128GB C300 raid1 or single 256

vroom_skies

[H]ard|Gawd
Joined
Sep 25, 2005
Messages
1,390
Hey all,
I'm just curious as to your thoughts on this.
If my understanding is correct trim still can't work when in raid. However I hear of more and more people raiding ssds, so I don't know what's up.

So here is what I'm thinking. I currently have a 128GB C300. I really need over 200GB of space on my SSD solution. So I was either thinking of going raid 1 or a single 256 M4 or 830 (or w/e is best).

Thanks in advance.
 
Given the low prices on current 240/256GB drives, that's your best bet. But you should still have a good backup solution.
 
Personally, until TRIM is supported in RAID, I'm going to go single SSD only. When that's sorted - you better believe there will be a large RAID 0 a'comin
 
Haha, my mistake.
I meant to type raid 0.

I have a good backup solution that the ssds are not a part of.

Thanks though
 
GC does 95% of TRIM and the only guys that shy away from RAID for that excuse have no idea what they're talking about.

RAID0 will give you much faster speeds but a back-up plan is a MUST.

I've run 2 256GB C300s in RAID0 for over 9 months with NO loss of performance.
 
yeah GC will get the jib done. RAID pretty awesome. But with the 250's getting cheaper you can grab one of those now. And then RAID it later.
 
I always recommend two smaller SSDs in RAID0 versus a single larger SSD.

1. There are next to no extra data integrity worries aside from maybe having a bad cable or something. An SSD is actually a self-contained RAID0 in it's own right - several striped banks of NAND flash memory. SSDs in a RAID0 are actually quite a bit more trustworthy than two mechanical drives in RAID0.
2. Performance advantages in everyday usage with two SSDs in RAID0 versus a single SSD are marked and noticable.

TRIM isn't NEARLY as big a concern as some people make it out to be - an SSD (in RAID or not in a RAID) without TRIM could be used for years without a noticable performance decrease. Add to that the fact that most current generation drives have decent garbage collection routines built in and it's just not something to be concerned about. If performance decreases due to lack of TRIM ever DID somehow become a problem all you have to do is perform a secure erase (use a utility that sends the ATA SECURE_ERASE command to the drive) and bam, the drive is brand new again.

Supposedly we are on the verge of TRIM working in RAID in an upcoming Intel RST release - I will believe if when I see it though. People have been saying that for almost three years now... lol
 
I always recommend two smaller SSDs in RAID0 versus a single larger SSD.

1. There are next to no extra data integrity worries aside from maybe having a bad cable or something. An SSD is actually a self-contained RAID0 in it's own right - several striped banks of NAND flash memory. SSDs in a RAID0 are actually quite a bit more trustworthy than two mechanical drives in RAID0.
2. Performance advantages in everyday usage with two SSDs in RAID0 versus a single SSD are marked and noticable.

TRIM isn't NEARLY as big a concern as some people make it out to be - an SSD (in RAID or not in a RAID) without TRIM could be used for years without a noticable performance decrease. Add to that the fact that most current generation drives have decent garbage collection routines built in and it's just not something to be concerned about. If performance decreases due to lack of TRIM ever DID somehow become a problem all you have to do is perform a secure erase (use a utility that sends the ATA SECURE_ERASE command to the drive) and bam, the drive is brand new again.

Supposedly we are on the verge of TRIM working in RAID in an upcoming Intel RST release - I will believe if when I see it though. People have been saying that for almost three years now... lol

TRIM is not super necessary, but many drives can and do get hammered in R0 without it. Some controllers are much better at active and idle GC than others, but you also don't want a drive with overly aggressive GC. Overprovisioning makes a huge difference in the ability of the drive to maintain itself without TRIM, so it's pretty much manadatory. Performance degradation could get better over time with idling or sequential writes, but steady state performance is going to suffer greatly on many drives in that environment. Some Marvell controlled drives perform quite well in the absence of TRIM, and the Samsung 830 does well too.

It is true that SSDs write to each device in parallel whenever possible. If the drive doesn't have some sort of redundancy scheme ala the Intel 320 series/ SF's RAID, then it can't survive the loss of die or plane. I have seen a couple of drives lose whole dice, and one survived while the other didn't. The drive that survived had a parity scheme and extra flash, while the dead drive lacked that particular feature.

When drives die, it's from a firmware issue mostly. Drives where nothing is wrong with the flash or the controller, but are rendered dead though fault of the FW. In that event, MLC or SLC, or RAISE or RAID-4 like parity scheme, the drive is dead regardless. So the chances of a single drive failure killing the whole R0 array are at least a little higher than one drive alone.

Intel RST 11.5 is supposed to add trim in RAID, but the current alpha drivers do not yet have this feature. I believe the 11.5 OROM is needed as well, and it's doubtful that many current boards will get that upgrade. As far as I know, just the drivers alone won't give you TRIM in RAID. But I do think you'll see 11.5 in the next few months.
 
Wow, thanks guys, this is great information!

So it looks like I'll go with a second 128GB C300 and throw them in RAID 0.
Correct me if I'm wrong, but does all this GC happen in the background or is there anything I need to do on the surface?

Also if anyone has a 128GB C300 for sale please let me know.

