Aces High Bulletin Board
General Forums => Hardware and Software => Topic started by: Chalenge on September 08, 2008, 12:16:31 PM
-
Skuzzy now that Intel has shipped SSDs I have to know if this would give a system boost over traditional hard drives. From the sound of things it will be a wonderful thing but we all know how ideas like this can go sour. Do you think there could be a problem in drivers or OSs that may present a major problem or rather then hurry to try one out should I just adopt the old wait-and-see attitude?
-
SSDs have been out for a while now. I've had a Samsung 32Gb for almost 6 months...
Read Average: 97.6MB/sec
Read Access Time: 0.5ms
Write Average: 36.6MB/sec
Write Access Time: 0.5ms
You decide. :)
Edit:Windows loads in about 5 seconds. That's from after Post to Desktop.:)
-
I thought the new Intel types were supposed to work directly with Intel chipsets? and speed things up even more? Feel free to inform me if thats not correct.
-
I am still on the fence about SSD's. They still seem to have a limit on the number of writes before they start failing. Of course, it is a huge number, but it is one of the stealth kind of failures. You have no idea it is about to happen until it does.
Most HD's give you a mechanical sense of a problem coming. Not always, but most of the time.
Do not get me wrong. SSD's have come a long way in the last couple of years. There is a new technology on the horizon which will make SSD's the drive of choice. I am waiting for it to hit the streets. It is based on a new type of memory module design.
-
Thanks Skuzzy 'wait-and-see' mode active.
-
I thought SSDs were relatively slow compared to a modern SATA drive (and not just slow, more expensive, significantly less space per $)?
-
SSD's are much, MUCH faster as there is no mechanical process to slow it down. Basically it works at the speed of memory.
EXTERNAL SSD's... such as USB controlled external drives tend to be quite a bit slower as their speed is limited due to the USB bandwidth.
I'm not much familiar with these drives (I think I want one now), but.. and correct me if I'm wrong.. but I'm assuming it installs internally.. possible through the SATA connection?
-
SSD's are much, MUCH faster as there is no mechanical process to slow it down. Basically it works at the speed of memory.
Flash based memory is slower on large writes/erases than HDD's though.
-
Point being to install the OS on an SSD and all programs on a typical HD. I think you can see the advantages there. Once the design Skuzzy mentioned is available and affordable then HDs might be a thing of the past. Might!
-
Point being to install the OS on an SSD and all programs on a typical HD. I think you can see the advantages there. Once the design Skuzzy mentioned is available and affordable then HDs might be a thing of the past. Might!
Yep, install OS and anything that doesn't write to the drive on the SSD. Make sure you put all writeables on an HD. That eliminates or limits the write life problem Skuzzy mentioned.
And right now they are limited in size and unlimited in price. So be prepared to whip out some cash.
-
In the past, when devices were a few megabytes and write limits were on the order of a few to at most a few 10's of thousands of cycles, the write limits were a primary design consideration. A poorly designed or implemented process could destroy an SDD in a few 10s to hundreds of hours of use. But on most current SSD's, capacities are in the gigabytes and write limits are at the 100's of thousands of writes or beyond, and they utilize algorithms to spread the writes evenly across the device (i.e. wear leveling), such that you'd have write to the device CONSTANTLY for several YEARS to begin to approach the limits of the device.
The constraining factor today is not the write limit as it once was, but the cost/benefit ratio. A RAID array can easily outperform an SSD on read throughput (and even a typical single HD implementation on write throughput) and yet will cost (literally) pennies on the dollar in comparison.
And I would have to disagree with Skuzzy - the primary benefit of an SSD over the typical hard disk IS reliability and decreased environmental sensitivity. I once read somewhere a maxim that there are only two kinds of hard disks - those that are dead and those that are in the process of dying. With current SSD devices (because of the wear leveling feature) you DO get warning that the device is approaching end of life - the wear leveling algorithms dynamically allocate storage space so that when you rewrite "the same block over and over" it's actually written to a different area of the device each time. And when the write fails, the bad block map is incremented to indicate that some part of the device is going south (although the location is by nature arbitrary, since the entire device is dynamically allocated to begin with), and the data is rewritten somewhere else. And as more of the device fails the number of "bad sectors" reported begins to increase.
