Sign in with
Sign up | Sign in
Your question

True Hardware Raid

Last response: in Storage
Share
April 3, 2007 5:53:21 AM

Evidently, some RAID controller cards are not a fully hardware-based. Can someone suggest a true hardware-based RAID conroller that's not outrageously expensive?

Windows Vista Ultimate
Intel Core 2 Duo E6600
BFG GeForce 8800 GTX OC
Corsair XMS2 TWIN2X2048-6400 2X1GB DDR2-800
Western Digital 150GB Raptor X
Sound Blaster X-Fi XtremeGamer Fatal1ty Pro Series
Dell E207WFP & Samsung SyncMaster 213T
Asus Striker Extreme Motherboard
PC Power & Cooling Turbo-Cool 1KW-SR
Thermaltake Kandalf LCS VD4000WBS (no leaks)
Pioneer DVR-212D DVD/CD Writer
APC Back-UPS XS 1000

More about : true hardware raid

April 3, 2007 11:20:16 AM

well, 'outrageously expensive' is pretty relative, especially compared to other enthusiast hardware (new cpu, new hdd, new gpu, new monitor, water cooling, etc)...

Areca ARC-1210 PCI-Express x 8 bus SATA II RAID Controller Card RAID level 0, 1, 1E, 3, 5, 6 (if RAID 6 engine supported) and JBOD - Retail $320
http://www.newegg.com/Product/Product.aspx?Item=N82E168...

Open Box: Areca ARC-1210 PCI-Express x 8 bus SATA II RAID Controller Card RAID level 0, 1, 1E, 3, 5, 6 (if RAID 6 engine supported) and JBOD - OEM $290
http://www.newegg.com/Product/Product.aspx?Item=N82E168...

LSI LOGIC LSI00110 PCI Express SATA / SAS Controller Card RAID 0/1/1E/10E - Retail $315
http://www.newegg.com/Product/Product.aspx?Item=N82E168...

LSI LOGIC LSI00051-F PCI Express SATA / SAS Controller Card RAID 0/1/5/10/50 - Retail
High-performance SAS/SATA controller, limited time offer $350 (normally $650)
http://www.newegg.com/Product/Product.aspx?Item=N82E168...

3ware 9650SE-4LPML PCI Express x4 SATA II Controller Card RAID Levels 0, 1, 5, 10, Single Disk, JBOD - Retail $335
http://www.newegg.com/Product/Product.aspx?Item=N82E168...

as always, it wont hurt at all to search for reviews and benchmarks on any of them that you might consider buying

edit: the highpoints i had listed i guess werent hardware based, which probably explains why they were so cheap to begin with, lol
April 3, 2007 1:21:06 PM

Yeah, I've been looking into them recently as well. Noticed a new one for over a grand and said..."Hmm."

Anything cheaper than $320? Still seems like a hell of a lot for a PCI card...I'd sooner wait for Oblivion levels to load, and go ahead and buy a new video card with that cash. If anyone can place a superb card with 2-4 SATA connectors for $150 or less I'll consider it.

MUST be completely hardware based. Using up system resources completely defeats the point of buying the card since motherboards now come with software raid.
Related resources
April 3, 2007 1:42:03 PM

well... its not so much about using resources or not, its the card itself being able to intelligently manage and organize the I/O requests with a built in logic chip (i.e. IOP333 I/O processor) and onboard memory buffer (among other things i presume), that an x86 cpu is pretty incompetant about handling all by itself, let alone managing efficiently

to loosely compare, its like comparing a specialized add on physics processor, to a general purpose single core x86 cpu having to calculate the physics instead... they both can do it, one just does it a lot better

a couple reviews THG did on some pcie hardware raid controllers
http://www.tomshardware.com/2005/10/31/sata_spells_trou...
http://www.tomshardware.com/2006/12/13/pci-express-sata...

http://www.xbitlabs.com/articles/storage/display/areca-...
a b G Storage
April 3, 2007 2:43:01 PM

Quote:
Anything cheaper than $320? Still seems like a hell of a lot for a PCI card...I'd sooner wait for Oblivion levels to load, and go ahead and buy a new video card with that cash. If anyone can place a superb card with 2-4 SATA connectors for $150 or less I'll consider it.


A quality true hardware RAID card is one of those things that fits into the "You Get What You Pay For" category. I am now using a cheap software RAID5 solution (in sig) and I have used 3Ware cards; the 8506 and 9500S. I will vouch for the fact that if you want the performance of real hardware RAID, it's worth the price you pay. For the silicon and technology you get on a hardware RAID card, paying ~$300 is actually reasonable.

Another thing to consider, how much does a cheap semi-hardware/software based RAID card really load down a high-end dual core system with lots of RAM. If all you want is some faster load times, then just having a RAID array will give you noticeable improvements compared to a single drive. Let alone noticing any system lag because the RAID solution is software based as opposed to hardware based. For example, I'm using my RAID5 as redundant storage and as a file server to 3 pc's thru my house. The semi-hardware/software solution I have offers performance enough with no noticeable system lag to my last gen dual opterons. Again, just something to consider...