Thanks
 
TRIM is not super necessary, but many drives can and do get hammered in R0 without it. Some controllers are much better at active and idle GC than others, but you also don't want a drive with overly aggressive GC. Overprovisioning makes a huge difference in the ability of the drive to maintain itself without TRIM, so it's pretty much manadatory. Performance degradation could get better over time with idling or sequential writes, but steady state performance is going to suffer greatly on many drives in that environment. Some Marvell controlled drives perform quite well in the absence of TRIM, and the Samsung 830 does well too.

Meh, after a time without TRIM there is always going to be a measurable performance difference as more Read-Erase-Modify-Write cycles have to be performed. Measurable yes, noticeable no. I would have to maintain that lack of TRIM isn't a huge concern for most enthusiast users as we tend to reformat (and would issue a secure_erase) at least once a year if not more often. I do agree that differences in the controller and GC can cause varied results.

It is true that SSDs write to each device in parallel whenever possible. If the drive doesn't have some sort of redundancy scheme ala the Intel 320 series/ SF's RAID, then it can't survive the loss of die or plane. I have seen a couple of drives lose whole dice, and one survived while the other didn't. The drive that survived had a parity scheme and extra flash, while the dead drive lacked that particular feature.

When drives die, it's from a firmware issue mostly. Drives where nothing is wrong with the flash or the controller, but are rendered dead though fault of the FW. In that event, MLC or SLC, or RAISE or RAID-4 like parity scheme, the drive is dead regardless. So the chances of a single drive failure killing the whole R0 array are at least a little higher than one drive alone.

True, but I'd still put those chances right up there with the chance that Adele is going to sing a song about finding a guy and being happy. :D If an SSD is going to fail it's going to fail regardless of whether it's in a RAID array or not. What I meant is that the pure fact of having two SSDs striped in a RAID0 array (again for performance, not redundancy) isn't really any more dangerous or prone to failure than a single SSD.

Intel RST 11.5 is supposed to add trim in RAID, but the current alpha drivers do not yet have this feature. I believe the 11.5 OROM is needed as well, and it's doubtful that many current boards will get that upgrade. As far as I know, just the drivers alone won't give you TRIM in RAID. But I do think you'll see 11.5 in the next few months.

Well for me inserting the OROM into an existing motherboard BIOS is usually pretty easy, so if it turns out that we need an 11.5 OROM for TRIM with the 11.5 RST driver I can handle it. It's kind of a bummer for the average joe though as I know of hardly any motherboard manufacturers that bother to update the OROM in their BIOSes.

Still, I'll believe TRIM in RAID when I see it happening. We've all been waiting on it for what like three years now? :D
 
if your worry about performance lost

you can always try diskeeper hyper fast defrag

its like a software trim program, this was actually release before trim came out, i believe

It's not software trim. Software trim can't exist, since the SSD isn't showing its guts to the OS.

What is does is capturing the writes and putting them in sequences if they're random. Might be good, might slow everything down.
 
Meh, after a time without TRIM there is always going to be a measurable performance difference as more Read-Erase-Modify-Write cycles have to be performed. Measurable yes, noticeable no. I would have to maintain that lack of TRIM isn't a huge concern for most enthusiast users as we tend to reformat (and would issue a secure_erase) at least once a year if not more often. I do agree that differences in the controller and GC can cause varied results.



True, but I'd still put those chances right up there with the chance that Adele is going to sing a song about finding a guy and being happy. :D If an SSD is going to fail it's going to fail regardless of whether it's in a RAID array or not. What I meant is that the pure fact of having two SSDs striped in a RAID0 array (again for performance, not redundancy) isn't really any more dangerous or prone to failure than a single SSD.



Well for me inserting the OROM into an existing motherboard BIOS is usually pretty easy, so if it turns out that we need an 11.5 OROM for TRIM with the 11.5 RST driver I can handle it. It's kind of a bummer for the average joe though as I know of hardly any motherboard manufacturers that bother to update the OROM in their BIOSes.

Still, I'll believe TRIM in RAID when I see it happening. We've all been waiting on it for what like three years now? :D

Yeah, that's all pretty much true. I put a higher value on TRIM, but less value on R0 arrays.

My Intel DP67BG motherboard gets frequent OROM updates in new BIOSs. My ASUS boards don't.

My Intel board just got a new UEFI with 11.0 OROMs, so hopefully 11.5 will only be a few months away.
 
Any free programs like that out there? Ones that have a good reputation in the community.
My C300 has honestly been running quite sluggish lately and I can't figure it out.

mydefrag has the same SSD defrag method as diskeeper and is free. Although read Aesma's reply as the same applies to mydefrag.
 
Any free programs like that out there? Ones that have a good reputation in the community.
My C300 has honestly been running quite sluggish lately and I can't figure it out.

And if you can't find a utility (and don't mind having to back your data up) you can always use HDDErase or GParted to send a SECURE_ERASE command to the drive - it will return it to factory new status as far as performance goes. Remember though that a secure erase completely wipes the drive so be prepared to repartition/reformat the drive.
 
Thanks you guy.
Got the C300 running great, but just picked up a 830 256.
So the C300 is now for sale.
 
Back
Top