And of course, just like a hard disk, "wearing out" is not the only possible failure - there is always the possibility of catastrophic component failure. But catastrophic failure is greatly minimized on a device that contains no moving parts, and has a greatly decreased environmental sensitivity.
But.... unless you truly need such and can justify 10 to 15 times the cost for it, it's (IMO) premature to use them as general purpose devices.
All IMhO, of course.
<S>
-
Ghastly, in theory you are correct. I said "in theory" as the real world has already proven the current crop of SSD's should not be trusted for any critical data due to abrupt failure modes occurring quite frequently.
All SSD's have limited amounts of 'spares' to handle the write failure modes. Once used up, the next failure is a hard failure. In some cases the EEC (particularly in MLC devices) will also fail at the same time.
Again, I am not against SSD. They keep getting better. They just have not reached the reliability of hard drives. I have run drives for over 9 years without a single failure. No one has been able to show that an SSD will last 9 years running 24/7 in a server environment. Matter of fact, in the last 9 years, I have never had a hard drive fail and I have a pretty fair number of them being used in pretty hostile conditions. When SSD's are that good, then I will happily admit your are 100% correct. Currently, I do not believe they have reached that level yet.
-
How expensive are we talking here? In general?
Also, isn't there a major flaw where XP has to write swap files to the same drive that it was installed on? Wouldn't that negate putting all programs on a normal HD, because the swap file is constantly being written to, regardless of the program directory?
-
Well for non-Intell SSD. The cheapest I've seen in the 2.5" form factor is $100 for 8gb. 32Gb runs about $150. 128gb runs about $400+
-
I wasn't aware of an exorbitantly high catastrophic failure rate. The ones we used have worked well enough, although admittedly we aren't yet even close to the one year mark - let alone 10.
And while hard drives have certainly gotten far more reliable than they once were in years past, we still regularly (several times a year, at best) see failures on every brand we have in use. Admittedly, we see 2 to 4 times as many failures as we might otherwise because everything that can be goes in in a mirrored configuration at a minimum (unless we use RAID 5, which we mostly don't anymore). What drives are you using that you see regularly last as long as 9 years without a failure - and how large is your install base?
In short though, I think SSD's mostly disqualify themselves as general purpose devices at this time purely on a price/performance basis. I can get a 1 TB drive (~ 150 or so) for less than a single 64 GB SSD (~ $250) - and if I need better than 1 disk performance I can put 4 of them on an Adaptec controller in a RAID 10 configuration (mirrored and striped pairs), get even better performance than an SSD for little more than the cost of 2 SSD's and get about 25 times the capacity - fully redundant.
SSD's - other than as a curiosity or where a hard disk won't tolerate vibration or acceleration - don't make sense.
<S>
-
Using SCSI drives that are still chugging away after 9yrs+!!!!
Thats 24/7/365 in servers running Oracle databases.
Admittedly we had a couple of failures, but nothing anywhere near the failure rates of IDE/SATA drives. No experience of SSD, looks good on paper but still to early to make a decision I think.
-
We use Seagate Cheetah SCSI drives, pretty much exclusively. I have been using those drives since the late 90's and never had a failure yet. I have replaced some just due to the sheer age of the drive, but never had one replaced due to failure.
IDE/SATA drives fail all the time. Not uncommon for me to replace 2 or 3 of those a year. I would trust an SSD over and IDE/SATA drive.
-
We use Seagate Cheetah SCSI drives, pretty much exclusively. I have been using those drives since the late 90's and never had a failure yet.
you know you just cursed your drive to imminent failure skuzzy :)
-
I am still on the fence about SSD's. They still seem to have a limit on the number of writes before they start failing. Of course, it is a huge number, but it is one of the stealth kind of failures. You have no idea it is about to happen until it does.
Intel is apparently rating their latest 80 gig drive at 100 gig worth of writes per day for 5 years... For a consumer drive that's a lot of use. There is talk of a utility that will provide data on drive health, and there will be 2 S.M.A.R.T. fields that also provide info on drive health. That ought to pretty much take care of your concerns, at least from the point of view of a consumer drive. Servers will of course will still need something with a higher rated duty cycle and as you say, the newer memory tech ought to directly address that.
-
Again, I am not against SSD. They keep getting better. They just have not reached the reliability of hard drives. I have run drives for over 9 years without a single failure. No one has been able to show that an SSD will last 9 years running 24/7 in a server environment. Matter of fact, in the last 9 years, I have never had a hard drive fail and I have a pretty fair number of them being used in pretty hostile conditions. When SSD's are that good, then I will happily admit your are 100% correct. Currently, I do not believe they have reached that level yet.
