P-7805u - Performance w/ 2nd Disk in RAID 1

I am loving my cheap P-7805u that I got a month ago. But it's time to make it better.

I am curious if there is any real-world performance increase to be gained from adding another Western Digital 320GB drive in RAID 1.

RAID 0 is not an option for me as I don't want to double the chances of losing the array. 320GB is plenty of storage for me.

This machine is currently used for work and not gaming. It currently runs Vista X86 Ultimate and has Office 2007, Visual Studio 2008, VB6, SQL Server 2000 Developer, Photoshop CS4, and Dreamweaver CS4 installed.

I have also considered the idea of adding a $150 64 GB SSD as my system drive, and using the existing 320 for documents and such. But it seems too risky at this point for a mission critical computer. My hope is that the RAID 1 hard disk based upgrade might help hold me over for now.

Thanks for any advice.

Tony

 

The concept of RAID 1 provides you with, essentially, a "real-time" backup since the drives are mirrored.

If that is your ultimate concern - data loss - than it is somewhat of a safeguard.

Remember, though - a virus or malware that caused data loss will effect both drives.

Of course, an external backup is highly recommended, and the eSATA port should come in handy for you in this regard. I expect that by the tone of your post - "mission critical" - that you are already doing an external backup.

 

RAID1 reduces performance, but it usually not noticeable. The Raid controller has to write the same data on both hard drives...and as a result there is going to be a tad slower. The benefit of RAID1 is that if one hard drive stops functioning, the same exact data will be retained on the second. As for SSD, they are more reliable for there is no moving parts,so no fear of head crashes, or bad bearings, etc...but they are still prone to failure. They can only handle "x" amount of read/writes before the memory block goes dead. So over the years of using a SSD, you will see your total hard drive space decrease, for the SSD is remapping dead blocks.

 

Isn't there a read improvement with RAID 1? I recognize writes are a bit slower. But in theory it can be reading half of the data from each disk at the same time. Maybe I'm giving the hardware too much credit here.

It's cool to know that SSDs kill off the blocks as they go bad. My fear was that I might lose the drive all at once.

I have my current 320 split 3 ways. The C drive is 48.8 GB and about 27 GB is in use. So even if the 64 GB SSD loses blocks, I would still be good to go for quite some time. Tempting...

Tony

 

Obviously, since they are new, no one know for sure, but from what I read, the useful life of an SSD will greatly outlive it's warranty. I think I read that an SSD would have to be written to with 20GB per day for 10 years, or something silly like that. At the rate of increasing capacities and decreasing prices, a new double-sized SSD for my notebook, every year, is a real possibility. I'm not too worried about the little nand-gates getting worn out!

 

Theorhetically, reads are supposed to improve, but it really depends on the RAID controller. The ones I've worked with lately (INTEL MATRIX), there hasn't been a real conceivable improvement on boot times or app launch times in RAID 1. RAID 0 was a definite improvement in my desktop (RAID 0 Velociraptors) with over 200Mbps read performance.

SSD's, as you can tell by my signature, are pretty cool. The single most improvement in performance you can make.

 

Forgot to add that my work documents and stuff are backed up both on my network and offline. So I am not worried about losing work product.

But there's still something to be said for not losing the OS if a drive fails. The great thing would be that I can fix it at my own rate.

 

The Matrix controller card was my concern with the RAID 1 thing. I'm guessing it's better than software RAID, but I wasn't expecting too much from it. I'll skip it for now.

I will go ahead and pick up a cheap SSD. Now I just need to pick one... There seem to be a few in the 60/64 GB range that are not too expensive.

Tony

 

honestly i have not heard this losing blocks theory before... do you have anything to back this up? just curious i want to learn more about this. i've been following the SSD thread and have not seen a mention of it. only problem i've seen really has been the stuttering issues with CHEAP MLC drives that cause system hangs while doing multiple writes at the same time do to the Jmicron controler...

 

Uh, I don't know about the intel controller, but on a GOOD raid controller RAID1 will have nearly double the read speed of a single drive.

 

I don't think the intel matrix raid controller is smart enough to double read speed. It is a very basic raid controller

Depends on what the SSD is rated in terms of rewrites. Each block is 16kb (common on budget mlc ssd), to write 1 byte, the SSD has to rewrite an entire 16kb block (as a result, the write stuttering occurs). MLC are typically rated for 10,000 rewrites (SLC is 100,000)...and 16kb x 10000 = about 20mb. So each block can only do 20mb of rewrite before going bad. As you see, it can get worn out faster than you think...but using a SSD for read only application can really increase an SSD longevity.

It's in the wikipedia page, along with other SSD reliability documentation out there (as reference in wikipedia)... http://en.wikipedia.org/wiki/Solid-state_drive#Comparison_with_hard_disk_drives

Look at the last advantage bullet, and the third to disadvantage bullet...