Good luck!
April 3, 2007 2:48:35 PM

Quote:
If anyone can place a superb card with 2-4 SATA connectors for $150 or less I'll consider it.


we aren't raid card salesman. i dont give a rats ass if you buy one or not. if you need one, you find the one you need and buy it. if you dont, you dont.

chances are whoever started this thread has no real need for raid in the first place.

Valis
April 3, 2007 2:48:59 PM

Valis, I'm the chap that started this thread. I don't *need* a RAID controller, I *want* a RAID controller. I have blue LEDs and UV lighting in my computer chassis too. Are you okay with that? :D 
April 3, 2007 3:31:13 PM

Quote:
Valis, I'm the chap that started this thread. I don't *need* a RAID controller, I *want* a RAID controller. I have blue LEDs and UV lighting in my computer chassis too. Are you okay with that? :D 


i wasn't responding directly to you, i was responding to the user i quoted. the one that made it appear as if he was doing us a favor by buying a raid controller.

(and i have blue and red AND uv lighting as well as el-tape in mine ;)  )

(i'm using highpoint rocketraid 2320, currently 2.5 terabyte raid 5, it's a dedicated storage box, upgradeable to 10+ terabytes)

Valis
April 3, 2007 4:10:42 PM

Quote:
Anything cheaper than $320? Still seems like a hell of a lot for a PCI card...I'd sooner wait for Oblivion levels to load, and go ahead and buy a new video card with that cash. If anyone can place a superb card with 2-4 SATA connectors for $150 or less I'll consider it.

MUST be completely hardware based. Using up system resources completely defeats the point of buying the card since motherboards now come with software raid.


You won't find a hardware RAID-5 card for $150, not new.

On the other hand, you can't do RAID-5 with only two drives, so perhaps you're only interested in RAID0 or RAID1, neither of which need significant amounts of hardware.
April 3, 2007 4:11:53 PM

directed @ both valis and chunkymonster: im wondering though, since both of you have used/own hardware raid controllers, how do you feel they perform compared to general software based onboard raid controllers as far as an improvement in overall performance? for things such as general windows usage, games, and the like?

im genuinely wondering if there really is an appreciable difference, or if onboard software raid could nearly be considered hardware raids equal in this respect

thanks
April 3, 2007 4:24:32 PM

Quote:
directed @ both valis and chunkymonster: im wondering though, since both of you have used/own hardware raid controllers, how do you feel they perform compared to general software based onboard raid controllers as far as an improvement in overall performance? for things such as general windows usage, games, and the like?

im genuinely wondering if there really is an appreciable difference, or if onboard software raid could nearly be considered hardware raids equal in this respect


I've used software RAID-0, RAID-1, and RAID-5 in Linux, and am currently using hardware RAID-6 with a 3ware 9650 card.

The benchmarks show a huge performance improvement going to hardware RAID. However, I don't actually notice all that much difference given my regular workloads in Linux. I'm a software developer so a lot of my computer's effort is spent compiling.

So really, it's going to depend on your workload. With software RAID, your CPU fills in the gap whenever you are writing anything to disk. Can you afford to give up some CPU power at that point? Often, you can.

Also, the way RAID5 (or even RAID6) is implemented, you aren't going to get a significant increase while READING data. Unless your array is degraded, of course.

So, what's your workload? Most people won't see much improvement with hardware RAID5 because most disk access is reading, not writing. And most of the time when they write data, they aren't using much CPU at the time. Some people will notice a huge improvement, mind you. I wouldn't want to run a database on software RAID-5, for example. Also, you'll have a much easier time with software RAID-5 with a dual core or quad core machine.

RAID6 is still another matter. It's going to use about twice the CPU of RAID5.

This is all assuming that the Windows driver provides a high quality implementation of software RAID. This may very well not be the case.

One final thing to consider is that a real hardware RAID5 or RAID6 card will have its own cache. Mine has 256 megs of RAM. This alone provides a non-trivial benefit (and tends to skew benchmarks unless they test with a substantial amount of data).
April 3, 2007 4:32:03 PM

Quote:
directed @ both valis and chunkymonster: im wondering though, since both of you have used/own hardware raid controllers, how do you feel they perform compared to general software based onboard raid controllers as far as an improvement in overall performance? for things such as general windows usage, games, and the like?

im genuinely wondering if there really is an appreciable difference, or if onboard software raid could nearly be considered hardware raids equal in this respect

thanks


as the poster above stated, it also depends on what raid level you're going to be using. raid 5 and 6 and others that utilize parity require much more intensive computations and processing power than other raid levels. so on my raid 5 arrays i use hardware based controllers, for my mirrored arrays i use software controllers.

i dont reccomend using onboard raid PERIOD. to much crap can go wrong and you're generally locked in, meaning if something goes belly up with your motherboard and you cant find the exact same one then you're pretty much hosed on the raid as well, it's generally not a walk in the park moving one raid setup from one controller to another (if not impossible).

