When is the Storage Industry going to "Get with it?"

[ahthurungnone]

As a computer enthusiast I am very disappointed that no matter how fast our cpu, memory, video cards, etc are, we still must settle for 90's Hard Drive speeds. Let's be honest, how long will we continue to spend hundreds of dollars upgrading our computers when the best the Storage Industry provides us are Pentium 4 speed hard drives (no matter how many RPMs they run) or ridiculously priced SSD's that have pathetic capacity?

I personally don't care what the issues are with the engineering problems. But if I was one of these Storage Manufacturers I would simply fire the entire engineering staff and bring in some fresh thinkers who could conjure up something better than either 1) spin the hard drive faster or 2) make bigger flash drives. Even these SSDs have their own speed barriers which are already being trampled by Intel's latest cpu's. So what happens when the average processor is running data twice as fast as it is now? Development in every other field will stall until someone else jumps into the Storage Industry to compensate for the present lack of creativity and ingenuity.

Ok, now I feel better. If venting helps you, feel free to reply.

Oh, and please don't defend the Storage Industry unless you have some breaking news of SSDs with 500GB under $150 or HDs with access and write times at no more than twice that of an SSD.

 

[sub mesa]

I can just add that you have a lot of learning to do. Not meant as an insult, but rather encouragement to inform yourself better so you understand the patterns that currently dominate storage technology. The breakthrough of NAND, the starting death of mechanical storage, the current era where we use both NAND and mechanical storage since neither has exclusively what we want, etc.

I'm not sure i'm willing to explain all this. But i've given you some pointers on which you could do research. Just using google.

Cheers,
-sub

 

[hang-the-9]

Why aren't you complaining that RAM is not as fast as the CPU? You don't think that they aren't falling over each other trying to make the fastest drive they can? And complaining that SSDs which is maybe a two year old consumer technology is not giving you huge storage for cheap is a bit narrow minded. You ever hear of phisical limitations to hardware? You can't have a huge amount of data stored in a medium that is hugely fast like a CPU or even RAM and not have some technical issues to overcome. Or are you looking for holographic storage this week?

You must have been one of the people that said the car is useless as a horse could have outrun it in the 1800s. Why this car thing, it's noisy and not as fast as a horse, sure it only came out this year but they may as well close up shop, it will never work.

I personally think that the military should get with it and invent a missle forcefield, cause you know those missles travel fast and hurt things.

 

[ahthurungnone]

Firstly, NAND has been around since 1989 (yea, I did your required research) so I wouldn't consider this a breakthru technology.

Secondly, I do agree that hybrid drives offer the best long term solution.

Finally, the problem with non-physical storage is that it is irretrievable once it loses electric current.

 

[ahthurungnone]

Hang,

Actually RAM does run as fast or faster than most current CPUs. You are making an error of equivocation when you measure "overall speed" with a hertz value. Is it true that RAM runs at 1.6Ghz whereas a CPU runs at 4.0Ghz? Yes, but the CPU must do much more work than the RAM to complete its task and thereby requires more "speed" to complete its task. Thus, even though the CPU has more speed it still completes its task in the same amount of time (i.e. overall speed or "reality").

What I fail to grasp is how you would rather divvy out "ad hominems" than address the impending problem with storage. Ad hominems only demonstrate that one does not possess a true argument but hopes to silence others (and the truth) through intimidation and the misappropriation of labels.

But, I always enjoy the company of a "debater", so I will humor you one time with the argument that I "would have been one to criticize the invention of automobiles."

Your contrived theory of my argument is as follows...

1) Automobiles are too noisy. (Storage is too limited.)
2) Automobiles are too slow. (Storage is too slow.)
3) Therefore, automobiles (storage) should not be made.

My true argument...

1) Automobiles are too noisy. (Storage is too limited.)
2) Automobiles are too slow. (Storage is too slow.)
3) Therefore, automobiles must become less noisy and faster! (storage must become increased and faster.)

If you actually read what I was saying is that we shouldn't abandon storage but that progress in the Storage Industry is lagging far behind the others. Consequently, any business owner would want this problem rectified even if that means "upsetting the apple cart" which is why something dramatic must be done to "wake them up."

Or we could be like most people and just accept the status quo and keep making and driving our Model T's because anyone who questions that anything better than a Model T could be made is crazy and just being negative and critical. (P.S. They don't make Model Ts anymore because Henry Ford was NOT satisfied with the status quo.)

 

[hang-the-9]

Firstly, NAND has been around since 1989 (yea, I did your required research) so I wouldn't consider this a breakthru technology.

Secondly, I do agree that hybrid drives offer the best long term solution.

Finally, the problem with non-physical storage is that it is irretrievable once it loses electric current.

Consumer SSDs were around since 2008, to me , that's pretty new. Just because some company made a 1 kb chip in a lab at some year does not mean x number of years from now we will be getting it free in our breakfeast cerial. It could be x + 20 years.

I figure in 3-4 years you'll see a 500gb SSD for $200. Maybe.

You can get all the drive speed you want, if you pay enough for new and/or expensive tech. There is a saying amongst probably every industry, you can have it Fast, Good or Cheap, pick 2. Althought in this case, Fast and Cheap really applies to how fast you can buy it, not how fast the product is. You want a fast disk with large storage, bring your $1,000 to the table. Or wait a few years.