As for the stuttering issue, it not the jmicron controller fault, but the lack of a onboard cache. As I stated, to write 1 byte, you have to rewrite an entire block. Due to the lack of cache, the SSD has to do this right away. With cache, the data can stay in cache until the NAND can finish erasing the block. The intel SSD has this cache, and future "version 3" ssd will have this also.

 

I don't think it's as doom-and-gloom for SSD's as wikipedia says. Another article here paints a much nicer, and hopefully more realistic picture of wear leveling and useful life of SSD's. Here is an excerpt:



So they say with more than normal use, it is expected that a 64GB SSD would last 51 years, and another study shows that a 32GB flash, completely overwritten 3 times day, would last 85 years. Even if they are 90% off, I'll still be happy.

 

Raid 0 increases performance speeds by splitting up the file between the two drives. Since Raid 1 mirrors the file, in order to provide a read speed increase, the raid controller would have to know how to read half of the file off of one drive while simultaneously reading half of the file off of the other drive, and more importantly, would have to be able to distinguish between the two halves to prevent redundancy.

Either way, Raid 1 will be slower in write time, and won't provide a great performance increase in read times either, even if it was a good enough controller to know how to split-read. Someone here said that raid 1 in theory can double read times. Sorry but that's just not true (I'm sure there are exceptions, but 98% of the time, that won't be true). Raid 0 will barely double read time, so Raid 1 really has little chance.

People do not use Raid 1 for performance, they use it for safety. If performance is your priority, use Raid 0.

If safety is your priority, don't use Raid 1. Instead just use an external hard drive, which is cheaper anyway. There is no point in using Raid 1 now that external hard drives are so cheap. It is more efficient and simply makes more sense to use your second internal drive in a Raid 0 format. You get double the size, and almost double the speeds, and then you can just back up your data more efficiently externally. Again, if safety is your priority, which is safer? Having your data backed up in another location, or risking having your data simply mirrored (what happens when your laptop falls or you spill on it, and then both drives fail?).

Using Raid 1:
Pros - Data is backed up and "safe." Small read speed increase.
Cons - If you spill on the laptop or drop it, both drives could fail. Creates more heat inside the laptop.

Using an external HDD:
Pros - Your second hard drive is freed up for Raid 0. Huge performance increase. Data is backed up. Cheaper and larger capacity than internal HDD's. Doesn't cause internal heat. In a separate location, so not at risk of being destroyed were the laptop to get destroyed.
Cons - None versus raid 1.

 

Raid0 requires the same "logic" to read together stripes of data as Raid1 does, so why would it be that difficult for a Raid1 array to interleave two disks to maximize read speed? As for write speed you shouldn't see a significant hit there because you're writing identical data to identical drives at the same time. There is a penalty but it's not that severe. A lot of cheap controllers don't effectively use both disks in a raid1 array though, they frequently use one drive unless two seperate IOs are called in which case they'll read the second data from the second drive, only benefiting a multitasking environment such as a large database server or the like, but frankly if your serving a database I'd hope your using raid 10 or a raid 5 parity set.

 

It doesn't require the same logic because Raid 0 splits a file into pieces and puts those pieces on separate drives, whereas Raid 1 copies the whole file to the separate drives. Raid 0 just has to combine those pieces when reading. But in order for Raid 1 to provide a read speed increase, it has to pull pieces from an entire file on one drive, and then it has to pull pieces from the other drive(s) knowing which piece(s) it already took from the first one. Then it has to combine those, so not only does it have to break up the files into pieces, but it has to know which pieces it took from one drive so that it doesn't pull those pieces from the other drive.

You might be writing identical data to identical drives, but you're still doing double the work that a non-raid setup would have done. In a non-raid setup, you're just writing one file to a drive. In a Raid 1 setup, you're doing that as many times as drives you have. I'm not saying that the write speed time will halve, but it will definitely take a performance hit.

 

Ok, and again, it takes the same amount of smarts to split those files when writing them in RAID0 as it would to split the reads in RAID1. It's not a real complex operation, I don't know how exactly they do it but I'm sure it's not hard when the files are written in blocks of X amount of data (4k in the case of most NTFS installs), and the drive is split further into tracks/cylinders/etc that could all be used as simple reference points and grab odds and evens or every other or whatever is necessary to get a good read quickly, the drives structures are written identically so any physical reference on the drive can be used as a logical split to recombine the data when reading.

I didn't say you don't take a hit writing, you do, but its not all that significant. I guess I just see people trying to disuade someone from using raid1 without a particularly compelling reason not to use it. Frankly for a lot of users raid0 really isn't much faster than a single disk would be, if someone wants to take a slight penalty for data security they're probably not going to get that much advantage using a raid0 since its best application is games and other programs that use very large files that need to be read frequently.

