Adaptec Storage Manager And MaxIQ

n individual SSD would still be significantly faster in our database benchmark. However, it would also limit the available capacity to 32GB and remove all redundancy.

What Redundancy? you are running RAID 0. if you consider the faliure rates of spinning drives the single SSD should outlast the RID setup since there is less parts that can fail.

There are software-based solutions for flash cache! Solaris has them in ZFS filesystem, it's called L2ARC. It's also available in Linux BTRFS filesystem.

Also if you use either SUN storage disk array, or NetApp storage array with mechanical disks you can use flash cache that's inside this array (available as option).

i think they wanted to do raid 5 with 4 drives, but can't due to not having 4 drives, so they simulated raid 5 performance with 3 drives and raid 0

Sounds like it's a larger cache with prefetching. It may be effective, but it's nothing new.
Then again, there's nothing wrong with improving an existing idea.

On a side note, if you are planning to use this in conjunction with SSD drives in a array configuration, triple check that you have ones that are fully supported by the controller. Adaptec is having issues with most major SSD manufacturers.

I'm curious if this works with the OCZ Agility EX drives? I already have an Adaptec 5805 controller and the OCZ drive.

True, but if you're using SUN or NetAPP you are talking about a solution that is hundreds of thousands of dollars. This solution would probably be under $1000 (depending on disk sizes and number of disks).

Bull****

1) Expensive. while they say it isn't.
2) Covers just a few applications like database
3) The SSDs, how much will they last? In a real environment outside read-only DBs... a month?
4) Can be done by software
5) RAM is faster. R/W. The only issue is that usually the RAID cards requiere expensive RAM and they charge you a freaking amount of money for the battery backups.

1. Enterprise level solutions aren't cheap. Enterprise SANs run in the hundreds of thousands (or millions, depending on capacity and features). This would be in place of buying either all SSDs or buying a SAN. Its cheaper then both and solves a problem (capacity vs performance). It's a niche product, but it could prove useful.

2. The graphic on MAXIQ details (and the benchmarks that follow) seem tos how it will work for any workload. Beside, in the enterprise your most storage intensive workloads (which this is for) are usually database anyway.

3. We're talking X-25Es here, enterprise level SSDs. 2 million hours MTBF. Even the second generation X-25 M's are 1.2 million hours MTBF.

4. What software can accelerate storage performance?

5. RAM is absolutely faster. But more expensive. High-end database servers can have 256GB (or more). RAM is essential to support high-performance work loads. But if you have a database (or application) of any significant size (e.g. 500GB or more) you will need fast disk to move transactions. And I haven't seen many RAID cards with more then 1GB of cache (Enterprise storage solutions generally don't have more then 64GB, which is very expensive). You need to go to disk at some point, this solution meets a mid-market need. Whether anyone sees enough benefit to use it is another question (but the benchmarks show that it could improve performance)

Costs more than my new build!!! Ouch!

Wouldn't it be cheaper to go with two SSD drives in a RAID-1 config? Then a 3rd as a nightly backup?

It absolutely would, but your capacity would be severely limited. By combining standard mechanical drives with SSDs (for caching) you get increased storage and a boost in performance.

You should really be testing arrays that are much bigger than the SSD cache - like 1TB in size (i.e. 4 WD Velociraptors). There is no reason I would have a 5GB or 25GB RAID array using hard drives if I already had a 32GB SSD. I would just use the SSD to begin with.

I will always Thumbs Up comments with Thumbs Down.

there's no question that when it comes to computer storage as well as
the best upgrade to speed up computers, swapping the hdd for a ssd is
the only way to go. currently the only problem is, ssd's are just too
expensive for most people while hdd's prices are dirt cheap. kind of
like comparing led's to ccfl's.

I've never liked Adaptecs controller boards which are prorpietary; If you build a RAID on an Adaptec controller you're bound to Adaptecs controllers unable to migrate your RAID cluster to other brand controllers. I would rather recommend a controller from 3Ware or LSI or any MegaRAID based controller such as Dell Perc, Fujitsu-Siemens, IBM ServeRAID or Intel Controllers.

The major reasons for the recommendations are good driver support and non-proprietary hardware.

But OpenSolaris use ZFS. It's free and is compatible with most PC hardware (but be careful with which SAS controller you choose HighPoint and Marvell don't agree well with Solaris). FreeNAS, which is FreeBSD based also use ZFS but is less stable and ZFS is not as mature and well-implemented as in OpenSolaris.

The benchmarks here all read and wrote to random areas of the disks, mostly negating the effect of the SSD Cache. Wouldn't a real world sysem be likely to be accessing certain sectors far more often, thereby creating a much bigger boost from the SSD?

Again, my own crazy idea I've mentioned before....instead of using RAID 5, A hardware RAID 3 ( or 4 )should be used with an X-25 as the dedicated parity drive. The SSD should negate the usual bottleneck of the dedicated parity drive, reducing overhead to the quality of xor engine used. 146GB SAS 15K's are common...use a bunch of them with a single ( still pricey ) 160GB X-25. Highpoint still supports what they call RAID 3, but it's really a RAID 4 with a 64K stripe size, which will do just fine.