Investigation: Is Your SSD More Reliable Than A Hard Drive?

[Guest]

I just wanted to say thank you SO much for your hard work on this great, well-researched article.

There's so much stuff being thrown around about SSDs versus HDs from people in both camps, that I think this kind of article is really very crucial. People are saying some stuff about SSDs you'd think they can cure cancer.

It's great technology but we should use it with the usual cautions.

You've made me realize that instead of relying completely on one or the other, I should maybe use both, and ensure that I have multiple copies of data for redundancy. It's a common sense solution but I didn't realize it until reading this article.

 

[Guest]

I monitored quality at Compaq many years ago and had access to proprietary data across their entire enterprise product line. Found the spread between the best and worst product was typically a factor of 4. For instance if the best hard drive had a 2% failure rate, the worst would have 8%. Customers also figured out which products are the most reliable and continue to create product demand over a much longer period then similar products; and the highest quality product typically had double the product life of the average or poor quality products.

Hard drive quality is hit-and-miss. Even a vendor like Seagate or WD will have a great drive and an unreliable one the same year, just different models. Sometimes a vendor will have 2 great years in a row then will have poor quality for the next 2 years.

The only reliable way I know is to browse the Newegg customer product reviews. Look at the overall egg score first and compare the detailed graph (not just what Newegg lists) focusing on the 1 and 5 egg scores. 1 egg typically means the drive failed or was DOA. 4 to 5 eggs means the product is good. Also a product really needs to have 50+ reviews before it is statistically valid and it is best to wait at least 3 months after a product is introduced before purchasing. This method has served me well over the past 12 years.

 

[bbartlow]

I can confirm what others have said: SSD's don't always like to sleep and wake up on the right side of the bit. I can live with that. The benefits of sheer speed far outweigh those quirks, imo, because speed translates into maintaining uninterrupted work flow. By comparison, HDD's can be just a little...well... well...(listen to it spin, access, spin, access) ...irritating! Even the slight difference in reliability is of little value to me, either way.

Coming in a close second place (to speed) is the huge reduction in noise. Love it. Except for large storage needs I will never go back. Not at these prices.

 

[Guest]

Loved the article, but not sure I agree. If an SSD is 15x faster (ex: MySQL queries per second), then for every year of life, it does 15 years of HDD work. Or, for a year, it does the work that would take 15 HDDs, and the probability to have a failure on at least one of those 15 drives would be a lot larger than the probability of failure of that single SSD.

I think the failure rate should not be time dependent only, but should also take into consideration the amount of data that flows through it before failing.

 

[bad_student]

a useful article. I'm definitely leaning towards a SSD for a future computer.

 

[Guest]

How about the reliability of 2.5" spinning disk?

 

[nov8]

Win 8 seems to be more geared for SSD with a boot tech that uses a half sleep mode. Maybe, just maybe software tweaks that can handle the SSD will make them more practical. Using defrag technology or swap files designed for linear read/write disks may be killing the SSD. Custom mode software in the OS may just prevent failures and enhance reliability of SSD.

 

[devBunny]

This article could do with an update. Any chance, Tom's?

 

[redtails]

This article has been written a couple months back and it's still a very interesting read as many of the same principles still apply. I can see that there is not yet enough data on SSDs to find statistical differences, though the multitude of data on mechanical spinners is very interesting!

However, I strongly disagree with the conclusions of this article concerning the reliability of SSDs being equal to those of HDDs. In a server environment, where technology is maintained properly, the environment is extremely steady and everyone knows exactly what they're doing, it is very difficult to find statistical differences.

Now take SSDs into the consumer world; laptops that are stuffed in the back packs of children while they're jumping around, camcorders that are taken on hiking tours through mountains, tablets that are handled roughly by users in public transportation.. I do not have actual statistical figures, but from empirical evidence I can tell that mechanical harddrives do NOT live beyond 2-3 years of life in a portable device due to their mechanical nature. Write endurance is completely irrelevant, SSDs last longer in consumer devices because they can take a beating. I've had so many laptop harddrives fail on me it's just not fun anymore. I'm sick and tired of having to buy new harddrives so I will never do it again. Yes, I am biased towards SSDs and yes I strongly believe (and hope) that harddrives will be gone for once and for all within 5 years.

Furthermore, the mistake is often made that SSDs may be just as reliable as harddrives in a server environment, but Tom's Hardware has made this mistake in the past! If you compare 1 year of HDD usage to 1 year of SSD usage, you'll see that an SSD can do the same task in less than half the time. In a truly controlled environment, you'd want to keep the number of variables as low as possible. So if you're using an SSD in a cache server and it's working near 100% of its maximum performance, well duh it'll fail earlier than a harddrive working at 100% of its maximum performance, the data cannot be compared

 

[abm]

This article is 1.5 years old. Is there any significant improvement in the last 18 months? Like Intel's S3700 http://www.anandtech.com/show/6433/intel-ssd-dc-s3700-200gb-review/2 (released last month) claiming the consistent performance and that it won't vary its IOPS by more than 10 - 15%, and that 99.9% of all 4KB random IOs are serviced in less than 500µs.

 

[Guest]

My Corsair SSD fried itself in less than a month of operation. I'm sure it was due to cheap electronic components (capacitors). Although it worked incredibly well up to that point, I'm no longer evangelizing SS. I've NEVER had a platter drive fry in that short of a period.

 

[tonyhall]

This article is OLD and outdated and does not apply anymore . It should be removed.

 

[Sparafucile]

Your comparison of speed of SSDs versus HDDs comes to a numerical conclusion that misrepresents the actual data you present (in graphs). You claimed that "low-end SSDs are only 85% faster than HDDs", while showing graphs that give busy times of 500sec versus 3500sec. That data indicates that those SSDs are not 85% faster, but 7x, or 700% faster. I'm sure I don't need to elaborate further.