 

Sorry I'm not getting through. First you said "Raid0 requires the same "logic" to read together stripes of data as Raid1 does" and in that you compare Raid0 reading to Raid 1 reading. Now that I explained that that's not true and how it works, you're trying to say that Raid 0 writing take's the same amount of "smarts" as Raid 1 reading, and in this you're comparing reading to writing. Firstly, it doesn't take the same amount of "smarts" and secondly, it's not even the same operation. You can't compare reading to writing, they're two completely different things. The process required to take a file from memory, split it up, and write it to two drives if very different from taking two identical files on two different drives, splitting them up so that there is no redundancy, and combining those pieces into the files. It's almost like you're saying the process of mitosis is the same as the process of meiosis.

Listen, I'm not trying to prove you wrong, I just don't want the OP or other readers having false hope is all.

Well now you're just talking semantics. What do you mean by significant, compared to what? I think the majority of the people would consider as much as a 50% decrease in write speed a significant hit.

As I said in my other post. Raid 1 SHOULD be dissuaded from use. External hard drives are cheap now. You can get MORE space for CHEAPER than you could an internal hard drive. Not only that, but people seem to need all the space they can get on internal drives now, why waste it on mirroring.

But more importantly, everyone should be backing up their data, and they should not be doing it on internal drives through Raid 1. The chance of one drive failing is just as likely as the chance of both drives failing at the same time, especially in laptops which are mobile and are much more prone to accidents. You can say "it's ok that I can't utilize my second hard drive, because now my data is twice as safe" but then woops, you spilled your coffee on your keyboard and you not only lost all your data, but it costs you twice as much with the second hard drive in there. That wouldn't happen if you had backed it up on the external drive instead. That being said, Raid 0 has NO dissadvantages since you're going to be backing up your data to an external HDD anyway, so why not both utilize the space of your second drive (in Raid 1 you cannot use the second drive) and also get a significant performance increase.

 

Again, data is written in BLOCKS, all you have to do is read 2 different groups of blocks. This is by NO MEANS a complicated operation to a computer regardless of whether you have 2 complete sets of blocks or two sets of different blocks split into two places. The way you explain it you make it sound like a computer is a couple of people trying to read a script where both people have the same script and their "parts" aren't noted so they're trying to figure out when to take turns reading. A computer uses blocks/sectors/tracks etc to file its information and uses those blocks whenever it retrieves information, telling two hard drives to cough up two different sets of blocks is about as elementary as it gets.

You find me a raid1 that takes even CLOSE to a 50% performance hit during writes that uses any type of hardware controller and I'll withdraw my entire argument, tell everyone your right and apologize to you personally.

So every notebook hard drive failure involves both drives now because everyone that's ever lost a notebook hard drive has spilled their coffee in the thing or dropped it down a flight of stairs?

 

sorry but this just isn't true. especially in a raid array, where its dependent upon stripe and cluster size, especially with variance in files size where files can fully or partially fill, or a file can be even smaller than what you would call a "block." i think where you're going wrong is you're taking what you think to be a general understanding of how normal non-raid setups work and then you're applying it not only to raid but to two completely different raid configurations and say that because the same general idea applies, the same process must be held in the configurations. this is shown in the fact that you at first tried to compare read times in raid 0 and 1 and then when i pointed out the differences you changed your argument to comparing raid 0 writing and raid 1 reading. Now that I pointed out why you're wrong in that aspect, you've digressed to pointing to a general understanding of disk writing.

This isn't an argument supporting you or against me, it's just an argument to the lack of data on the web regarding raid 1 setups. Hmm, I wonder why (could it be people realize that Raid 1 is currently obsolete?). If you can three instances on the web where people compare their non-raid write times to their raid 1 write times, I'll be impressed. Regardless, even IF raid 1 had a ZERO% performance decrease, my argument that Raid 1 is little more than waste versus the other options would still be valid.

You quoted something I wrote, and then in your response, completely disregarded what I wrote. Did you even read it? Here: " The chance of one drive failing is just as likely as the chance of both drives failing at the same time, especially in laptops which are mobile and are much more prone to accidents. You can say "it's ok that I can't utilize my second hard drive, because now my data is twice as safe" but then woops, you spilled your coffee on your keyboard and you not only lost all your data, but it costs you twice as much with the second hard drive in there."

And you take "just as likely" to mean "every notebook hard drive failure involves both drives"? And what of the fact that Raid 0 + external hard drive will not only save you money, but allow a significant performance increase AND allow you to utilize twice the space of your internal hard drives, without the risk of data loss were an accident to happen?