RAID 1.5 - Speeds of RAID0, security of RAID1, all in 2 drives? Or just marketing gimmick?

[Confused]

THG has the article



Confused

 

[Pariah]

It's marketing BS. What they are calling RAID 1.5 is really RAID 1 with loadbalancing reads. This is a feature that has been around for awhile in the SCSI realm, and random ATA controllers (3Ware), but this is the first company to try an market it as a "new" feature. It is a legit feature that improves the read speed of RAID 1 arrays, but it isn't anything new, and there is no such thing as RAID 1.5.

 

[First]

Funnily enough, THG initially called this article "RAID 15 With IDE: Added Value Or Eyewash?" instead of RAID 1.5

 

[Confused]

Originally posted by: Evan Lieb
Funnily enough, THG initially called this article "RAID 15 With IDE: Added Value Or Eyewash?" instead of RAID 1.5

I'd noticed that they had put 15 in a load of places, as well as 1.5, and the graphs showed 15 as well!!

I guess someone made a boo-boo!


Confused

 

[LED]

RAID 1.5 - Speeds of RAID0 No, security of RAID1 Yes, all in 2 drives Yes? Or just marketing gimmick? Yes

 

[Pariah]

Originally posted by: Evan Lieb
Funnily enough, THG initially called this article "RAID 15 With IDE: Added Value Or Eyewash?" instead of RAID 1.5

I didn't read the entire article, but just browsing through parts they do talk about "true" RAID 15 as well as the made up 1.5.

Speeds of RAID0 No,

Depending on the implementation, it can get pretty close in some situations. It certainly will be faster than a single drive. Noticable? Unlikely, but I don't think RAID 0 is useful to the average user either.

 

[Lord Evermore]

Wow, that THG article is unnecessarily confusing the issue and making it seem like there IS something special to RAID 1.5 marketing. There's no need to make comparisons to RAID 15, the 1.5 does absolutely nothing regarding "parity data" like RAID 5, it is just a mirrored array with balanced reads, which I think most IDE RAID controllers do now (Promise at least does it). No parity information is being processed or stored (THG does mention no parity processing, I wonder where they expected the information to come from; they also never do mention parity being involved in the actual transfers, despite saying that parity is used).

By even mentioning RAID 51 or RAID 51, THG just lends more respectability to a marketing gimmick that is NOT actually any better for the user.

HighPoint doesn't even mention the 372n on their site. I think it's just a 372 chip with a modified ID so that DFI can claim to have something special. Apparently Tyan has used it as well.

I would like to know how the HPT374 is supposed to support RAID 5 without an XOR engine, but they removed RAID 01 capability (but not RAID 10).

 

[Shooters]

Okay, I'm officially confused. The person that wrote that article needs to take some classes in technical writing. So, does RAID 1.5 use parity or not? The last sentence on page 3 states "The capacity available with the RAID 1.5 array equals the capacity of a single hard drive, leaving the other half of the total capacity for parity data." So, does that mean that the controller is striping the user's data with true parity data by using the XOR operator, or does the controller just write two mirrored drives like a RAID 1 and read from both of them simultaneously like a RAID 0? In the case of the latter, I don't see how it functions any differently than most RAID 1 implementations that are on the market.

 

[Pariah]

the controller just write two mirrored drives like a RAID 1 and read from both of them simultaneously like a RAID 0?

yes

In the case of the latter, I don't see how it functions any differently than most RAID 1 implementations that are on the market.

ATA RAID controllers do not read from both drives in RAID 1 configurations, with rare exceptions (3Ware).

which I think most IDE RAID controllers do now (Promise at least does it).

Unless they just started doing it, up until about a year ago, no Promise controller including their highend ones did.

I would like to know how the HPT374 is supposed to support RAID 5 without an XOR engine, but they removed RAID 01 capability (but not RAID 10).

Any RAID controller is capable of RAID 5 with a software driver.

 

[Lord Evermore]

Promise lists load balancing, elevator seek and tagged command queuing as features (at least for the TX2000 and up). I presumed that load balancing would include reading from mirrors but perhaps not. I can't imagine why they wouldn't include that if Highpoint is capable of doing it.

I know RAID5 could be done in software, but the performance hit is always claimed to be so tremendous that it wouldn't be worth using, and I can't imagine anybody actually using it on a consumer board (this is after all a "LANParty" board where performance is key). THG mentions the HPT374 supporting it, but then says that their "experience" with parity storage is transferred to the 372, but then never mentions parity again with this chipset.

The horribly written sentence regarding "leaving the other half of the total capacity for parity data" is very misleading again. There is no parity data. The visible drive size to the OS is the space of one drive. The "other half of the total capacity" doesn't contain "parity data", the two drives are simply mirror copies of each other. That is different from being "parity data".

 

[AgaBoogaBoo]

That is one of the worst articles written... they made it very confusing to write. I wish Tom himself would write some good articles because he is pretty good but that person that wrote it doesn't seem to be trying to make it easy and simple to understand. I was going to go for Raid 1.5 but thankfully I posted here and found out the truth about it. Since I don't know much about Raid devices archtiecturally I was thinking like a n00bie and thinking that it mirrors like Raid 1 on both drives and then splits the data requests up so if the total size is 1024 then it gets a stripe from each at a time. And yes, I found out it doesn't work like that!

 

[Pariah]

Since I don't know much about Raid devices archtiecturally I was thinking like a n00bie and thinking that it mirrors like Raid 1 on both drives and then splits the data requests up so if the total size is 1024 then it gets a stripe from each at a time. And yes, I found out it doesn't work like that!

In a nutshell, that is what it does. It's a standard RAID 1 mirror array that splits up read requests among the 2 drives.

 

[Acanthus]

just get Raid5.

 

[Lord Evermore]

Essentially it is just a RAID1 array, with features not found on most ATA RAID controllers (or at least not advertised as a special feature) but which SCSI RAID has always done, and DFI and Highpoint are marketing it as some sort of special RAID configuration when it's really just RAID1 with another feature included. Calling it RAID 1.5 isn't "wrong" because there's no official RAID 1.5, but it gives the customer the impression it's something special, and for those that have ever heard of RAID 5 or RAID 15/51 they might think this involves the same features, so basically it's misleading, and the author of the article seems to have fallen hook line and sinker.