Page 1 of 3 1 2 3 LastLast
Results 1 to 20 of 50

Thread: RAID FAQ

  1. #1
    Join Date
    Jan 2003
    Location
    Mountain View, CA USA
    Posts
    2

    Lightbulb Raid Reference

    A place for RAID resources, links, etc.

    AMUG review
    SoftRAID 3
    has photos of setting up RAID. Also, you can download the quick start guide and manual for SoftRAID 3, and while unique to their program, guides you on setting up a RAID.

    Guide to SCSI
    RAID by the numbers
    RAID 5 Scaling Tests With Up To Eight Drives
    Barefeats 4-drive SATA RAID
    Miglia MediaBank HS-R - Hot Swap Raid
    HUGE Systems
    RAID-5 Hardware vs. Onboard Compared
    AMD Zone: FibreChannel array
    RocketRAID 1820A 8-channel SATA
    HighPoint
    Guide to SCSI Pages of Doom!
    Adaptec About SATA
    ATTO SCSI FAQ
    Official SCSI FAQ
    Hitachi SCSI Checklist
    Adaptec SCSI Basics
    Linux 2.4 SCSI subsystem
    Linux RAID
    SoftRAID vs. Apple
    RAID.edu Benchmarking IO
    Benchmarks FAQ
    MacIntouch RAID Reference
    MacIntouch: RAID block sizes
    PCGuide: RAID Defined
    RAID Terminology
    Storage Review Guide to RAID
    SR: Defrag RAID
    Hardmac: Software RAID
    What is RAID Primer
    Single vs RAID0 Boot drive
    benchmark 4 Raptors RAID0
    RAID DB
    Everything Mac
    Adaptec DuraStor RAID
    Storage Subsystems RAID
    Google: RAID/
    RAID How-To
    RAID Glossary
    Should I boot from RAID?
    Google: SCSI + RAID
    FirmTek SeriTek/1SE2
    HighPoint RocketRAID 8CH SATA
    Maxtor MaXLine III 300GB
    axtor DiamondMax 10 300GB
    Review 23" LCD
    _________________________
    MacGurus SATA Drives
    SATA Enclosures
    SATA Cables
    SATA Host Adapter
    HUGE MediaVault320 Data Sheet pdf

    Reducing noise:
    Seagate: Using Isolators to reduce noise, hum, and vibrations.
    Hard drive silencing

    FireWire 800 RAID probably obsolete at this point
    Last edited by TZ; 04-25-2007 at 06:31 PM.

  2. #2
    Join Date
    Feb 2001
    Location
    on the landline, Mr. Smith
    Posts
    7,787

    Lightbulb

    Quote Originally Posted by David
    ]Hello everyone, I am trying to find out if there is a need for true hardware RAID for the Mac OSX platform? I receive calls periodically asking if our Escalade RAID controllers are supported in the Mac OS and wanted to confirm the need. I also wanted to know if it is true that Mac OSX is similar to FreeBSD. I am trying to see about getting some support for our products in this OS. Thanks for your input.

    David
    Hi David,

    The short answer is YES.

    As for your question, "is...OSX similar to FreeBSD?", OS X *is* FreeBSD coupled to perhaps the best GUI in the world.

    I am an old Mac guy, not a Unix guy, so others can give you specifics if you like, or check these links to get started: http://www.apple.com/macosx/jaguar/unix.html http://www.apple.com/macosx/technologies/darwin.html

    and perhaps: http://www.osxfaq.com]http://www.osxfaq.com]http://www.osxfaq.com

    Yell if you need more details...

    These are the sort of discussions you find around forums like these every few weeks: http://www.macgurus.com/ubb/Forum1/HTML/000331.html

    As OS X evolves, and Apple makes inroads to enterprise and education server markets, there will be a demand for more and more hardware RAID. Think of all those Graphic shops, schools, and small businesses out there handling ever increasing volumes of data; they probably shy away from the $3000+ high end controller, but would find a $500 - $1000 to be a good value.

    Traditionally Apple has included SCSI controllers OEMed by ATTO or Adaptec, but the best performance and value has alway been with 3rd party products (mostly from the same venders, but also from others including Acard). Based on the history, I have little hope that Apple will provide anyhting themselves, except possibly for the top of the line hardware. Apple due to ship a rack mount device called Xraid in the near future to compliment the this http://www.apple.com/xserve , but this will most likely only be targeted at the high end markets. The low to mid level buyers (like my company) will not be able to afford it.

    Although there are high end controllers available for external storage that are OS indepentant like these http://www.bellmicro.com/fibrechanne...d_controll.asp , it seems to me that the real market would for an internal controller that could handle 0,1,3,5 and/or 0+1.

    As I am sure you are aware, Dell sells a low end RAID 5 contoller as a $500 option. Where is this for the Mac market?

    Care to take a stab at it?

    Just one mac admin opinion... thanks for asking!
    Last edited by TZ; 10-06-2005 at 09:30 AM.

