The Evolution of Disk Storage and an Introduction to NVMe

https://www.linuxjournal.com/content/data-flash-part-i-evolution-disk-storage-and-introduction-nvme

by workrockin • 7 years ago

92 36 7 years ago

linuxjournal.com

3xblah• 7 years ago

Will the expression "spin up" ever fall out of usage?

NikolaeVarius• 7 years ago

Doubt it, though I don't think that I've used it in the context of SDD's recently since latency is that low.

I think colloquially, it's pretty much understood to mean "time between turning on and being useful"

modzu• 7 years ago

language is symbolic so there doesn't have to be a literal connection between the reference and referent. for example you still "boot" a computer, which comes from the concept of "bootstrapping", ie, pulling oneself up by the straps of their boots. no boots are involved.

bluedino• 7 years ago

We 'spin up' virtual machines or instances now

NullPrefix• 7 years ago

Because once you hit the launch button your mouse cursors starts spinning.

jyrkesh• 7 years ago

Contrary to what others said, it will likely fall out of favor the way the floppy icon faded for "Save" functionality in GUIs. It's everywhere until one day you realize you can't find it in any of the apps you use every day.

SketchySeaBeast• 7 years ago

Doing a quick survey of the apps on my machine, it seems like most have just forgone the button altogether. For those that still have the button, the floppy icon seems to still be there.

vbezhenar• 7 years ago

Latest Word: it's there; paint brush: check; paint.net: check; editpad: floppy. Even Intellij Idea uses floppy. Actually from a quick glance I did not find a single application which did not use floppy icon as a save pictogram.

May be it's Mac world? Apple is eager to forget old ways.

tyingq• 7 years ago

It was in wide use for jet and turbine engines, centrifuges, gyroscopes, etc, before disk drives. I don't think it will fade away.

niccl• 7 years ago

At least one of the D's in the acronym usually means Disk, and there isn't a disk in many of these. Will that ever go, do you think?

spectramax• 7 years ago

I am surprised the article doesn't talk about Optane technology: https://en.wikipedia.org/wiki/3D_XPoint

Nursie• 7 years ago

Is Optane really 'all that' compared to other PCIe storage?

I know that Intel are positioning it as some sort of system accelerator, and I know it does tend to score well in comparisons... but not "This is in a completely different category" well, AFAICT.

What's special about it compared to, for instance, Samsung's NVMe offerings?

sithadmin• 7 years ago

Optane tends to crush the competition in terms of latency and random I/O, and mixed workload benchmarking. Samsung Z-NAND performs just slightly better on sequential I/O.

synack• 7 years ago

Optane is persistent and bit-addressable, with mean latency under 1 microsecond, which is an order of magnitude faster than other SSDs.

I don't think the market has figured out the right use case for Optane yet. The majority of desktop applications won't benefit from lower latency IOPS and it's too expensive to use for general purpose storage on servers. It does make sense for constant write or seek-heavy applications like database journals, but most databases are optimized for doing bulk sequential reads/writes and won't take advantage of Optane's bit addressable storage.

Intel's recently started shipping Optane DIMM modules that act like slow, cheap, high density RAM. This is an interesting option as it allows in-memory databases to be atomically persistent without having to add any code.

chapium• 7 years ago

Why is Optane storage always referred to as "Optane technology" (big announcer voice)?

ip26• 7 years ago

They have a big marketing team. "Now with XYZ Technology" is standard marketing for jazzing up something either not actually all that radical, or too technical to explain.

edoo• 7 years ago

I use LVM and have enough storage that all SSD would be a little expensive so for the last few years I've been using the block caching LVM feature with mirrored NVME drives as a writeback cache to my large HDD array. It greatly increases the performance of the storage with little hassle.

If haven't played with gluster for a few years but it has a tiered storage feature so the most recently used files actually get moved onto the high speed storage medium, which depending on your workloads might be even better, assuming gluster is now performent and stable enough.

brianpaul• 7 years ago

FWIW, I believe Cray had an SSD option for the X-MP back in the early 80s.

theideaofcoffee• 7 years ago

They sure did! It came in 256-, 512- and 1024-MB flavors, if i recall, in its own cabinet off to the side at a blazing 1000MB/s.

I don’t have a hunk of the SSD, but I have a (30lb) board of the central memory perched against my desk right now.

wolf550e• 7 years ago

From the brochure, it had a few million of 64-bit words worth of RAM that was used as regular RAM ("central memory") and a lot more of ECC DRAM that was used as fast storage ("SSD"). What was the API for the SSD? Was it mapped into virtual memory as RAM, or did the ISA have I/O in,out opcodes for it with word address, or did it use the API for disk storage, or what?

mechagodzilla• 7 years ago

Even the Cray-1S supported it! I believe it was connected via a 128-bit wide bus operating at the system clock speed (80 MHz in the case of the Cray-1), which made it a ~10 gbit interface. From a low-level software perspective, it was connected to a pair of DMA "channels" (1 input / 1 output), and there were instructions to do bulk copies from SRAM->CHANNEL or CHANNEL->SRAM. It may have used another slower channel pair as a command interface, but I'm not sure (nor how COS or UNICOS exposed it from a higher level software perspective).

jcoffland• 7 years ago

Some really early computers stored bits in waves in mercury. A series of waves would be started at one end of a tube full of mercury and they would be read back when they arrived at the other end. The memory of course then had to be refreshed. Core memory was however much more successful.

https://en.m.wikipedia.org/wiki/Delay_line_memory

agumonkey2• 7 years ago

Same idea was used between satellite and ground, wave was used as a memory that could be reused when back to station.

winrid2• 7 years ago

Holy crap that's so cool. Thanks for that info.

Avery3R• 7 years ago

An example of delay line memory using a wire that vibrates: https://www.youtube.com/watch?v=2BIx2x-Q2fE

sliken2• 7 years ago

Register latency is usually around 4ns (2.5-3.5 Ghz with a pipeline depth of 12-16). The article claims main memory latency is 7x more, which is around 24ns to 45ns. I tried to come up with plausible numbers that fix the 7x ratio they mention.

These days main memory latencies are more around 70-80ns (single sockets) and 80-100 ns (dual sockets).

Being off by so much does make me wonder about the accuracy of the rest of the article.

sliken2• 7 years ago

My dad had a 256bit memory module on his desk. Made out of sandwich of two pieces of glass and 16 vertical and 12 horizontal wires. Each intersection had a small ferrous ring.

Around that time he was programming and early computer and you could actually hear the memory bits changing. He was excited listening to it, figuring each click was a loop. Then her heard a different sound and realized he as hearing instructions and not loops.

ntw1103• 7 years ago

I recently upgraded my laptop to a Thinkpad T470s, which supports M.2 NVMe, and M.2 Sata. I upgraded from a 500gb NVMe drive to a 1TB M.2 sata drive. Why do I say upgraded if it is slower? Well, in a laptop, power usage is hugely important, as is heat. To prevent throttling on the CPU, less heat is better. I did a lot of research, and found that the sata based M.2 drives consume less power, and generate less heat. For my purposes the sata speed is sufficient. :)

The Evolution of Disk Storage and an Introduction to NVMe

Related Stories