i dont use raid in a general setup, there just isn't enough boost to justify it. i use a 120 gig raid 0 as a scratch and speed disk in my main machine, it's fastest at unrar'ing and unzipping and compressing and uncompressing multiple gig size files. when you're unrar'ing something 8 gig in size there is definately a speed increase, by several minutes. however, if you just want to load your game faster, you'll be sitting at the "loading" screen ~10 seconds less than otherwise. meh. also, if one drive fails, the whole thing goes, and so nothing of importance is kept on the drive.

on another computer i use a software raid 1 for security, if one drive fails there's an automatic backup.

i have two storage boxes both with hardware based raid controllers, one with 4 x 160 gig drives, one with 6 x 500 gig drives, 450 gig and 2.5 terabyte respectively. they both sit on their own machines, the 2.5 terabyte is a dedicated storage box, expandable to 10+ terabytes. it's a pci-express raid card, but is software based. i didn't need a hardware based raid since this is a seperate box and the machine can work on calculating parity and it doesn't bother me over on my main machine.

so in a nutshell, if you are using raid 5 or 6 or such for storage you want a card that can calculate the parity bit in hardware if you're also using that machine for other tasks, like video editing. if it's a standalone box, software is fine (but not onboard, be sure you know the difference).

if you want application speed you'll probably do fine with onboard since you'll be using raid 1, striping, which isn't computationally intensive, but dont put anything you can't bear to lose on the drive, and i wouldn't make it my OS. add in two drives, stripe them, and use them for installed games, with your other applications and especially important data elsewhere.

valis
hope that helps a bit. this has probably been a bit rambling.
April 3, 2007 4:33:48 PM

well, i was mainly asking due to supposed benefits that a hardware controller could offer particularly when used for games (as thats usually the reason people ask for raid advice on here)

if i remember correctly, another member had said that one of the main differences was in how the controller dealt with the I/O requests, compared to how the cpu would instead (be unable to) properly handle it... and with the proper handling of the random I/Os, you could get a significant performance boost (even in load times), considering optimized striping and cluster sizes as well

also because my friend is considering purchasing one of the above, to test on his own, and it would be a waste of his money, if it really didnt benefit much at all for general uses

edit: ah, okay. that clears things up a bit then. thanks :) 
April 3, 2007 4:57:55 PM

Quote:
well, i was mainly asking due to supposed benefits that a hardware controller could offer particularly when used for games (as thats usually the reason people ask for raid advice on here)

if i remember correctly, another member had said that one of the main differences was in how the controller dealt with the I/O requests, compared to how the cpu would instead (be unable to) properly handle it... and with the proper handling of the random I/Os, you could get a significant performance boost (even in load times), considering optimized striping and cluster sizes as well

also because my friend is considering purchasing one of the above, to test on his own, and it would be a waste of his money, if it really didnt benefit much at all for general uses

edit: ah, okay. that clears things up a bit then. thanks :) 


i usually tell people that my general rule is that if you dont know what raid is or what the levels are then you dont need it. if you're playing games and reading your email you're not going to see much of a difference.

if your storing multiple terabytes, programming, video editing on a large scale, compressing and un-compressing multigigabyte files then you will see a difference, but chances are that if you're doing those things you already have a good knowledge of raid and it's benefits.

it's not condescending, it's just the way things work out. if you aren't doing anything specialized then you dont need raid, which is itself a specialized setup.

if you WANT it of course then go for it and more power to ya.

but just like anything, the more you find out the more you may find out you dont know! like optimal cluster size for your use, stripe sizes, interface bandwidth, bios upgrades....
April 3, 2007 5:28:41 PM

I've settled with 3ware's 9650SE-4LPML. Thanks to all for replying. :D 
a c 116 G Storage
April 3, 2007 5:43:46 PM

At those prices, seems to me a SCSI RAID solution with 15k drives would be well worth considering.
April 3, 2007 6:29:08 PM

Quote:
I've settled with 3ware's 9650SE-4LPML. Thanks to all for replying. :D 


I have the 8 port version of this card and it works very well. Native support in Linux, though this requires a 2.6.19 or later kernel, and this was very important to me. It's a great little card.
a b G Storage
April 3, 2007 7:10:45 PM

Quote:
directed @ both valis and chunkymonster: im wondering though, since both of you have used/own hardware raid controllers, how do you feel they perform compared to general software based onboard raid controllers as far as an improvement in overall performance? for things such as general windows usage, games, and the like?

im genuinely wondering if there really is an appreciable difference, or if onboard software raid could nearly be considered hardware raids equal in this respect

thanks


I would not say that hardware and software RAID are equal, but for everyday gaming and general computer usage, my experience has been that using onboard software RAID0 or RAID1 is adequate. Even with just going from a single drive to software based onboard RAID0 will show an improvement in reads/writes. I have not used any onboard software RAID5 so I can't comment on that.

I think there is a general misconception about RAID 5. True hardware RAID5 is not an everyday desktop solution, hardware based RAID5 it is really a server/enterprise level solution. When you've got a file/web server with 1000's of hits requesting data, a multiple drive RAID5 (or RAID0/1) is key to ensuring stability, reliability, and minimizing downtime. Or, like those in my position, I've got a RAID5 array because I've collected over 90GB of mp3's, over 200GB of movies and videos, do some video editing and DVD authoring, as well as use it to serve/stream media to my HTPC and another computer in the house; so redundancy and quick reads are what matters to me.