  3. #3
    Join Date
    Jan 2003
    Location
    Mountain View, CA USA
    Posts
    2

    Lightbulb

    Okay, it definately sounds like the interest is there for Mac OSX support. Now my question is who would be the best resource for me to work through on having drivers developed for OSX? We do not currently have plans for direct support but I would like to find a way to develop third party support like what was done for FreeBSD. I have included the FreeBSD support pages below for reference. If anyone can assist with this project I can do my best to help on my end. David

    Development Journal and 3DM http://people.freebsd.org/~msmith/RAID/index.html#3ware
    Driver Download http://www.freebsd.org/cgi/cvsweb.cgi/src/sys/dev/twe/

    The current GUI management interface for our Escalade is 3DM and is HTML based. The configuration is BIOS based and independent of the OS. I hope this would make the Mac OSX support option more likely. If anyone can help let me know.

    The one thing I would like to add is that this is a pet project of mine based on some customers that have contacted 3ware for Mac OSX support. This is not an official program at 3ware so it would need to be a third party developed solution that would not require funding or engineering support from 3ware itself. This is how FreeBSD support was created so I was hoping for the same situation for all of you Mac users. I will personally do what I can on my end. Any leads to someone that can do this would be much appreciated. I will try contacting Softraid but I am not sure if they would charge a fee for this support.
    ______________________
    David Graas
    3ware Inside Sales Manager
    Last edited by TZ; 10-06-2005 at 10:01 AM.

  4. #4
    Join Date
    Feb 2001
    Location
    on the landline, Mr. Smith
    Posts
    7,787

    Lightbulb

    Hi again David -

    Just checked out the 3ware sight, and pricing at CDW (bought 66 imacs from them last year). The 7500-4 and 7500-8 sound like what I need. $320 to $600? Sold. I have 6 servers in this room with me that all need RAID 5 (currently software mirrors), 1 upstairs, and a video production station next door. And SerialATA too...looks good.

    Frankly, as I have only used software RAID on Macs (there are several available, included a very limited 0 or 1 which is native in OS X), I am not sure what a user needs as far a GUI for a hardware solution.

    I may be off here, but I am thinking that many Mac users would be willing to pay a reasonable price for drivers and a GUI that is as solid as their current product. My point is that there *are* a large numnber of unix savy folks out there.... RAID 5 is more about reduncancy (and all the better if it's bootable - the system is redundant too) for my needs.

    If I can afford to even consider an external box for $2500 to $4000, then surely I can afford at least a G4 466... Perhaps an oversimplification, but we keep buying new(er) servers at work every couple years *because* of increased I/O performace. CPU is secondary.

    Again I can't speak for the tiny shops/home office users on a very tight budget, but I would not flinch if 3ware only supported 733's and newer. *If* it was a solid product, with OS X support and a reasonable value...
    Last edited by TZ; 10-06-2005 at 11:38 AM.

  5. #5
    Join Date
    Feb 2003
    Location
    ex-Cupertino, CA, USA. Now HU+SRB (Europe).
    Posts
    163

    Lightbulb

    No promise, but who knows, maybe it actually WILL be a controller which is fast even on the "troublesome" machines? ;-)

    IMHO (and I am pretty firm on my opinion) a multi-channel (> 2 channels) Parallel ATA approach in ANY PowerMac is a very bad idea. It will "fry" the processor and the drives pretty soon due the cable salad. (That would be a sad lunch: fried BigMac with Cable Salad!)

    Do you imagine the PowerMac box fully stuffed with 4 or, God forbid, 8(!) 80-pin ATA cables? You may get RAID 1+0 - but also a dead processor and dead drives very soon due the insufficient air flow. Further, even now to put a 6880M or PDC20269 in a G4 "MD" is quite a challange - even more to close the door afterwards. (I cannot get rid of the idea the morons who made that G4 case never had an ATA controller in their hands).

    The biggest question for me is the speed of the card. Macs are a very specific breed. G3 B&W, G4 PCI and G4 AGP (3 PCI slots) are all damn difficult to support with a card if it aims at significant boost of the speed. A two channel PCI ATA card is not a big deal to make to be "fast enough" on these guys. A 4 channel ATA (or SATA or whatever) card is a damn difficult task because the speed users expect simply won't be there. Yes, these machines have a very difficult PCI bus to support high speed.

    The question:

    How would it look like if there is a nice & fast but somewhat pricey 3Ware ATA (or SATA) card which does work reasonably fast ONLY on G4 AGP Macs with 4 PCI slots (that is, only beginning with G4/466)? I expect that the WAST majority of the Mac users have a G4 "Sawtooth" or even the previous G4/G3 models. Very likely they won't see a SINGLE benefit of using most of the multi-channel cards offered today. What would be the market? I am scratchng my head.

    Using a proper 64-bit SCSI controller with four drives will be fine. Alternatively, a SATA controller is even better from cabling point of view and the new WD Raptor drives should be perhaps OK... Let's see, what happens.

    A 32-bit controller is capable of only 128 MB/Sec sustained because it is driven by 33 MHz PCI bus. Do not confuse the PCI speed with the ATA speed.

    TZ,

    I agree with your comments regarding SoftRaid. One thing has to be said: the RAID driver and the controller driver is not the same thing. As I understand, more often, than not the RAID driver is talking to the controller driver. So to some degree, the SoftRaid sits "atop" of a controller driver. For instance, ATTO makes a very decent controller driver for their SCSI cards, but personally I prefere SoftRaid.