You picked an odd time to complain just when everyone is introducing hard drives that are way faster than they ever were. All you are really saying is that SSDs are expensive in large capacity.

And RAM modules have a latency about 200+ times a CPU cycle. http://duartes.org/gustavo/blog/post/what-your-computer-does-while-you-wait

 

[ahthurungnone]

I guess I will have to wait.

 

[sminlal]

> 1) spin the hard drive faster

You can buy 10,000 and even 15,000 RPM drives if you're willing to pay the money. But access times are still going to be multi-milliseconds because fundamentally you still have to wait for platter rotation and access arm movement. It's like complaining about how years of engineering haven't made bicycles very much faster than they were 100 years ago.


> 2) make bigger flash drives.

You can buy solid state storage in capacities as large as you want - if you're willing to pay for it. For example, Nimbus has a 2.5TB array of solid state drives for $25,000: http://www.enterprisestorageforum.com/technology/news/article.php/3878441/Nimbus-Unveils-Nifty-Solid-State-Drive-Storage-Array.htm

SSD costs will come down as chip fabrication facilities scale the technology. That's been going on for many decades now - it's not as though they haven't been making progress on it.

 

[sminlal]

Actually RAM does run as fast or faster than most current CPUs. You are making an error of equivocation when you measure "overall speed" with a hertz value. Is it true that RAM runs at 1.6Ghz whereas a CPU runs at 4.0Ghz? Yes, but the CPU must do much more work than the RAM to complete its task and thereby requires more "speed" to complete its task.

The CPU really is a LOT faster than DRAM memory. Typical DRAM memory takes about 10ns to deliver requested data to the CPU chip - in that time a stock Core i7 920 CPU can execute about 40 instructions. And since EVERY instruction needs to be read from memory BEFORE the CPU can execute it, that's a serious problem.

DRAM memory simply can't keep up with the CPU, even with all it's optimizations for sequential access. The only way CPUs can cope is to use not one, not two, but THREE levels of very high-speed on-chip cache to reduce the need to access DRAM memory as much as possible. Fortunately these caches are very effective, particularly for program instructions (which are usually accessed sequentially and repeatedly). So for run-of-the-mill programs, the caches minimize the demands on memory and the CPUs are mostly free to execute instructions as fast as they can.

But there are certain classes of programs (image processing, for example) for which the cache just can't solve the problem. Those kinds of programs run a lot slower than they otherwise could because of the memory bottleneck. And the problem is worse than you'd expect because those kinds of programs typically use the CPU's SIMD hardware to process multiple data items in parallel with each instruction, so they need even more data bandwidth.

So saying that "RAM runs as fast as current CPUs" is kind of like saying "Disks run as fast as memory". You can use DRAM memory caching to make disks LOOK faster (just as on-chip static RAM caches make DRAM LOOK faster), but when push comes to shove and a program actually has to fetch and wait for data that isn't cached then the CPU will sit there twiddling it's thumbs until the slower device gets around to delivering it.

 

[MRFS]

If you are willing to do some "manual" data management,
there are certain things you can do right now that will
accelerate storage across the board.

First of all, in order to implement an effective data management
strategy, a careful and realistic distinction needs to be made
between "most recently used" data and "least recently used" data
-- and possibly a third data set in between those 2.

Modern computers already make decisions like that,
in the absence of any direct intervention by the user:
demand paging comes to mind, but there are other examples.

A well designed ramdisk can do wonders: we have moved
all of our browser caches into ramdisks managed by
RamDisk Plus, developed by SuperSpeed LLC:

http://www.supremelaw.org/systems/superspeed/RamDiskPlus.Review.htm

Short-stroking the OS partition on a rotating platter
will reduce armature wear, minimize access times,
and exploit the fact that there is always more data
on the outermost tracks: this is so because modern
HDDs do maintain at or near the same recording density
from outermost to innermost tracks.

Moving the swap file e.g. pagefile.sys on Windows systems,
to a short-stroked partition, created with the Contig freeware
so as to be perfectly contiguous internally, is another way to
harness parallelism when a system must do concurrent I/O
to/from data files at or near the same time it is doing I/O
to/from the swap file.

Now, at the other extreme, we find lots of examples like
image files of the OS partition: these tend to be large,
sequential files that will probably not be READ very often,
if ever, relatively speaking.

Although there may be situations in which the WRITE speed
when creating a drive image is very important, in our experience
this task tends to be done best as a background / batch task
that can finish when it finishes: its sheer speed is a lot less
important that the integrity of each drive image created by that task.


The lesson here is to learn as much as you can about the
kinds of storage your system(s) do, and allocate files
in an intelligent fashion so that the most frequently used files
are stored on the fastest available storage, while less frequently
used files are stored in slower storage.

The predictable benefits of such an intelligent allocation
are measurable, and substantial.


MRFS

 

[ahthurungnone]

Best answer selected by ahthurungnone.

 

[xxsk8er101xx]

There's little demand for speedy hard drives for the home user. If you want to complain about cost look at what it costs for 15k SAS drives that businesses purchase. They're more than an SSD.

 

[sub mesa]

Firstly, NAND has been around since 1989 (yea, I did your required research) so I wouldn't consider this a breakthru technology.

The breakthrough is in price: