Load AH on SSD drive?

[AKIron]

We're gonna hafta disagree on capacity/size being a limiting factor. Cost is the difference and only obstacle to a complete replacement today.

https://www.extremetech.com/extreme/233358-seagate-unveils-the-worlds-largest-ssd-60tb

 

[Chalenge]

Just a little paper on this topic:

https://www.usenix.org/legacy/events/fast12/tech/full_papers/Grupp.pdf

[Skuzzy]

We're gonna hafta disagree on capacity/size being a limiting factor. Cost is the difference and only obstacle to a complete replacement today.

https://www.extremetech.com/extreme/233358-seagate-unveils-the-worlds-largest-ssd-60tb

Did you read the article?  It is a "technology demonstration" and not for sale yet.  That aside, it is still not going to keep up with the advances in platter based storage.

There are two new technologies in the labs which will take the 3.5" hard drive up to 50PB.  The culmination of those efforts will start showing up in 3 years and be complete in 5 years.  During that period of time new technology designs will be starting in the labs.

Hard drive technology is nowhere near the end of its life and without a major breakthrough in chip design/manufacturing, there is no way SSD costs will ever catch up to the hard drive in the next 5 years.

There are still other mitigating issues with SSD's keeping them from the data centers.  The few Google does have in place are only used for static data as write performance/reliability, over time, is still an issue, but those are more difficult to quantify as it would depend on the application.

[zack1234]

In 10 years time we will have ports in the back of our heads and have AH direct to our brains.

No need for hard drives.

Image that Skuzzy exploring our ports

[AKIron]

Did you read the article?  It is a "technology demonstration" and not for sale yet.  That aside, it is still not going to keep up with the advances in platter based storage.

There are two new technologies in the labs which will take the 3.5" hard drive up to 50PB.  The culmination of those efforts will start showing up in 3 years and be complete in 5 years.  During that period of time new technology designs will be starting in the labs.

Hard drive technology is nowhere near the end of its life and without a major breakthrough in chip design/manufacturing, there is no way SSD costs will ever catch up to the hard drive in the next 5 years.

There are still other mitigating issues with SSD's keeping them from the data centers.  The few Google does have in place are only used for static data as write performance/reliability, over time, is still an issue, but those are more difficult to quantify as it would depend on the application.

Yes, I read it. It is a counter to the idea that there is an inherent limit to the capacity of SSD per volume that is more restrictive than HDD. There isn't and that is one proof. Have you ever seen a 60TB HDD in 3.5 inch form factor? I haven't. The speed difference between SSD and HDD is amazing if you have never seen it. Even the cheapest SATA  SSDs can make a big difference for an aging dekstop/laptop over HDD.

[AKIron]

Just a little paper on this topic:

https://www.usenix.org/legacy/events/fast12/tech/full_papers/Grupp.pdf

That looks like an old article, 2012? Hardly relevant today.

[AKIron]

If the desktop/laptop market shifted from HDD to SSD (inevitable imo and soon) what do you suppose would happen to the market for HDD? Those high end plattered drives for data centers will go up in cost, maybe dramatically.

[save]

Laptops already did to a large extent, remember moving / dropping a laptop with a spinning harddisk in it can be bad news, even with new technology.
For data centers there is simply no replacement in sight.

I have both SSD and hardisks in my game system, of my laptops 4 out of 5 are SSD (3 of them work laptops for different types of work).

[AKIron]

A mostly unrelated anecdote. In the mid 90's I installed a couple of new HP servers for a private school down your way Skuzzy, St Marks of Texas. Had trouble with the HP branded drives, think they were 4GB. So I'm on the phone with HP support who had me update the firmware on the drives to a new release which was supposed to solve the problem. The conversation with the tech got kind of chatty and I asked him why the drives cost about 4 times the desktop equivalent. Of course I knew he was just a tech but he said the cost is due to all the testing HP does for the specific hardware. I laughed and said, you mean like these drives we're dealing with right now. He chuckled, I think he saw the irony.

[Chalenge]

That looks like an old article, 2012? Hardly relevant today.

Yeah, because the physics behind electronics changes every day. (sarcasm off).

[AKIron]

While our understanding of physics does change that's not my objection to an outdated report. Technology changes rapidly. We'll revisit this in 2020.

[Chalenge]

There are more reasons to consider the platter HDD approach to have a longer lifespan than just track size. Cabling, controllers, and even internal array structures. The problem with SSDs is that miniaturizing flash memory itself is limited by design. Both HDD and SSDs have temperature limitations. SSDs actually use more power than HDDs (an OCZ Agility 3 SSD uses 1.5W at idle and 2.7W under load, whereas a 1TB WD Scorpio Blue HDD uses only 1.4W under load and a mere 0.6W at idle!). Even if you look at the most economical SSD available it will still use more power than an HDD of any size. I realize that most people look at the speed of the SSD and think that's all they need to know. Well, except if some of the early hybrid projects hit the market the HDD will literally leave the SSD in the dust! Even if that doesn't happen, would you really want a 4TB drive, when for the same price you can buy a 100 TB HDD? a 1 PB HDD? That day is coming.

[AKIron]

According to Google's testing, based on no longer current technology, SSD has a better replacement rate than HDD. I know that doesn't necessarily mean a longer potential lifespan.

Yes, I will trade capacity and price for performance.

[100Coogn]

I'm waiting for a 1 Geopbyte hdd, then I will be all set. (that's 1000 Brontobytes to you and me)

Coogan

[Skuzzy]

According to Google's testing, based on no longer current technology, SSD has a better replacement rate than HDD. I know that doesn't necessarily mean a longer potential lifespan.

Yes, I will trade capacity and price for performance.

Lifespan and performance become secondary concerns when you have to buy 300,000 drives for a single data center.  The last numbers I saw from Google was they were collecting 15PB/day of data.

Google is advocating for a smaller form factor hard drive.  They would like to see 8TB 2" hard drives.  They would rather replace smaller drives, than larger drives.  Makes sense when you think about it.