Don't take this in a bad way, but that's the key difference I think between 99% of users and you... Few people, including hardcore gamers and even a great number of developers, need to get 9 years of server grade duty cycles out of their drives. I expect a lot from my computer hardware, but even as much as I hate to throw out old hardware I don't have a single hard drive in use that is more than 4 years old even though in the last 16 years only 3 of about 30 hard drives I used have outright failed on me. All of the "spare" hard drives in my parts bin were removed in good working condition because they were too slow or small, not because they failed.
I used to only recommend ultra-high quality stuff but gave that up when I realized that only one of the people I provide tech support for (family and friends) have ever used a computer for longer than 4 years before upgrading everything. They don't need server grade hardware, so I quit recommending it and they now get faster stuff for less cost, and can upgrade or replace their computers more often. And they STILL almost never suffer critical hardware failures, including an extremely low rate of true hard drive failure (as opposed to OS corruption, malware-related data destruction, or user error).
-
How expensive are we talking here? In general?
Also, isn't there a major flaw where XP has to write swap files to the same drive that it was installed on? Wouldn't that negate putting all programs on a normal HD, because the swap file is constantly being written to, regardless of the program directory?
According to the Anandtech article on the new Intel SSD, the 80 gig drive will be a bit under $600. The drive has wear leveling logic that should allow it to serve as a primary drive (including swap) for 5 years. Some smart guy did the math (I forget the web site) and he calculated that to use up one of these drives before their rated lifetime, you would have to write to them at the drive controller's maximum data rate for something like 8 hrs a day continuously. That article convinced me that even an expert user ought to have no problems getting the rated life out of these things. Everyone worries about SSD write lifecycles, but nobody seems to be reporting early SSDs failing in consumer applications due to exceeding the memory write count. It's a theoretical problem that doesn't seem to be an issue in real life.
For Skuzzy's uses, yea they're still not suitable. So just don't use them in a server and like any other hard drive, use a good backup scheme.
-
<-- just saying I have in place an IDE HD in use, 40gb segate barracuda for 7 years now, and I have POUNDED it with data. In those 7 years I have re-formatted it 2 times for fresh installs, otherwise data in and out like crazy. it currently has 3% free space, and has hovered around there for 3 years now. I'll kill off enough to get 20% and do a defrag then fill'er back up.
I also have a 20GB IDE western digital still chugging along after 10 years now, about 95% up time (PC almost never rebooted... it is my step-fathers).
I personally have never had a hard drive fail that I've owned, but I have bought "premium" drives. that WD was top of the line back in 1998, same with the seagate.
-
The difference in work rate between a server and a home computers hard drive is enormous.
As I said earlier we have SCSI drives in a RAID 5 configuration that are running 12 x Oracle 8 databases, and another RAID 5 that are running 8 x Oracle 10 databases.
Theres no way IDE or SATA drives would hold up under that workload.
One of our biggest problems was with 2.5" laptop drives. Some of the guys ran multiple demo databases on them, lucky to last 18-24 months before they died.
-
Eagl, I am aware server use is very different from typical users. I do use server load testing to qualify any mass storage device I want to purchase. Yes, I am very critical of mass storage devices. I have spent a good deal of my career focused on mass storage technology. From creating test platforms to giving seminars to hard drive design engineers. I have been around the block a few times in this area.
Personally, I do an enormous amount of video and audio work. When you are writing 16GB to 30GB of data at a time and then re-writing it two to three times after that, it is pretty easy to kill an SSD device. Write-leveling is not a panacea for solving the write problem as it depends on the spares available in the device. It does help extend the life of the device. The projected life of the memory is at best, a guess. We have no idea how the manufacturer rates their devices. There are many factors which can contribute to the failure mode of a memory.
Right now, Intel's SSD's are mostly market-speak. I am not a bleeding edge technology type. I prefer hardware that has a history of success behind it. There is a new memory tech which will obsolete the current SSD's tech and it is showing promise to be a very practical and cost efficient design as well as far more reliable than the current flock of flash based SSD devices.