For the average every day email checking, web surfing, and game playing computer user, a software based onboard RAID0 solution for the OS and a single drive for back-ups is more than enough. For the developer, CAD workstation, A/V editing machine that is encoding, decoding, compiling, and otherwise has large I/O requirements, a hardware based RAID solution is reasonable, even recommended, because it off-loads the RAID work to the RAID card and allow the system to keep it's RAM and cpu cycles for itself and apps.

Quote:
i usually tell people that my general rule is that if you dont know what raid is or what the levels are then you dont need it.
Ain't that the truth!!!!
April 3, 2007 7:33:43 PM

Quote:
I've settled with 3ware's 9650SE-4LPML. Thanks to all for replying. :D 

Be aware, you stand to get some performance improvements if you buy the battery module for this, but if you install that you may discover like I did that unless you have substantial forced air coming from the front of the case, that card may require some spot cooling.

Good luck,
-Brad
April 3, 2007 7:44:22 PM

performance improvements for buying a battery? :lol: 

Hopefully you meant having a battery is a good idea if you plan on using the 256mb of onboard cache, which isnt required but highly suggested as a power outtage will kill any data in cache without the battery and probably corrupt all kinds of data.
April 3, 2007 8:27:17 PM

Quote:
performance improvements for buying a battery? :lol:  Hopefully you meant having a battery is a good idea if you plan on using the 256mb of onboard cache, which isnt required but highly suggested as a power outtage will kill any data in cache without the battery and probably corrupt all kinds of data.

For someone who's so busy "Laughing Out Loud" over me skipping the explanation, I see you yourself failed to mention that it has nothing to do with using the onboard cache in general but plenty to do with using it to cache writes. The 3ware is one of the few that will permit caching writes without a battery backup, but that doesn't mean it's a good idea.

Something else you also glossed over is the matter of hardware RAID controllers typically defeating Windows' own write caching, which usually translates to REDUCED everyday performance when going to hardware RAID if the controller has write caching disabled.

We now return you to our regularly scheduled programming...

-Brad
April 3, 2007 8:29:13 PM

Keep talking guys, I'm reading :D 
April 3, 2007 10:09:06 PM

Quote:
performance improvements for buying a battery? :lol:  Hopefully you meant having a battery is a good idea if you plan on using the 256mb of onboard cache, which isnt required but highly suggested as a power outtage will kill any data in cache without the battery and probably corrupt all kinds of data.

For someone who's so busy "Laughing Out Loud" over me skipping the explanation, I see you yourself failed to mention that it has nothing to do with using the onboard cache in general but plenty to do with using it to cache writes. The 3ware is one of the few that will permit caching writes without a battery backup, but that doesn't mean it's a good idea.

Something else you also glossed over is the matter of hardware RAID controllers typically defeating Windows' own write caching, which usually translates to REDUCED everyday performance when going to hardware RAID if the controller has write caching disabled.

We now return you to our regularly scheduled programming...

-Brad
Funny, I'm pretty sure you just destroyed your previous statement there all on your lonesome.

I don't believe I differentiated between write/read on the controller or windows caching period, but you are more than welcome to get all delusional for mine and everyone else's personal enjoyment if you so choose. Just be sure to point out how the magical battery boosts performance next time will yah! I can always use a good laugh.
April 3, 2007 10:36:11 PM

Quote:
I don't believe I differentiated between write/read on the controller or windows caching period, but you are more than welcome to get all delusional for mine and everyone else's personal enjoyment if you so choose.

Nope, no delusion here. Just facts. I chose to go from point A to point C knowing that the OP - who wants the controller not because he needs one but because he wants one - may get lost in the noise. He was always welcome to question why, if he needed to. Then you come in thinking you're a riot with "performance improvements for buying a battery?"

Which, by the way, is a fact unless you over-ride the controller's default settings AND ignore the warnings when you do so.

You wanna fan the flames, go ahead, but just because your getting hit with your own embers is no reason to accuse anyone else of being delusional. Laugh all you want. It only feeds your childish ignorance.

-Brad
April 3, 2007 10:42:37 PM

I'm learning a lot more about hardware/software RAID & specifically 3ware's 9650SE series than I thought I would, but I don't want to get caught in the crossfire. Thanks, guys :D 
April 4, 2007 2:22:03 PM

You're not learning anything from this guy. He clearly doesnt understand that the battery plays absolutely no role whatsoever in the performance of a raid card. He slipped up and made a complete imbecile of himself with that statement and is too pig headed to correct his statement.

I need only give one reason why you would not require the raid card battery, ie., a system UPS, and his entire "over-ride the controllers defaults and warning" statement becomes a fallacy. Unless he wants to change is argument that somehow a UPS will increase you system performance or whatever other crap he wants to pull out of his butt.
April 4, 2007 4:29:23 PM

Quote:
You're not learning anything from this guy. He clearly doesnt understand that the battery plays absolutely no role whatsoever in the performance of a raid card. He slipped up and made a complete imbecile of himself with that statement and is too pig headed to correct his statement.