    Santilli,

    According to some of my friends they got sick of "many API changes", " "manager idiocity" and that "their paycheck shrunk over the years compared to the Windows developers" - so they moved to the "evil" side. Can you blame them? The developer folks do not lilke being tied to a mothership which changes its course every few month and creates an other "reorg". More $$$ donot hurt either. Now it's hard to get a driver developer for Apple. What I know, the support for "9" is the last priority and the lurning curve for the IOKit on "X" is steep. Old folks are gone, the new ones cannot reuse their BSD drivers.
    Last edited by TZ; 10-06-2005 at 11:47 AM.

  6. #6
    Join Date
    Feb 2003
    Location
    ex-Cupertino, CA, USA. Now HU+SRB (Europe).
    Posts
    163

    Lightbulb

    There is NO such thining, again NO such thing as "ATA Hard RAID" on the Mac. Period. Basta. I can say that with 100% certainity and with 100% proof.

    What people call "ATA Hard RAID" and is called "True Hardware RAID" by the manufacturer is a lie. Their product is is a controller driver with smartly distributed I/O scheme and drive array simulation for booting. They should be proud to make this thing instead of providing false information to customers.

    Perhaps they assume, people will buy a product because it is called "True Hardware RAID" better, than if they would call it by the correct name. The overall misconception is, a "True Hardware RAID" would be a superior solution - but it is NOT.

    In the case of the well-known ATA controller the distribution of I/O is done by a software (conroller driver) means simulating a single disk drive. The engine of this action is your G4 processor, which can run up to 1.42 GHz. Only integer operations are involved in the I/O distribution process - and it is what that G4 can do the best. Don't feel sorry for your G4.
    A "True Hardware RAID" is usually performed by on-board I/O processor, completely unneeded for RAID-0, RAID-1 or RAID-10. These I/O prcessors are in theory great for RAID-5 if parity calculation (XOR engine) is needed. These I/O processors run at 33 MHz, sometimes at 66 MHz speed.

    The ATP865 chip never claims to be able to perform "True Hardware RAID" - but the controllers built with it do claim it. Do these controllers invoke some Chinese spirits to do that RAID or what?

    The lack of I/O processor and the incapability of ATP865 of performing XOR operations is the proof. Do you want me to post the Linux sources of the ATP865?

    Let's do it in small steps.

    I explained, how I would define a "True Hardware RAID". Let me do it with perhaps better detail.

    1) If the controller driver detects that there is a RAID, and addresses each individual RAID member individually, we have a software RAID.

    2) If the controller driver uses a single hardware access for all involved RAID members and the ASIC distributes the call to each individual RAID member, we have a "True Hardware RAID".

    3) If an ASIC is not capable of doing 2), the controller having only such ASIC and no other processor has no legal right to be called "True Hardware RAID".

    Are you agree with this terminology?

    It looks like there is a well-established definition what is "hardware RAID"
    and what is not.

    Here are the links to the major manufacturers:

    Promise:
    http://www.promise.com/product/produ...d=8&familyId=2
    http://www.promise.com/product/produ...=88&familyId=2
    http://www.promise.com/product/produ...104&familyId=2

    I do not see Promise mentioning "hardware RAID". Even here, where a parity engine does assist the software RAID: http://www.promise.com/product/produ...=94&familyId=2 I cannot find any.
    --------
    Neither does it HighPoint:
    http://www.highpoint-tech.com/haproducts.htm
    -------
    Nor SiliconImage:
    http://www.siliconimage.com/products/storage.asp
    (They even explicitly say, it is a SOFTWARE RAID)
    -------
    Nor Intel:
    http://www.intel.com/design/storage/...storage_spot6&
    --------
    Not even Aralion - but they have 6-channel RAID http://www.aralion.com/products/raid...timaMAX133.htm
    ------
    In contrary to their competitors, Acard proudly tells everybody, they using
    "True Hardware RAID" on both PC and Mac.
    --------
    We can find cards called "HARDWARE RAID" here:
    http://developer.intel.com/design/se...locks/srcs14l/ (this an Intel card but it is NOT based on INTL 31244. It has two Sil3112A chips and an I/O processor)

    and here:
    http://www.3ware.com/products/parallel_ata.asp
    and here:
    http://www.promise.com/product/produ...=86&familyId=2

    They have in common at least an i960RM-class I/O processor and the features I described. Nobody (aside Acard) dares to call a card without an I/O processor a "hardware RAID" card. The "Hardware RAID" cards not just feature an expensive I/O processor, but also a much more complex PCB and a price usually higher, than $250 for the simplest design.

    Wording "True Hardware RAID" is even worse. That implies, that the 12-channel 3Ware card is "just" a hardware RAID, however maybe not a real one (since it's called hardware RAID only). But a controller from Acard based on a sub-$10 ATP865 chip is a true hardware RAID?

    What Acard is doing is simply misleading the customer. It is an other
    question, is a hardware RAID worth the money or not.
    ___________________
    Some clarification: MacOS-X has a BSD layer above the kernel, but the PCI controller drivers are in Kernel. In order to live in that world, the drivers should be based on Apple's IOKit framework. It is a proprietary C++ framework based on DriverKit from NeXT's OpenStep. The learning curve of IOKit is pretty steep.
    Regarding your cards - it might be confusing, but Sonnet has both Promise PDC 20269-based and Acard ATP865-based cards. The ATP865-based card does feature a controller-based RAID driver, the other does not (and is significantly cheaper, too). Please take a look at the chipset. Perhaps all your cards are from Acard.
    Last edited by TZ; 10-06-2005 at 10:35 AM.