It is also worth noting that most failures are not write related, but for other problems, although there have been early write failure issues as well. That is the current state of things though. In the real world, they have been failing at a higher rate than they should. Intel's are new, and they may be better. Only way to know is either to jump on the bandwagon early or wait and let others test them for you.
From my perspective, it does not hurt to wait.
-
Want reliability?
I still have an HP server drive. It's on wheels, weighs about 100lbs. The platters are belt driven and it sucks more power than a washing machine. When I got it, I obtained the last 2 cards that were available to interface it to a PC from HP.
It's well over 25 years old and still works. :)
Sometimes just for nostalgia sake I power it up and play with it.
-
ROFL! I still have a couple of 8" Winchester HD's. They store a massive 10MB and are suitable space heaters as well! :)
-
Oh the story behind that drive is something that's hard to believe...
The brilliant people where I was working put the servers in a room that was in the lowest part of the building. For "security" they said.
Murphy struck one day...the server room was flooded with 3 feet of water. The elevated floor did little to protect the drives and they were running when the water hit them.
Insurance replaced the entire computer system and we were wheeling the drives out to the compactor when I asked if I could have one. They were trash so they said it would be OK.
I took it home, dried it out, replace a blown fuse, checked the belts and motor to be sure they were free, plugged it into the dryer outlet (yes 220v operation) and flipped the switch. This thing spun up and I swear I about died laughing. Got the interface cards and found it to be 100% operational. It ran on my BBS as a file server for years after that.
Try that with any drive today and I almost guarantee it won't work.
-
So, what's the difference, interface-wise, between SCSI and IDE/ATA?
I ordered a Gateway a long time ago and one of my options was to pick a SCSI 20GB HD. I chose that option. Later when I got into building/configuring my own hardware I opened her up. She had the HD running through a IDE cable to a PCI card. I don't know why. I eventually removed the card and put the HD directly into the IDE port on the motherboard, and had no problems whatsoever.
Does SCSI use the same plug as IDE, or did Gateway lie to me and give me a second IDE controller instead of a SCSI controller?
-
Gateway lied. The interfaces are completely different.
<S>
-
:furious :mad:
-
So, what's the difference, interface-wise, between SCSI and IDE/ATA?
I ordered a Gateway a long time ago and one of my options was to pick a SCSI 20GB HD. I chose that option. Later when I got into building/configuring my own hardware I opened her up. She had the HD running through a IDE cable to a PCI card. I don't know why. I eventually removed the card and put the HD directly into the IDE port on the motherboard, and had no problems whatsoever.
Does SCSI use the same plug as IDE, or did Gateway lie to me and give me a second IDE controller instead of a SCSI controller?
Nope completely different -
ATA - 40 pin and either 40 wires or later cables use 80 wires.
SATA - Small plug
SCSI - Latest uses either 68 pin, 68 pin high density or 80 pin SCA.
Big diff is max transfers speeds (best avail devices)-
Ultra ATA - 133Mbs
SATA II - 300Mbs
Ultra 640 SCSI - 640Mbs
-
I forgot how expensive SCSI controller cards were.
-
I forgot how expensive SCSI controller cards were.
U320 cards even dual channel ones are a lot cheaper. I have a dual channel U320 PERC (Dell branded Adaptec) in my home system.
-
U320 cards even dual channel ones are a lot cheaper. I have a dual channel U320 PERC (Dell branded Adaptec) in my home system.
Cheapest controller card I found was $70 (next was an $130) and a 74gb 320 SCSI drive is about $150. It'd be faster than a raptor which runs about $150 for 74gb, would just need the controller card.
-
Advertisement. ;)
We have these: http://h18004.www1.hp.com/products/blades/components/c-class-bladeservers.html (http://h18004.www1.hp.com/products/blades/components/c-class-bladeservers.html)
Connected to these: http://h18006.www1.hp.com/products/storageworks/eva4400/specs.html (http://h18006.www1.hp.com/products/storageworks/eva4400/specs.html)
-
ROFL! I still have a couple of 8" Winchester HD's. They store a massive 10MB and are suitable space heaters as well! :)
My school has an old 15MB hard drive that's bigger than a brick. I suspect it might be like the ones you're talking about. The only way I found out what it was was by throwing the SN into Google and it came up with a shop that repairs them.
It only has one dead sector written on it and looks to be in great condition on the outside. I'm questioning if I should hijack it or not, since the lab seems to be throwing everything out. :D