I need only give one reason why you would not require the raid card battery, ie., a system UPS, and his entire "over-ride the controllers defaults and warning" statement becomes a fallacy. Unless he wants to change is argument that somehow a UPS will increase you system performance or whatever other crap he wants to pull out of his butt.

You, "sandmanwn", are the one who is an imbecile. I would have given you more credit and just labeled you a moron but you are far too obstinate to be allowed that. That's OK though, there is some entertainment value to be had in slapping you back down. It's like the arcade game "Whack-a-Mole".

There's no slip-up and the facts remain. Worse yet, in your futile attempts to prove to everyone that you're not an imbecile, you're handing out horribly bad advice and performing an unfortunate dis-service to the original poster and to anyone else who may be reading.

You are correct on one thing however - that I can pull crap out of my butt. You on the other hand should pull your painfully swollen head out from yours.

Anyway, Babaghan: A UPS, though a great idea on its own merit, is absolutely no substitute for the RAID controller battery module. First of all remember the UPS itself is not fail-proof. But more importantly, a battery-equipped RAID controller can operate completely independently of the computer. You can even remove the controller and transplant it and the drives to another computer and the data will be retained. So whether you errantly hit a reset button, or Windows suffers an unrecoverable error, or the computer's own power supply fails, or the motherboard dies, the battery on the 3ware controller facilitates retention of un-written cache data for several days (3ware guarantees three days but in practice you can get well over a week) until you can get the system back up. This reduces most of the file system damage that occurs in those situations.

One important additional detail that sandmanwn lacks the capacity to offer you is that the "battery" on the 9650SE is really not just a battery. It is an electronic module with a completely independent CPU, that includes a battery. That CPU module simultaneously maintains several snapshots of the data moving through the cache so that it can back up not only the controller cache but the on-board hard drive cache(s) as well. Hence the $100+ street cost.

Another interesting detail, since sandmanwn writes rather generically "the battery plays absolutely no role whatsoever in the performance of a raid card", is that the 3ware is actually somewhat unique in allowing you to enable write caching without a battery backup installed. Most RAID controllers won't allow this at all, try as you might. Since most software performance - Windows performance in particular - benefits immensely from write caching, and since Windows assumes that RAID controllers will do any necessary write caching and disables its own in their presence, that statement of his is incorrect.

Getting back to the 3ware, the controller firmware will see whether or not you have the battery module installed and will automatically choose the most sensible settings for your RAID volume(s) when you configure it.

If on the other hand you are foolish enough to over-ride the controller defaults (I have one on the bench here and I can disconnect the battery module to show you the screen shots and the warnings) and choose to enable write caching without the BBU, and your system suffers from any variety of faults, your computer's file system will be far more likely to be seriously damaged in the event of a component or software failure than if you were running no RAID controller at all, due to the differences in caching implementations.

As you gain experience in this field you will discover that there are many ways to gain performance, but that many come with increased risk. You will encounter people - particularly in forums like THG's where overclocking is a popular pastime - who will advise that instead of buying a higher performance part you should just save the money and over-utilize the existing one by toying with settings. And this usually does work. At least for a while...

What you are seeing now from sandmanwn is a lightly veiled example of exactly that kind of advice.

Default settings are usually programmed to what they are for good reasons. Choose your risks carefully.

-Brad
April 4, 2007 6:45:12 PM

Oh looky here, hes using bold and really really long worthless paragraphs that basically reiterate the same concept over and over and over again.

Read carefully how he keeps reiterating that there is a performance increase when using the cache on the controller versus windows cache. Although this has absolutely no bearing on his statement about the battery somehow giving your controller a performance increase, he continues to use this as his defense of an utterly stupid and moronic statement. Yes there is a performance increase when using the controller cache versus windows cache, no one has denied this or even vaguely attempted to counter this point, although what this has to do with the battery however seems to be a figment in his small demented little mind.

Also, look at how he tries to divert the conversation to some random topic like overclocking which had absolutely no context in this discussion at all. You can read the entire topic from start to finish and notice that the only mention of overclocking is made by this delusional person. But somehow I have become the overclocking fool in his wacky little mind. You can check my entire post history, all 500+ posts, and youll see very quickly that never have I mentioned overclocking as a viable alternative to anything whatsoever.

Heres a test for you since you seem so ready to post something. Test your controller under full load with the battery and caching on versus without the battery with caching on and show us all how the battery magically improves performance.

Lets all go buy batteries and increase our performance everyone! :roll:
April 4, 2007 8:44:04 PM

Awesome, I get to whack the imbecile mole once again!

By the way, have you noticed that your own LSI Megaraid 150-6 doesn't even allow you to turn on write-back caching unless you have the BBU installed?

Oh that's right, silly me, you're an imbecile so you probably would never notice that. And yeah, I checked a few of your other posts. Nice to see I don't have to take this personally though, and that you are just as much an imbecile in other parts of the forum too. Yep, the big bold type and name-calling really drive the point home, that there's Enterprise and then there's enterprise, and let's not forget ENTERPRISE hard drives. Whatever. Then you sit back and watch the war erupt. Funny guy. From which psychiatric hospital are you posting?

