September 11, 2014 at 03:37Regarding the appearance of 8 and 10TB hard drives
News: deliveries of 8TB of hard drives have begun, 10TB of hard drives with a “tiled” recording system ( shingled magnetic recording ) have been announced , which promises an increase in volume by reducing the speed of random access ( details ).
SSD almost completely gobbled up everything in the range of tens / hundreds of gigabytes. The HDD is used only as an ultra-cheap plug, and I don’t understand why not a single dodgy Chinese guessed to make an SD card reader with an mSATA interface. For the low-end (just to boot), this is enough for the eyes and ears, and it will cost less. If the person is "slow", then it is obvious that the SSD will save him. As you move for 300-400GB, a picture becomes economically unattractive for many segments (1Tb costs € 50 per HDD versus € 350 for the cheapest SSD). But everyone understands that access to a budget terabyte for SSD is a matter of time. Although the demand for a place is growing (games / movies are getting more and more resolutions), at the same time, the area of needs is shrinking (why download, when you can watch online), that is, the desktop market can be considered lost.
But, besides him, there are other markets.
HDD, in addition to low-end, continue to live:
Here is the last factor that determines a unique niche for HDD - they are better suited for linear recording. This is much worse with SSDs - if you constantly write in large blocks on SSDs, then housekeeeping ceases to cope (especially if the record is circular inside a file, such as video surveillance servers), the cache gets clogged, and the SSD degrades to WD Green levels or even worse . At the same time, they are more expensive, their write-offs are multiplied by write amplification (when a disk is full of data, SSDs have to move several blocks to write one), the cheaper the SSD, the worse its resource (and unlike the household “oh , my SSD is wearing out "the potential wear and tear of the SSD on the surveillance server is a completely objective question).
If the client does not care about seek time for reading (which is also linear), then we get a net loss of SSD. Note that SSDs are very difficult to analyze and design, with an extremely uneven curve. If you run the IO test on the HDD, then its graph (in time) will look like a perfect straight line, albeit not a high one, which barely trembles. For SSDs, this is always a jitter from “very bad” to “awesome” - and this complicates the math or gives “weird sticks” at high loads.
Thus, the target audience of expensive discs is customers who need very cheap storage (in terms of $ / Tb) mostly linear writing. Logs, backup backup systems, videos, archives, lower tier for multi-tier storage. The key to market success here is the combination of a low terabyte price along with a high copy price. Note that shingled magnetic recording finally removes the issue of performance in random access. But, in the light of the target audience, this is not a special problem (as long as random access can be done in a reasonable amount of time - file systems still want to read from strange places sometimes).
And this market has quite bright prospects for itself. There will be more and more data, you will not save all of them on the SSD - the need for archive storage with random access will grow. Industrial standards are well developed, the infrastructure is ready and generally accepted, there is backward compatibility. Current trends show that the HDD is still managing to grow in size, and after a strange “slowdown” in the 1-2TB region, it is quite likely that Moore’s debts are catching up. Skeptics may say that the HDD will soon run into quantum problems and this will stop, but the SSD is not doing so well - the more levels the SSD cells have, the larger the block size for rewriting, that is, less resource. But the capacity grows like log 2(N) then the number of levels ... And there are quantum effects there too (perhaps you should start introducing the concept of "bit by mole", or even "atoms per bit"). In other words, the race comes with obstacles on both sides.
In light of this, the thorny future can be roughly described: Hard drives from the “essential part” for the entire computer industry are turning into a completely niche product. Which is not going to disappear anywhere. Just as streamers do not disappear (how many streamers do you have at home?), Nothing threatens the life of hard drives until SSDs can offer such capacities for such money with such performance indicators.
A comment
SSD almost completely gobbled up everything in the range of tens / hundreds of gigabytes. The HDD is used only as an ultra-cheap plug, and I don’t understand why not a single dodgy Chinese guessed to make an SD card reader with an mSATA interface. For the low-end (just to boot), this is enough for the eyes and ears, and it will cost less. If the person is "slow", then it is obvious that the SSD will save him. As you move for 300-400GB, a picture becomes economically unattractive for many segments (1Tb costs € 50 per HDD versus € 350 for the cheapest SSD). But everyone understands that access to a budget terabyte for SSD is a matter of time. Although the demand for a place is growing (games / movies are getting more and more resolutions), at the same time, the area of needs is shrinking (why download, when you can watch online), that is, the desktop market can be considered lost.
But, besides him, there are other markets.
HDD, in addition to low-end, continue to live:
- Because of the completely crazy fleet of servers that live a lot more than drives. Disks need to be changed.
- Due to gigantic volumes for reasonable money
- Due to better performance on established line recording.
Here is the last factor that determines a unique niche for HDD - they are better suited for linear recording. This is much worse with SSDs - if you constantly write in large blocks on SSDs, then housekeeeping ceases to cope (especially if the record is circular inside a file, such as video surveillance servers), the cache gets clogged, and the SSD degrades to WD Green levels or even worse . At the same time, they are more expensive, their write-offs are multiplied by write amplification (when a disk is full of data, SSDs have to move several blocks to write one), the cheaper the SSD, the worse its resource (and unlike the household “oh , my SSD is wearing out "the potential wear and tear of the SSD on the surveillance server is a completely objective question).
If the client does not care about seek time for reading (which is also linear), then we get a net loss of SSD. Note that SSDs are very difficult to analyze and design, with an extremely uneven curve. If you run the IO test on the HDD, then its graph (in time) will look like a perfect straight line, albeit not a high one, which barely trembles. For SSDs, this is always a jitter from “very bad” to “awesome” - and this complicates the math or gives “weird sticks” at high loads.
Thus, the target audience of expensive discs is customers who need very cheap storage (in terms of $ / Tb) mostly linear writing. Logs, backup backup systems, videos, archives, lower tier for multi-tier storage. The key to market success here is the combination of a low terabyte price along with a high copy price. Note that shingled magnetic recording finally removes the issue of performance in random access. But, in the light of the target audience, this is not a special problem (as long as random access can be done in a reasonable amount of time - file systems still want to read from strange places sometimes).
And this market has quite bright prospects for itself. There will be more and more data, you will not save all of them on the SSD - the need for archive storage with random access will grow. Industrial standards are well developed, the infrastructure is ready and generally accepted, there is backward compatibility. Current trends show that the HDD is still managing to grow in size, and after a strange “slowdown” in the 1-2TB region, it is quite likely that Moore’s debts are catching up. Skeptics may say that the HDD will soon run into quantum problems and this will stop, but the SSD is not doing so well - the more levels the SSD cells have, the larger the block size for rewriting, that is, less resource. But the capacity grows like log 2(N) then the number of levels ... And there are quantum effects there too (perhaps you should start introducing the concept of "bit by mole", or even "atoms per bit"). In other words, the race comes with obstacles on both sides.
In light of this, the thorny future can be roughly described: Hard drives from the “essential part” for the entire computer industry are turning into a completely niche product. Which is not going to disappear anywhere. Just as streamers do not disappear (how many streamers do you have at home?), Nothing threatens the life of hard drives until SSDs can offer such capacities for such money with such performance indicators.