where each is doing only what he feels like or comfortable with,
i see you going on with that FP thread and just not paying attention to what is being said here, and it's just like WTH is going on here?! why? this is such an important step forward that would influence everyone, it is a step toward transparency, transparency is what brings trust and better stronger relationship between the individuals involved, this is a huge step.

and why not saying "hey, onex, this is a horrible way to start a dialogue"...
id he supposed to go over the thread 5 times to figure out what is wrong?
and if he won't understand, then? people would just ignore him and go on until he understand?
THIS is the problem! why not simply saying, hey, man, what's going on? where is caring? just to care.

we need to understand what people are thinking and feeling over here and not just share reviews of hardware,
there should be a bond between the members and not just respect based on they're knowledge.

and there seems to be lack of assisting others to understand, like saying "go figure it out by yourself, come back when you are up to it" "this is our level", not answering to people, there are groups at some forums and others at other , there are also some who seem to be talking only with people at "they're class", the yellow badges etc. .
there is a sense of distance, this is absurd where there should be more commitment to others..


this is it in general.

[mbreslin]

i see you going on with that FP thread and just not paying attention to what is being said here

This thread is about storage news it doesn't require a response to every post, I pay very close attention.

When there is a clear next gen storage winner (people disagree but I doubt it will be this year) I will be happy to be the guinea pig and purchase way more drives than I could ever need just to get to play with new hardware again.

[onex]

o.k, good luck with your way.

[GullLars]

I will also comment on this point i see mbreslin made and i thought was implicit.
This is a thread mainly for posting links to news articles and keeping a well structured log of interesting news, and maybe comment on them or elaborate a bit. Long rants and discussions about topics or news posted here would fit nicely in the SSD discussion/rant thread.

I'm sorry i've been kinda inactive the last week, and probably will for one more or so, i've just moved and only have internet access from my laptop ATM.

Here's a couple of news stories:

Kaminario launches RAM SSD ASAP

Editor:- June 14, 2010 - Kaminario launched its 1st product - an FC SAN connected acceleration appliance in which a grid of blade servers access upto terabytes of shared memory.

Link to datasheet
There is an expanded editors comment on storagesearch.com/ssd.html

And also from storagesearch:

Virident Systems launches enterprise PCIe SSD
Editor:- June 15, 2010 - Virident Systems today announced the immediate availability of its tachIOn - a fast PCIe SSD using SLC flash - with 800MB/s sustained R/W throughput, 200K sustained random IOPS (320K peak) and capacity options of 200 / 300 / 400GB.

Aimed at the enterprise acceleration market - the tachIOn's data intergity features include end to end error correction. Endurance is quoted as 24 years at 5TB writes / day.

As a curiosity, the name is likely derived from the elementary particle Tachyon, wich is a hypothetical FTL (faster than light) particle. A nice name choice for an SSD with SUSTAINED 800MB/s R/W throughput and 200K SUSTAINED random IOPS. (note that most manufacturers like to quote "up to").
Still, i would have liked sequential read to be higher, but then again "typical seq read" might be, since sustained is worst case.

[Computurd]

jesus lord for one device that is unbelievable.!