As my civic duty I should assume the responsibility to chase you around the forums and make sure people know when you screw up. Finding the time would be the only challenge. Maybe you post intelligently on other subjects, but storage most certainly is not your strong point.

Quote:
Read carefully how he keeps reiterating that there is a performance increase when using the cache on the controller versus windows cache.

Mis-quoting me won't help you. But I know better than to expect reading comprehension from an imbecile. Sorry that you have trouble reading a paragraph longer than two lines, truly.

Quote:
Yes there is a performance increase when using the controller cache versus windows cache, no one has denied this or even vaguely attempted to counter this point, although what this has to do with the battery however seems to be a figment in his small demented little mind.

Since imbeciles have been shown to be able to eventually learn through sheer repetition, I'll take another stab at this for your benefit. Try to focus for a moment. Pop a Ritalin or a Lithium or whatever you need to simulate normalcy and wait forty minutes if you need.

1. Windows by default performs some write-back caching. (Domain Controllers excepted, but let's try to stay relevant for now)

2. Windows, when presented with a RAID controller, disables that write-back caching because it assumes the controller will do it. (obligatory anti Microsoft diatribe thoughtfully omitted here, you're welcome)

3. The RAID controller in the absence of a BBU will either dissuade (in the case of the 3ware being discussed) or prevent entirely (in the case of your own LSI controller) the activation of any write-back caching mechanism. Or so indicates the documentation.

4. So a hardware RAID controller without a BBU typically leaves you with no write-back cache anywhere.

Quote:
Heres a test for you since you seem so ready to post something. Test your controller under full load with the battery and caching on versus without the battery with caching on and show us all how the battery magically improves performance.

Why don't you try that? Oh that's right, you can't, your LSI controller won't let you be the fool that you are and permit write-back caching without a battery installed.

On the other hand, I did test the controller over a month ago, using JBOD and four different RAID configurations, with three different benchmarking utilities and a rather simplistic measurement of the time required to complete the final phase of Windows' installation, using the default settings presented by the controller during container configuration with and without the BBU, and the performance increase with the battery was remarkable.

Since I would never entertain the inherently dangerous idea of using hardware write-back cache without a BBU, and since never in my life would I recommend such a dangerous configuration to anyone, even imbeciles like yourself, that scenario was never tested.

-Brad
April 4, 2007 9:30:46 PM

Figured you would make up some sort of excuse and start another additional mindless rant. :roll:

All that writing and still cant show the battery actually has some sort of magical performance enhancements.

It has come to the point where it is pointless to listen to your garbage. You are like one of those imbeciles on television when in the course of an interview your idiotic statements come to light and the interviewer catches your crap, so in turn you babble on incessantly to cover up the facts. Instead of just admitting the statement was worded incorrectly and the performance is completely a function of added cache.

I have two HP 641 64mb smart array controllers manufactured by LSI sitting here in two ML370 systems. One is the hot-spare server of which I detached the battery unit for the controller. Ive just concluded the test and remarkably both ran the exact same performance benchmark of 155mb/s with a 4 disk ultra 320 scsi setup.

Part two of the test. I have an additional 128mb memory module bringing the total to 196. Low and behold I get a higher burst rate from the controller.

Part three. I have yet another 128-MB BBWC (battery backed write cache) addon for the controller. But what do the test show, it adds absolutely no performance AT ALL over test two and the added cache addon.

I was hoping at some point you would realize that the battery isnt the cause of the performance increase its simply the added memory that you can just as easily get for the controller without the battery that gives the performance increase. The battery does absolutely nothing other than maintain the power to the memory modules during an outtage. Thats it! It has no inherent performance properties.
April 4, 2007 10:12:52 PM

Quote:
I'm learning a lot more about hardware/software RAID & specifically 3ware's 9650SE series than I thought I would, but I don't want to get caught in the crossfire. Thanks, guys :D 

Oh, don't worry, you're not in the cross-fire. You're more like watching it on CNN, safe on your couch, far away from the bombs and blood. The only thing missing is the feel-good propaganda.

Here's a little more food for thought, meanwhile...

After a couple of decades of dicking around with everything from midrange DEC PDPs to Wintel boxes, in and out of the Big Glass Rooms and Raised Floors, I'm now responsible for several hundred servers in five domestic regions from four different vendors, their operating systems and their application software. They all are purchased, configured and maintained according to guidelines set by a number of experienced engineers, put through a peer review process and then double-checked by the hardware and operating system vendors. Per our minimum standards they all have at least two power supplies and use varying levels of RAID depending on the application necessities and cost/performance trade-offs. They all sit in rooms with two air conditioners whose power is supplied on two branch circuits from two different feed lines from two separate power company transformer vaults, supported by two UPSs and two diesel generators.

And yet our minimum standards no matter which vendor the server nor which level of RAID is implemented, all mandate the inclusion of BBU modules. It's not because we're splitting atoms or splicing genes here - it's because we don't want to see performance go utterly into the crapper.

