The Register® — Biting the hand that feeds IT

Comments on: Samsung swaggers in with 100GB SSD

Oh hurry up... 

Posted Wednesday 7th January 2009 13:07 GMT

Paris Hilton

and adopt these things already! The soooner the big-boys start adopting the technology the sooner the prices can come down enough to make them competitively-priced replacements for the consumer market..

I've been teetering on getting a SLC but the price/capacity argument rang true in my head, and i cannot for the life of me find the intel X-25M @ greater than 80GB available online...

Paris, because i know she appreciates a well-positioned fast hard drive...

You've confused me. 

Posted Wednesday 7th January 2009 13:21 GMT

Boffin

While that isn't difficult I was wondering about this paragraph.

"The idea is to replace four 15,000rpm hard drives with one of these 100GB SS805 SSDs which offer 25,000 random read IOPS and 6,000 random write IOPS. Samsung says that a 15K, 146GB, 3.5-inch HDD will cost $0.43 per IOPS, a 76GB 2.5-inch one $0.82/IOPS whilst its SS805 SSD costs $0.05/IOPS."

So you replace 4 * 146GB HDD with 1 * 100GB SSD. And where does the other 484GB go?

Double glazing analogy 

Posted Wednesday 7th January 2009 14:09 GMT

Can't possibly work. Having looked into this I found (1) there is no pay-back for double glazing in the life-time of the sealed units, (2) HMG used the AGW bilge to pass a law which in effect gave Pilkingtons a monopoly position in the supply of low-emissivity glass necessary when the front-back depth of units is constrained - feck um. Both of um.

OTOH these SSDs look tasty. Especially to non-carbon based life forms.

Huh? 

Posted Wednesday 7th January 2009 14:34 GMT

Stop

Sorry for the huh? title. I'm just a little bit confused by the math in the article.

They say replace 4 older drives with one new one?

Ok, so lets say your older drive is a 73GB 15K drive. How do you get a 100GB drive to equal 4 times the storage? Assuming that they were talking about 34GB drives, maybe the math might be close. (132GB vs 100 GB)

And they don't mention the cost or how long the drive is expected to last.

You can't really do a TCO analysis if you don't know the life span of the product, right?

Easy to wrap. 

Posted Wednesday 7th January 2009 14:49 GMT

Paris Hilton

"... compare kilowatt hours per year per 100,000 IOPS... Get your head around that."

It's easy to get your head around really. The trick is wrapping it right. See here, 2767 kilowatt hours per year is just 9,961,200 kilojoules per year. This is a little more than half the electricity consumed by the average household* in the same timeframe. Breaking it down a bit further, given 1 year is 31536000 seconds we get 0.31587 kilojoules per second per 100,000 IOPS. Now, _everyone_ knows that a kilojoule per second is a kilowatt which in turn is just 1000 watts. So it now reads 315.87 watts per 100,000 IOPS or a mere 3.16 milliwatts per IOPS**. There now, wasn't that easy?

Paris wants to know is it really IOPS or is that just the plural of IOP and is the "I" pronounced as "y" or is it plosive fricative?

*average of 4700 kilowatt hours per year from npower renewables

** of course this doesn't imply an actual power use per IOPS but more an average power draw based on the given an annual usage of 100,000 IOPS. That said, I don't want to hear someone jump up and say that it could use 99 kilojoules per IOPS and if each IOPS takes 50 milliseconds the power draw would be closer to 2000 watts per IOPS and nearly zero the rest of the time. Ok?

I/O not same as capacity 

Posted Wednesday 7th January 2009 15:14 GMT

When they say that it replaces 4 drives they don't count the capacity but the I/O only. It does happens that you end up with many drives just to get the right I/O number but only need a fraction of the space.

@You've confused me, and @Huh? 

Posted Wednesday 7th January 2009 15:15 GMT

Pirate

i do high end database and storage work with Symmetrix and other arrays, where capacity is often less important than IOPS. high-end apps are often very performance oriented in banking and similar sectors. Samsung is talking to people that work in storage, not to people that want a fast HDD to play games on, so their comparison is excellent. you are both right that they are comparing lower storage to higher storage, but it's a much higher IOPS solution, so for the applications they are addressing, capacity is not really relevant. drives like this are not for ye or me to have in our workstations ... yet ;) hopefully 2 years will bring all this goodness filtering down to us though :)

SSDs have limited rewrites! 

Posted Wednesday 7th January 2009 20:24 GMT

Thumb Down

Surely this discussion is just academic? SSDs have a much shorter lifespan than HDDs in terms of number of re-writes per sector. Servers typically have a lot more I/O than personal computers making SSDs unsuitable for this application.

What enterprise cares about lifetime? 

Posted Wednesday 7th January 2009 21:59 GMT

Flame

The typical enterprise lifetime is 3 years - the length of the product lease - then it's goes back to the lease company in exchange for new kit, then it's off to ebay.

The built-in wear leveling should extend the drive's useful lifetime far beyond that.

Aren't these drives built using SLC not MLC - that extends the lifetime too - or so it has been said.

We won't fully know until about 5 years from now if and when we start seeing unusual mortality rates on these types of drives.

With capacity increasing and prices going lower each day, consumers rarely hold onto computers more than 5 years - unless they hold onto it forever.

Who cares about capacity? 

Posted Wednesday 7th January 2009 23:21 GMT

Look disk space is free, you pay for disk performance. If you want to get better performance out of your disk drive you just use less of it. Using a whole 15K RPM disk gives you approx 200 IO/S if you only use a small part of it you can get 500 IO/S. If you care about the performance you use lots of mostly empty disks. So where does the extra 400GB go... answer you never used it in the first place in a desperate attempt to get an acceptable level of performance out of your drives.

The fact that these drives give better read performance is a also a big win, most systems do more reads than writes (why else bother to write the data, if no one is reading it) Disk arrays on the other hand give much better write performance because write behind caching usually works where as random read caching only works when your whole data base is smaller than your cache.

The sooner we can dump mechanical disks out of computers the better. We can finally leave Babbage behind.

In reply to 'You've confused me' 

Posted Thursday 8th January 2009 16:11 GMT

In reply to 'You've confused me' - the four drives being replaced will, I understand, be short-stroked ones and the used capacity will be less than the nominal capacity. Don't know what Samsung thinks the used capacity wil be but, if one 100GB SSD can replace 4 HDDs then the used capacity would be 25GB/HDD. That's my understanding and we all know that Samsung is putting its absolute best case forward.

Chris.

@Chris Mellor 

Posted Friday 9th January 2009 01:10 GMT

Paris Hilton

"Short stroked"

Fnar fnar

Don’t Miss

Amazon logo 75Amazon cloud heads for Asian sky

Steals MS thunder with Redmondian SDK

EMCData Domain-besotted EMC dumps Quantum

Not the girl for me, even after lending her $100m

Intel logo teaserBrace of Intel SSDs imminent

Low- or high-end as you prefer

AMD unmasks Opterons of servers future

Faces for 'Magny Cours' and 'San Marino'