If "TLC NAND flash is typically rated at about 100–500 cycles" how will this technology ever be used for anything other than very low end low reliability products?
I'm not sure about Amazon's EC2. It talks about striping across volumes on their site though, and measures IOPS. Their low end is about 100 IOPS. My guess is that it's a massive mixture of different types of hardware and that based on what users need, they can custom configure just about anything.
I don't really agree with all of your analysis of the future of the market. I do agree that demand will continue to increase, and that it will be difficult to meet that demand. That will drive prices up, as well as margins, as there is higher demand / lower supply. It's kind of a win in either situation. If there is lots of NAND available at low prices, they can make lots of drives and replace HDD's more while still making a good margin. If NAND supply is more constrained, it will slow adoption of SSD's, but it will increase prices, and there will be a continuing steady upward growth in demand as more NAND production comes online. Either way is fast growth for SSD companies.
I actually agree with you that OCZ should consolidate products. I think there are too many consumer SSD's. I support their decision to move into enterprise and consumer simultaneously, and I support the decision to buy Indilinx (I think that will turn out to be a huge success - and already has probably). Where I take issue is the number of consumer drives out there. There's the octanes, petrols, vertex, agility, vertex plus, and like 5 other models, in many sizes, and in too many versions. It would cut costs I believe, and improve margins to make 2-3 models tops (in varying sizes). I think they may be going in that direction but it's unclear.
It's certainly true they are aggressively spending. If they were not in a secular growth market where all the companies combined will not be able to meet demand, I don't think it would be the right strategy. However, I think they see an opportunity and are rolling the dice. It takes a pair, but I like that in a CEO. I don't want them to be boring and play it safe - although I admit I have a very high risk (volatility?) tolerance. Playing it safe in this situation gets you nowhere.
What other alternatives are there? Clearly STEC is just a clusterfuck. That thing is a sinking ship. Only reason they are worth anything is the $200 mill or so on their balance sheet. WDC has crappy SATA2 drives and little tech expertise - they are shipping drives, but they are slow and old AND more importantly, it's a small part of their business, so the total revenue numbers don't move much (also because each SSD shipped can mean 1+ HDD's not shipped). FIO is a great company, but it's a different play - it's more software. You have the Woz factor, so you know you can get funding and automatic respect, but look at the valuation. Their revenues are lower than OCZ's, they are not profitable yet, and they are valued something like 8x OCZ P/S (that's a very rough number, have to check current financials to be more precise). I like FIO - but every company has a price at which it's cheap and at which it's expensive - OCZ is undervalued. FIO is overvalued relative to OCZ - sure, it might have upside, but not as much. All the other vendors have other reasons not to play them for the SSD story. Intel and Samsung are too big and diversified, doesn't move the needle. Corsair is private (maybe not for long - but also diversified into lots of computer components). I like MU - I think they make great drives that are very reliable based on that same Marvell hardware as current pre-BF3 Indilinx controllers. These drives are all reliable - read the Crucial reviews on Newegg or Amazon. I'm also long MU, so I get NAND exposure too. Also a cheap, undervalued company that is reinventing itself to take advantage of the secular growth story in NAND and SSD's. OCZ has the controller that MU doesn't, but MU has the ability to make its own NAND - I like both companies.
I would not recommend OCZ if you are risk averse, or at least not a large position. Then again I wouldn't recommend FIO either. It's a small, early stage, pre-profitability growth company that is still expanding rapidly and burning cash. Quite normal for a company at that stage, but still not for widows and orphans and all that stuff. Reminds me of Amazon years ago - Bezos regularly #$%$ off Wall St, even now, by crushing on revenue but showing little or no profit because he invests in growth. He even did this years ago when it wasn't so clear that Amazon would make it. That growth strategy can pay off - but not always. I like the fundamental business though, although I understand your concerns about cash burn. Pretty hard to raise much cash at the current valuation. I think they know that though and will avoid it.
Lastly, I expect consolidation in the space. I don't know if the STX rumor is true, and I don't know which companies will pair with which, but I expect consolidation as players w/ money but no growth and players with growth but no money pair off.
Sentiment: Strong Buy
i agree with most of what you say. however ryan is no good.
he can be ballzy when he has proven himself, being cocky now just gets the stock nailed. i haven't bothered to see if he always been like this or is it a new thing?
now this is what i would do to ocz. sell off one of the arms. enterprise or consumer, one should go. then quit competing with yourself. selling a agility 4 and a vertex 4 20 bucks apart is just plain stupid. the money you spent into R&D for one or the other would of closed up that $20 gap. then you can sell the better one for $20 cheaper.
then you get into controllers you see hitachi says they made one that is 4 times faster now? who knows if it is true but that don't bode well for companies like ocz who just entering the market.
"New SSD controllers are always big news, and it doesn’t get much bigger than this, as Hitachi has stated that their latest solution, using MLC flash, will offer sustained throughput four times greater than what’s available in existing products. This likely means that SSDs based on the new processor will be PCIe based in order to facilitate the extra bandwidth. This makes perfect sense, even in the wake of SAS 12GBps, which offers twice the speed of consumer oriented SATA, yet is still seen as a primary bottleneck when it comes to SSDs"
they built this with intel so?
like i say ocz best chance is to sell off one arm and focus on a niche market as that is where you will make it in the world of SSD's.
I think you guys have a tendency to dismiss Pipster out of hand because he's bearish on OCZ. However, at least on some of his points on this thread, I think he's correct.
SSD's do not make sense to me - right now, with today's economics and pricing structure - as backup media. They are more expensive per GB, and since backup media is just that - for backup - it is by definition not frequently accessed and does not need to be powered full time. The faster read speed provides little to no benefit (only faster restores). HDD's or even tapes are better, more economical choices for backup, and that's why today's data centers use all those technologies. This is not a bad thing for OCZ or SSD companies in general. Realistically, it will be hard for every HDD and SSD maker, combined, to ship enough storage to meet demand (per Seagate at the FMS). Storage needs are ramping so fast that companies cannot keep up. Storage in general, and SSD's even moreso, are in a massive secular growth cycle.
The netflix quote makes a lot of sense. The most common titles are streamed off SSD's. They gain economy when you are maximizing throughput. I think that the amount that goes on SSD's vs HDD's will continue to shift towards SSD's as price goes down and also if energy costs increase. I'm not sure how old that example is - things are rapidly changing. If OCZ meets their stated goal of 10GB/$ for SSD drives, that's a real game changer, and HDD's start to lose a lot of economy. Also pointed out is that they use commodity drives - not super expensive enterprise drives. Better to use software RAID 5/6 for redundancy and cheap drives with sophisticated software to manage the storage.
schrodinger's kitty - The enhanced durability technologies mean that that low number of cycles is no longer applicable. MLC drives have very long lives now, and TLC are coming along rapidly. The 100ish cycles is not even close to modern reality in anything that would be sold. As you noticed, such a drive would fail so often so as to be useless (and destructive even). Just ignore the number - it was only true in the very earliest stages when the technology was highly experimental.
Sentiment: Strong Buy
"netflix uses this server so not sure what you are spewing"
Well, they DON'T according to one of your earlier posts. Open Connect uses - wait for it - 8 (eight!) GB of RAM.
Is that the "TON" of RAM you were making up earlier.
Funny how calling you out on your made up stuff is suddenly "spewing".
Man this is totally awesome 8 GB IS NOW A TON OF RAM!!!!!
Let me repeat: 8 (EIGHT!) GB IS NOW A TON OF RAM!!!!
Unbelieeeeevable!! - ROFL LMA (t-o-t-a-l-ly) O - BWAAAAAAAAAAAAAAAAAAAAAAAAAAHAAAAAAAAHAAAAA!!!!!!!!
Very cool, the stuff you are making up - without equal!!!!
longocz isn't the amazons ec2 just a giant RAM warehouse?
i agree there will be more drives needed then can be built, but this also puts a floor on SSD. once they start catching on NAND will go back up in price. or you will have to build your own plant.
you ever wonder why i am so bearish here? ryan is one bad CEO. tbh they should of never bought a controller company this early. sure it compresses margins a bit, but also removes a ton of unneeded R&D they have no money for.
they should make less products, they should focus on one area to excel in and i think they be fine. but they trying to do everything at once when the whole market just sucks.
Well, whether they are max efficiency or not, replacing low energy efficiency hard drives for smaller numbers of higher efficiency SSD's would be equally beneficial. Sure, there may be lots of other ways to improve efficiency - no one is claiming that SSD's are the only way, but ideally, they can all be used together.
If data were rarely accessed and could sit in an off state, HDD or tape backup storage would seem a logical choice. That doesn't change the economy of storing the active data on SSD's - again, the two can, and should, coexist at current pricing structures. If SSD's get cheap enough then it might change, but for now, there are times when each is a lower cost choice.
I'm not sure on what basis you are claiming that SSD's "start getting errors the second you cut off main power." Probably not important from the point of view of the above discussion, as I think we both agree if you can turn the power off and only occasionally access the data, SSD's are not your ideal choice for storage. Actually tapes are probably still the cheapest for that.
Sentiment: Strong Buy
My understanding is that 100-500 is the raw endurance out of the factory for a single cell. When you layer on things like adaptive DSP, CellCare, memory modems, etc., it greatly increases endurance.
It's the same way they are able to take 3,000 p/e MLC and make it 50,000 or 100,000 for enterprise uses.
That's not a practical erase/write cycle limit. There are a few companies out there, including OCZ, that have been pioneering technology to enhance durability dramatically (I think there was some stuff on this in the FMS presentation notes). I think the other company is actually STEC - but not 100% sure.
Obviously that would not be practical for any drive. I remember when MLC was considered too unreliable for use, but it now completely dominates consumer level drives and is used in many enterprise drives. SLC really is just too expensive in most cases if you need high capacity - if we want to expand capacity, we need to increase data density and we need to control costs, and that means MLC, TLC, and whatever is after that. The nature of technology is to take things that don't work at first and engineer around the problems.
Sentiment: Strong Buy
even at those rates it is only good for approx 3-5 years. not sure how often you replace drives in a server farm but i know it can't be that soon.
the only bad thing about a SSD is they typically don't give a warning before they go bad, they just die.
btw you do know "memory modems" is a seagate program?