The multi-user, multi-tasking nature of what goes into the server room and the huge amounts of data involved all magnify immensely the difference in performance. But what about you? The performance decrease you'll see at the desktop level without a BBU on the RAID controller will be bad if you use RAID 5, but tolerable if you use RAID 0, 1 or 10. Any disk-intensive multitasking work will show the shortcomings of either configuration right away. When you look at what you spend on the RAID controller hardware, it seems silly not to spring the extra few bucks on the BBU.

Be very careful of the advice you get on these forums, not just here but everywhere. Just as in face-to-face scenarios, there are people who will convincingly tell you to try things that may very well lead to disaster. But on the forums you don't have the benefit of eye contact and body language to clue you in on whether the person is knowledgeable, malicious, well-meaning but ignorant, or just a self-important buffoon who can barely conjure up an original thought but gains his only sense of self-worth from attempting to belittle others.

And don't take just my word for it! Do some independent research. It'll be good for ya', I promise.

Good luck and have fun,
-Brad
April 4, 2007 10:25:52 PM

Quote:
It has come to the point where it is pointless to listen to your garbage.

And yet you're still listening, imbecile!

Quote:
I have two HP 641 64mb smart array controllers

Very good - two more controllers that won't let you do what you're recommending - enable write-back caching without the BBU.

Nice coincidence, I just got done upgrading some poor sap's ML370 running Microsoft Small Business Server, because Exchange was crawling due to the miserable I/O write performance. Solution? BBWC module!

That was not long after having to install similar modules in three Compaq DL580s because of massive Citrix performance issues that would literally stop everything in every other session while users logged on and off.

Quote:
155mb/s with a 4 disk ultra 320 scsi setup.

Your tests are invalid. Tools like Atto, HDtach, etc. are useless for testing performance revolving around write-back cache. Next time read the documentation, if you can.

[edit...]

Got home a little while ago and rooted around for some scribbles from the testing I did several weeks ago. These are my favorite kind of benchmarks, timing the interval in a Windows XP Pro SP2 Select install, between the entry of the workgroup and initiation of the final reboot. It's a very I/O -intensive phase with lots of write operations, and I feel it is a good real-world illustration of the more or less single-threaded heavy utilization that would be consistent with typical heavy desktop use. It also avoids all the idiosyncrasies with benchmark utilities.

These tests were run on a 3ware RAID-10 set consisting of four 150GB Raptors, using the controller's default settings for the container.

BBU not installed, it took seven minutes and fifty-three seconds.

BBU installed, it took three minutes and twenty-eight seconds.

That's TWICE the performance with the BBU installed!

[...end edit]

Quote:
I was hoping at some point you would realize that the battery isnt the cause of the performance increase its simply the added memory that you can just as easily get for the controller without the battery that gives the performance increase. The battery does absolutely nothing other than maintain the power to the memory modules during an outtage. Thats it! It has no inherent performance properties.

Sorry, it's hopeless, I can't subscribe to your ill-founded fantasies, and you're wrong about the utility of the battery when it's easily demonstrated that write caching is prevented or dissuaded when the battery is absent.

And by the way, the BBU on the 3ware doesn't add any memory into the controller's cache pipeline.

Also, I never stated that the battery itself had inherent performance properties. My statement was "you stand to get some performance improvements if you buy the battery module".

You get an F for reading comprehension once again.

Whack!

Whack!
April 5, 2007 1:47:13 PM

Whats the matter with all the edits. Cant figure out how your going to change you story to make it sound like your first initially idiotic statement yet steer its way back towards the correct assumption that was pointed out by me long before.

Write caching has nothing to do with the battery. It can function without the battery. You look like a f@cking tool by doing this really long mindless rant after rant that eventually leads to the conclusion you should have come up with in the first place.

Your pretty good at the endless BS, its almost Baronesque! Next time get to the point sooner.

P.S. your test is useless as it doesnt disprove anything other than additional cache is beneficial to the controller. Your battery statement is still a mindless brain fart on your behalf.

Have a good day. :lol: 
April 5, 2007 2:52:49 PM

Quote:
Write caching has nothing to do with the battery. It can function without the battery. You look like a f@cking tool by doing this really long mindless rant after rant that eventually leads to the conclusion you should have come up with in the first place.

You saying it over and over again and punctuating it with insults and name-calling does not make it true, imbecile. Try it on your own controller!

And you get yet another "F" for reading comprehension.

Whack!

Quote:
Your pretty good at the endless BS, its almost Baronesque!

If it's so "pointless to listen to [my] garbage", why are you still listening?

Whack!

Quote:
P.S. your test is useless as it doesnt disprove anything other than additional cache is beneficial to the controller. Your battery statement is still a mindless brain fart on your behalf.


Another "F" for reading comprehension and an "F" for subject matter understanding, too. Remember, the 3ware's BBU has no additional cache.

Whack!

You should learn to use the tools properly. Being able to click on Start, Programs, etc. does not mean you're capable of making sense of their results.

Whack! Whack!

I could play this game all day!

-Brad
April 5, 2007 4:21:43 PM

You like whackin' it dont you.

Booyah!

Youve already disproved your own statement by saying the write back cache can be enabled without the use of the battery.

Quote:
I'm now responsible for several hundred servers in five domestic regions from four different vendors, their operating systems and their application software.

Is that all? Im sitting here in front of my console looking at the monitoring services for a small portion of our network that spans 5 CONTINENTS.

Five "domestic regions". Pff, what a joke. I'm doing this on a global scale. Hell, we only have 6 multi-floor datacenters and two Mobile backbone platforms.

Quote:
BBU not installed, it took seven minutes and fifty-three seconds.

BBU installed, it took three minutes and twenty-eight seconds.

That's TWICE the performance with the BBU installed!

OMG, Ohs Nozs. Just as expected. He did the test without write back cache enabled on the unit without the BBU. Your tests are scewed and the perfect example of straight FUD. Do the test properly next time and your results will be an exact mirror each time. Or are you scared to perform the test properly and disprove your own statements. You shouldnt be, youve already done it multiple times already.

Quote:
These are my favorite kind of benchmarks, timing the interval in a Windows XP Pro SP2 Select install

BWAHAHAHA OMG no you didnt! :lol:  Your performance benchmark is an Installation timing of XP? Oh, I just have to pass that around the office today.

Imbecile

Quote:
Oh that's right, silly me, you're an imbecile so you probably would never notice that. And yeah, I checked a few of your other posts. Nice to see I don't have to take this personally though, and that you are just as much an imbecile in other parts of the forum too. Yep, the big bold type and name-calling really drive the point home, that there's Enterprise and then there's enterprise, and let's not forget ENTERPRISE hard drives.

Oh I see now, you are the type of moron that thinks a Raptor is on the same level as an SAS drive. That explains everything. You should probably read the whole topic instead of skimming for what you want to see. By the end of the topic it has pretty much become unanimous that the Raptor isnt in the same league as SAS and my point held true. But hey, I suppose you read what you want to read to justify your own points. Perhaps you can pick up a battery for that ragged brain of yours, at this rate your going to need the performance boost. :p  :lol: 

And the band plays on...
April 5, 2007 4:57:50 PM

Quote:
BWAHAHAHA OMG no you didnt! :lol:  Your performance benchmark is an Installation timing of XP? Oh, I just have to pass that around the office today.

Yes, terrible of me to use a real-life and perfectly repeatable method of benchmarking, isn't it, imbecile.

Quote:
Oh I see now, you are the type of moron that thinks a Raptor is on the same level as an SAS drive.

Another "F" for reading comprehension. You're racking 'em up!

Whack!

Quote:
Im sitting here in front of my console looking at the monitoring services

That explains everything! You're the data center jockey that maintains all the equipment after people who actually know what they're doing have architected and implemented the systems therein. Well no wonder...

Yes, please do mention my techniques to your buddies who feed the tape library and baby-sit the monitoring consoles, I'm sure it will get their careers right back on track.

By the way, remember writing the following?

Quote:
You can check my entire post history, all 500+ posts, and youll see very quickly that never have I mentioned overclocking as a viable alternative to anything whatsoever.

You were right about checking your post history. It was a real education...

First, there's this post.

So much for never mentioning overclocking.

And I really like the one where you tried to help the fellow whose desktop was reloading itself when he closed his browser. You wrote "I would suggest it may be time for you to reinstall windows. Its just one of those things you need to do every so often."

Very creative advice. But I'm confused now - do you work in a data center, or for Dell's technical support services?

Also amusing was the optical lens cleaner thread, where you chose to criticize the writing skills of the article's author and wrote taking vendor claims with a "ton" of salt, is just that, a ton. makes it sound like you value their opinion greatly as if it were a ton of salt bearing down on your conscious. theres a reason why the saying is "grain of salt."

Ignoring a couple of basic grammar shortcomings that are typical to casual forum posts, your critique failed to capitalize the first letters of two consecutive sentences, missed the apostrophe in "there's", mistook "conscience" for "conscious" and worst of all, mis-interpreted the etymology of the "grain of salt" idiom. Maybe you should study that "glass houses" saying?

Actually we've crossed paths before. I first encountered your reading comprehension problems in a case choice thread. It was nice that you tried to be helpful, though, in spite of your assertion that "its difficult to find top mounted usb/firewire [...] because the usb/firewire headers are generally located at the bottom of the motherboard".

I'm enjoying my Mystique case, thanks. Apparently they managed to circumvent this massive obstacle by using [drumroll...] longer wires! I hope they didn't violate any patents by making those wires five inches longer.

Whack!


-Brad
April 5, 2007 5:37:13 PM

Quote:
Write caching has nothing to do with the battery. It can function without the battery.


That may not always be true. I have 4 MSA 20's in which the batteries failed. After that, the entire cache (read and write, ugh) was disabled. I have surely noticed a difference. :) 
April 10, 2007 2:32:40 PM

Hello bberson, sandmanwn,

thank You for the hard but very informative diskussion.

I found a lot of Informatiom, and i am now going to order an adaptec 3805 Raid WITH Battery for my families Data on our new server running under XP.

Thank and greating from cologne/germany,
Frank
April 10, 2007 3:05:52 PM

Quote:
Hello bberson, sandmanwn,
thank You for the hard but very informative diskussion.

You are very welcome, and best of luck to you.

-Brad
!