  7. #7
    Join Date
    Jan 2001
    Location
    Mobius Strip
    Posts
    13,045

    Lightbulb Medea debuts RAID arrays w/2Gb FC interface

    Medea debuted two new fail-safe VideoRaid disk arrays featuring a 2Gbit fibre channel interface today at NAB.

    VideoRaid FCR2 and VideoRaid FCRX2 disk arrays
    transfer rates up to 200 MB/second per channel
    multiple workstations share storage

    Both feature a 2Gbit fibre channel interface, real-time RAID controllers, built-in 4- or 8-port fibre channel hub with HSSDC connectors, optional optical fibre channel ports, removable disk drive modules, RAID controllers and power supplies, and background array reconstruction.

    VideoRaid FCR2 available in storage capacities up to 720GB (starting at $4,900). VideoRaid FCRX2 (starting at $5,600) features a 3U rack mountable design that can be populated with 5 or 10 disk drive modules with capacities to 1440GB per rack.

    Medea VideoRAID FCR

    VideoRaid scsi desktop
    disk arrays are the ideal low-cost storage solutions for digital content creators who work with DV/MJPEG non-linear editing and DVD authoring systems. VideoRaid scsi supports real-time editing and DVD authoring with products including the RT2500, DV500 Plus and Pro-ONE.

    4-drive desktop enclosure with capacities to 640 GB
    Zone Stripe Technology (ZST)
    for sustained data transfer rates up to 40 MB/second ???
    Scalable to multi-terabyte capacities
    Sneaker Net enabled
    Hardware based RAID appears as a single SCSI drive - no third party disk striping software required.
    Last edited by TZ; 03-21-2005 at 03:39 PM.

  8. #8
    Join Date
    Aug 2001
    Location
    Grangeville, ID USA
    Posts
    9,123

    Default

    ?Every major data storage facility that I have ever heard of uses RAID5. It is scalable up to huge sizes and awesome speeds. How big and how fast can you afford? They are also getting to the point that the fault protection is incredible. Some can even handle two drives in failure. Auto replacement of failed drives from mounted spares.

    ?0+1 is in my very humble opinion is for small scale. It comes no where near the fault protection and is much harder to repopulate a replaced drive. Probably cheaper to implement a given speed. Would handle small file transfers faster than a RAID5.

    ?RAID5 will handle smaller transfers at a faster speed than a RAID3. RAID3 is excellent for Audio or DV recording with huge file transfers.

    Rick

  9. #9
    Join Date
    Feb 2003
    Location
    ex-Cupertino, CA, USA. Now HU+SRB (Europe).
    Posts
    163

    Default

    Kaye,

    I did remove the chip cover (it was plain paper) of the Tempo Raid133 long ago. I was not aware that the cards newly ship with a glossy covering. Neverless, the card layout I have seem to be identical to the one on the picture and it has the ATP865 chip.

    There is actually no need to remove the paper or the glossy "goo". The card made by Acard has 3 ASIC - one big is the ATA chip (ATP865-A), the middle is the Flash Memory, usually of Winbond 29EE512 series (but that could change) and ultimately the smallest, ICS332M. The Promise-based cards do have a single ASIC because the Flash Memory is integrated into it.

    The ICS332M is actually a very interesting part. It seem to be discontinued by Integrated Circuit Systems (http://www.icst.com/) because they do not have any documentation about it anymore. Fortunately the document was cached on a broker's web site:
    http://smartdata.usbid.com/Datasheet...pdf\ics332.pdf

    Looking at it we see, that this chip is "just" a frequency generator - but it features an OTP (One time Programmable) ROM. It could be that the output frequencies of this chip are used to determine the sub-system VID of the 865. But perhaps not and it is used only to generate the ATA clock. I am not sure.
    --------------------------------------------------
    I think you completely misunderstand me. I have no trouble if you, the end-user call your OWN card "True Hardware RAID" or "Truest & Ever-best Thing Mankind Ever Made". You can name your favorite hardware you own the way you like it. You have no legal responsibility.

    The problems begin if the manufacturer uses the naming primarily to get unfair advanage. It is more, than pure semantics. As I recall, I did praise Acard several times for their work, but I also dare to criticize them here for misleading the customer. The customer in this case is not just the end-user (i.e. us) but also Miglia, Sonnet, SIIG.

    Here is one more issue which probably escaped the attention of Acard marketing folks. If they would one day produce a controller with I/O processor how would they name it? "Even More True Hardware RAID" sounds silly for me.

    Therefore I propose to call the cards this way:

    1) Not RAID-able Card. currently there is no such thing being offered (for Mac), but that horrible thing can happen (let me not to disclose the manufacturer who makes this "great" product for PC folks). It is a card which is unable to transfer simultanously on all channels due bad hardware design. Such card can support RAID, but the performance would be unacceptable. Hope, we never see such thing on Mac.
    I hope, Kaye, you won't mind if I am rough towards such product

    2) RAID-able Card - it's the Acard AEC 6280M, Tempo/100/133 from Sonnet, VST UltraTek/66, etc. Software RAID works fine, both channels active during the transfer, no data loss.

    3) RAID Card. I't's like 2), but the RAID driver is at the controller level. Acard AEC 6880M and its incarnations are the example.

    4) True Hardware RAID - this would be the Ultimate Thing (perhaps), but I am not aware anybody ever made it for the Mac.
    ---------
    Finally, Kaye - there is ABSOLUTELY no way to determine based on performance data of a particular card, it is a "True Hardware RAID", RAID or RAID-able Card. A good software RAID should be competitive with a "True Hardware RAID" in RAID-0, RAID-1 and RAID-10 on modern machines because of the fast main processor. The story could change with RAID-3 and RAID-5.

    Therefore please do not expect me to provide you with some URL-s where somebody would prove, AEC 6880M has performance issues because it is ultimately software-based RAID card. Instead, I can prove you it either simple way (looking at the ATP 865 ASIC spec and speculating about it) or somewhat hard way. The "somewhat hard way" would involed both of us starting from the same AEC6880M firmware and looking with the MacsBug how the I/O request is handled. It actually sounds a bit more scary than it is in the reality. The I/O is usually done very soon after the PrepareMemoryForIO() call which very easy to locate. After setting the breakpoint at PrepareMemoryForIO() we would just watch, when does the driver set the Scatter/Gather list pointer(s). It is very easy since the position of S/G list register is standartized (in the case of ATP 865 it is at BaseReg#5+0x04 for channel 0 or BaseReg#5+0x0C for channel 1). If the driver sets these registers individually, sends the 0xCA or 0xC8 command to both Command Registers individually and starts the busmaster operations individually on both channels, we have a software RAID masqueraded as H/W RAID. If it ignores the standard ATA registers of ATP865 and calls some "magic" "who-knows-what-the-heck" register perhaps submitting there the pointer to the entire command, we have a true hardware RAID.

    If you would like to do it, let's do it and let's see, I am wrong accusing Acard of misrepresenting their product or not.

  10. #10
    Join Date
    Jan 2001
    Location
    Mobius Strip
    Posts
    13,045

    Lightbulb Removing SoftRAID (2.2.2, OS 9) driver

    People using RAID often choose SoftRAID. However, there are times when you want to remove the SoftRAID driver, and it isn't obvious how. Attempts to initialize a drive with Drive Setup aren't allowed and you get an error message.

    The most common reason was that people wanted to install OS X on their drive.

    This is what SoftRAID folks recommend:
    quote:
    NOTE: SoftRAID uses a single driver to control all devices, rather than load a separate driver for each device. You may find that after installing SoftRAID on a device, it is difficult to remove the driver with other utilities.

    SoftRAID allows you to completely remove its driver, by holding the option key when selecting Initialize. Then select quick initialize, and SoftRAID will leave the device blank with no driver. AFTER a reboot, any formatting utility will be able to control the drive or cartridge. www.softraid.com
    Last edited by TZ; 03-21-2005 at 03:38 PM.

  11. #11
    Join Date
    Jan 2001
    Location
    Mobius Strip
    Posts
    13,045

    Lightbulb

    RAID Basics

    RAID is an acronym that refers to "Redundant Array of Inexpensive Disks," often abbreviated as "an array."

    Using multiple drives, it is possible to create volumes which distribute data across several drives. The reasons for doing so include speed, data security, and data volume capacity.

    While individual mechanisms in an array, meaning individual drives, all hold data from a volume created using one of the forms of RAID, no single drive holds all the data in the volume. Even though you may have up to thirty drives in an array, it is an option to stripe or mirror them into one logical unit, meaning one Macintosh HFS or HFS+ volume, mounting on your desktop. Of course, you can always partition, if multiple volumes are more useful for your particular application.

    As mentioned, arrays are useful for two primary reasons: speed and data redundancy. One other useful aspect of RAID is that it is possible to create very large volumes: we have ourselves built 584GB arrays using eight 73GB LVD Cheetah drives, striped into one logical volume.

    On the Macintosh, arrays are commonly implemented primarily either in software, or in hardware. Both have strengths and weaknesses.

    Software-based RAID commonly take the form of RAID 0 or RAID 1. Assuming mature, sophisticated RAID formatting utilities are used, software-based RAID is fast, simple, robust, reliable, and relatively inexpensive.

    Hardware RAID provides the means to implement RAID 0, RAID 1, or RAID 5 using a hardware abstraction backplane which offloads processor-intensive tasks from the CPU to enhance performance and enable fault-tolerant features which are not feasible using RAID software alone. Appropriate primarily for the enterprise, hardware RAID is relatively expensive, putting it beyond the reach of the average power user and most small businesses.

    Levels of RAID Pertaining to Macintosh

    * RAID 0, aka "striping," requires a minimum of two drives used as single or multiple HFS or HFS+ logical volumes, yielding very high performance, but comparatively lower data security. Adding drives in pairs increases logical volume capacity, as well as thruput. RAID 0 scales up nicely, to the limits of PCI bandwidth, the limits of SCSI enclosure technology, and the maximum number of devices permissible on a single bus. Bus duplexing can be used to bond channels, theoretically enabling dual-channel RAID incorporating up to 30 distinct drive mechanisms and over two terabytes (2190GB) in logical volume capacity. Used properly, with high-quality SCSI accelerators, hard drives, and appropriate cabling, RAID 0 can be robust, reliable, and extremely fast.

    * RAID 1, aka "mirroring," requires a minimum of two drives containing redundant data, yielding excellent data security, but poorer performance compared to RAID 0, particularly during sustained writes to disk. Sustained reads, conversely, are comparatively fast, making RAID 1 extremely useful for file server applications. Drives must be used in pairs in a "master-slave" paradigm. RAID 1 also scales well, and it is possible to build RAID 1 incorporating 30 drives on dual SCSI buses yielding up to 1095GB in logical volume capacity with current hardware, with the limitation that such a configuration would require 15 separate logical volumes, each including two drives per mirror. Two drives are used to yield the logical volume capacity of one drive, since redundant data is mirrored on two separate devices simultaneously. While literally twice as expensive as RAID 0 when considered in terms of megabyte per dollar, used appropriately, RAID 1 delivers superb data security at a price point far cheaper than hardware RAID.

    * RAID 0/1, combining the best features of RAID 0 and RAID 1, involves the mirroring of a striped volume, yielding the best compromise in terms of performance with data redundancy. An example would be 14 drives striped as a single HFS+ RAID 0 volume, then mirrored, requiring a total of 28 drives to yield the capacity of fourteen drives. RAID 0/1 delivers the best combination of performance and data security theoretically attainable. Not yet feasible on the Macintosh using software, but forthcoming from SoftRAID LLC in a future update to SoftRAID.

    * RAID 5, requiring a minimum of three drives (four, with a hot-swap spare), involves data striping across multiple drive mechanisms with parity data distributed among drives, enabling background regeneration of absent data. More efficient than RAID 1, three drives in a RAID 5 configuration yield the data capacity of two drives, with one additional drive used as a hot-spare in the event of drive failure. RAID 5 scales nicely, though enclosures with intelligent backplanes are expensive. While software-based RAID 5 implementations exist, RAID 5 can only be implemented reliably on the Macintosh in hardware.

    Rear inside-view of a Gurus-built 144GB RAID 0.

    Eight ST318203LW 18GB LVD Cheetahs mounted inside an MAP5081R enclosure with dual hot-swappable redundant power supplies. Cabling is Granite TPO. Striped in a 144GB Level 0 RAID, this array delivered MacBench 5 Disk Scores of 2750, and sustained thruput of 59-69MB/sec, peaking at 79MB/sec, using SoftRAID 2.1.6 drivers and an Initio Miles2.

    RAID Terminology

    * Array: Two or more hard drives addressed as a single logical unit, or volume.
    * Duplexing: Bonding two separate SCSI channels so multiple disks can be addressed as a single logical volume. Useful both in RAID 0 and RAID 1, and readily implemented using a dual-channel SCSI card, or two independent boards in a six slot PCI Power Macintosh.
    * Fault-tolerance: Refers to the ability to keep a computer system operating despite the failure of a hard drive. Virtually synonymous with RAID 5, also refers to redundant auto-switching power supplies.
    * Hot-spare: Refers to a drive which is preformatted and spun up and ready to accept reconstructed parity data from other drives in a RAID 5 array.
    * Hot-swap: The ability to remove and replace a failed drive in a RAID 5 array without interrupting data transfers to the logical volume.
    * Mirroring: Duplication of data from a primary drive on a secondary drive, aka RAID 1.
    * Parity: A method of recreating the data contained on a failed drive in a RAID 5 disk array.
    * Striping: Distribution of data across multiple pairs of drives to enhance performance.

    Another source of helpful information is www.SOFTRAID.com Support - Setup and Installation.
    Last edited by TZ; 10-06-2005 at 10:03 AM.

  12. #12
    Join Date
    Feb 2003
    Location
    ex-Cupertino, CA, USA. Now HU+SRB (Europe).
    Posts
    163

    Lightbulb

    You are 100% correct. It's a software RAID residing in ROM. Nothing wrong about it only...

    1) The performance of the "regular" card is "cripled" compared with the the "RAID"-card

    2) The 6880M is called "True Hardware RAID" on Acard web side. There is one fine thing in this country and it's name is "truth in advertisement". Most Mac users do not browse MacGurus site and it is a challenge for them to understand this. All they know, the "hardware RAID" card is better than the "software RAID" card esp. if it is "true hardware RAID". This foul business practice did cause us 7-digit damage - and it's still on-going.

    By accident today I got what I believe a simple proof, what the "True Hardware RAID" in reality is.

    Who can, please check it out.

    You will need:

    - One AEC 6880M "true hardware RAID" controller
    - One AEC 6820M "regular" controller
    - Latest Acard firmware and if MacOS-X used, latest drivers with "deep sleep" support.
    - A pair of (possibly identical) ATA drives for RAID-0 set-up
    - Any G4 machine with AGP graphics (earlier machines won't do the trick because they do not have "deep sleep" support)
    - MacOS-9.x or MacOS-X 10.2

    The procedure:

    1) Set up the "true hardware RAID" under MacOS-9 or MacOS-X by using the 6880M DIP switch
    2) Put the machine into "deep sleep". Deep sleep entirely removes the power from the PCI bus, so you CAN physically remove the "True hardware RAID" card from the machine.
    3) Substitute the "True hardware RAID" card with the AEC 6820M - be carefull, the non-RAID card has to be put in the same slot where the RAID card was!
    4) Wake the machine up
    5) Enjoy the "True hardware RAID" while using the regular card.

    Therefore: the "true hardware RAID" is just inside of the driver which gets fooled and fails to detect, the supposingly "RAID" hardware was removed and substituted with the "non-RAID" hardware.
    Last edited by TZ; 10-06-2005 at 10:56 AM.

  13. #13
    Join Date
    Jan 2001
    Location
    Mobius Strip
    Posts
    13,045

    Default

    quote:
    The story so far: RAID

    Frank Hayes, Computerworld

    19/11/2003 09:44:13

    David A. Patterson led the team at the University of California, Berkeley, that developed the idea of RAID storage. In an interview with Frank Hayes, Patterson recalled the beginnings of his RAID project in 1987.

    "We had just been working on RISC processors, and we consciously said, 'Processors are going to start getting fast, improving faster than they have in the past. So what are we going to do about I/O?' That was one motivation.

    "The other one was that Randy Katz (one of Patterson's colleagues at Berkeley) got a Macintosh, and it had a hard disk in a separate box next to it. And he said, 'That's kind of interesting; here's a much smaller disk than I'm used to. What could we do with that as a building block?'

    "So we held a graduate course where we started off with some rough ideas, and then we and the graduate students -- Garth Gibson, Pete Chen, Ed Lee, Ann Chevernak, Ethan Miller -- met and talked and read papers, and the ideas evolved from there.

    "But when we tried to tell people our ideas, they couldn't understand. They'd say, 'Oh yeah, that's the same thing that IBM's been doing forever in terms of mirroring.' Or, 'Oh yeah, Thinking Machines, they've got a product in this area.' And so when we tried to explain things, they assumed what we'd done had already been subsumed by other work.

    "That motivated us to write a paper ('The Case for Redundant Arrays of Inexpensive Disks'). It advocated that we should be replacing these big disks by lots of small disks. Basically, a big, relatively thick disk that has to spin fast is much less efficient than lots of small disks, and we get all these benefits in terms of volume and footprint and power. We submitted the paper to the database conference SIGMOD, and Garth Gibson (the lead graduate student on the project) and I went to a short course that was given at Santa Clara University by Al Hoagland, who was kind of the godfather of the disk industry. We came with 20 or 30 copies of our report and handed it out at that meeting, and that was a good thing to do. The paper just clicked. It was a good time, I guess, for that set of arguments.

    "We built the RAID I (in 1989) to try the ideas in software. For RAID II (in 1993), we said, 'Let's try to build a high-performance I/O system that connects over a network.' Then at the end of the project, we had a little demo where we pulled the disk out and the thing kept working.

    "We were still performance-oriented, thinking RAID was for performance, so we were shocked to see somebody write this up in Byte magazine. The PC community was obviously not so performance-oriented as it was dependability-oriented, and they thought, Hey, less-expensive dependable computing.

    "It really just took off after that. EMC (Corp.) decided to build mainframe storage out of PC disks. Compaq (Computer Corp.) had RAID early, and Data General (Corp.). And of course IBM had its own. We didn't know IBM had its own RAID 5 set of ideas in the AS/400 line. IBM had completely independently done the same RAID part of the ideas but used large disks.

    "One of the surprises about RAID was it was so expensive. The I in the name when we coined the term was for inexpensive disks. But the system was so expensive, that was kind of awkward for marketing people. So Randy blessed the change to independent for I. Since the RAID boxes weren't cheap, that was probably a better name.

    "The current project I'm working on is ROC, Recovery-Oriented Computing. With the RAID stuff, we were always thinking performance, but obviously, dependability is the reason people are doing it. People get mad if their program crashes, but they just go berserk if they lose data. The ROC philosophy is recovering fast when outages happen. That's a different engineering ethic. Hardware will break, software has bugs, people will make mistakes. And if you believe that, then it makes sense to recover fast, rather than just try to make things that never break."
    The story so far: RAID

  14. #14
    Join Date
    Jan 2001
    Location
    Mobius Strip
    Posts
    13,045

    Lightbulb

    ATTO Mac OS X Support
    http://www.attotech.com/macsup.html

    Unleash the power of Mac G5 with optimized ATTO host adapters.
    Click here to learn which ATTO host adapter is best suited for your video system.
    Compatibility matrix:
    http://www.attotech.com/epciview.html
    http://www.attotech.com/unleash.html

    ATTO Hot Topics:

    Will ATTO ExpressPCI host adapters with a standard 33 MHz or 66 MHz PCI interface (not PCI-X) work in a G5 system? Certainly. ATTO has released updated drivers to support our Ultra3 and FC 3321 host adapters. These drivers are available on our driver download site.

    quote:
    How does using a PCI adapter in a PCI-X slot affect performance? The three G5 PCI-X slots are configured as two independent banks. One bank (comprising one slot) operates at 133 MHz, and the remaining bank (comprising two slots) operates at 100 MHz. Refer to your G5 documentation to properly identify the location and performance characteristics of these slots for your G5 model.

    Using a PCI adapter in any of the PCI-X slots will result in a performance degradation for that specific bank since the bus will revert to PCI speeds, either 33 MHz or 66 MHz PCI, depending on the bus and adapters that you are using. As such, the UL3D-66 or UL3S-66 is the more appropriate match for the high-end G5 with a PCI-X bus because these will result in the least amount of performance degradation.

    The standard UL3S/UL3D host adapter is appropriate for the entry-level G5 because both support 33 MHz PCI, which results in a good match. Each of our PCI Fibre Channel adapters supports 66 MHz PCI operation. For best performance, use these with the high-end G5 systems. However, they will work in the entry-level G5 at 33 MHz speeds.

    Unfortunately, the ExpressPCI DC and PSC SCSI host adapters are not supported in the G5 because of signal voltage limitations.

    Additional questions about G5 support should be directed to our technical support staff (available between 8am and 8pm ET by telephone at 716-691-1999 or e-mail at techsupp@attotech.com). They will be able to answer additional questions about G5 support and guide you to the best host adapter solution for your system. http://www.attotech.com/hot.html
    Last edited by TZ; 10-06-2005 at 11:22 AM.

  15. #15
    Join Date
    Jan 2001
    Location
    Mobius Strip
    Posts
    13,045

    Default

    quote:
    Do the PowerDomain SCSI cards support the new Apple PowerMac G5?

    Question: Does Adaptec support the new PowerMac G5 introduced by Apple? These PowerMac G5s have PCI or PCI-X compatible slots and I need to install a SCSI card, do the Adaptec SCSI cards work in these systems.

    This information applies to the following product(s):
    - AVA-2906, Power Domain 2930U, Power Domain 2940 series, Power Domain 3940 series, Power Domain 2940U2W, AHA-8945, Power Domain 29160, Power Domain 29160N, Power Domain 39160 - PowerDomain PCI SCSI Card(s)

    This information applies to the following Operating System(s):
    - Mac OS X 10.2.x

    Answer The Adaptec SCSI Card 2906 (ASC-2906, AVA-2906) and Adaptec PowerDomain 2930 (APD-2930, APD-2930C, APD-2930CU, AHA-2930CU) are 5 V PCI cards and will not fit into the PCI or PCI-X slots in the new Apple Power Mac G5 according to page 75 of Apple's Power Mac G5 Developer documentation. Due to this hardware incompatibility the 2906 and 2930 will not function in these system. Adaptec has no acceptable replacement at this time.

    The Adaptec PowerDomain 29160N (APD-29160N), PowerDomain 29160 (APD-29160) and PowerDomain 39160 (APD-39160) are Universally keyed and will fit into either 5 V or 3.3 V PCI slots, so these cards will fit into PCI or PCI-X slot but have only been tested minimally in the Power Mac G5. No new issues have been seen in this configuration.

    The HotConnect 8945 (AHA-8945), PowerDomain 2940 (AHA-2940), PowerDomain 2940W (AHA-2940W), PowerDomain 2940UW (AHA-2940UW), PowerDomain 3940 (AHA-3940), PowerDomain 3940UW (AHA-3940UW) are not support(-ed) in any system since the the Blue and White G3.

    G5 Compatible Adaptec controllers

    Adaptec Mac OS X Support and G5's
    --------------------

    Question: Do existing drivers for Mac OS X available on the Adaptec web site for the Adaptec PowerDomain SCSI cards function under Mac OS X 10.3 (Panther)?

    This information applies to the following product(s):
    - AVA-2906, Power Domain 2930U, Power Domain 29160, Power Domain 29160N, Power Domain 39160 - PowerDomain PCI SCSI Card(s)

    This information applies to the following Operating System(s):
    - Mac OS X 10.3.x (Panther)

    Answer The currently existing drivers for Mac OS X available on the Adaptec web site for the Adaptec PowerDomain SCSI cards have been tested minimally with Mac OS X 10.3 (Panther). As of today we have found no new issues with these drivers. Ask Adaptec


    5 V PCI cards and will not fit into the PCI or PCI-X slots in the new Apple Power Mac G5 according to page 75 of Apple's Power Mac G5 Developer documentation. Due to this hardware incompatibility the 2906 and 2930 will not function in these system. Adaptec has no acceptable replacement at this time. Apple G5 Hardware Developer Notes (.pdf)

    The issue exhibited in some G4's with Deep Sleep (aka, Advanced Power Management) with the SeritTek/1S2 from FirmTek is also noted when using Adaptec SCSI controller 29/39160's: G4 Digital Audio Macs: Deep Sleep

    [This message was edited by TZ on Fri February 13, 2004 PT at 5:00.]

  16. #16
    Join Date
    Jan 2001
    Location
    Mobius Strip
    Posts
    13,045

    Default

    A "Universal" design that will plug into a 32-bit slot keyed for 5V.
    Many new cards do not fit this category.

    You can tell a universal card (all 64-bit here) by looking at the card edge.

    A universal looks like:

    === ========= === ======= (64-bit 3.3v and 5V)

    While a 3.3V only card looks like:

    === ============= ======= (64-bit 3.3v-only)

  17. #17
    Join Date
    Jan 2001
    Location
    Mobius Strip
    Posts
    13,045

    Default

    LSI developing 4Gb/s Fibre Channel HBA

    LSI Logic today announced that it will deliver a 4Gb/s Fibre Channel host bus adapter (HBA) product family, supporting the Fibre Channel Industry Association's (FCIA) vote to extend 4Gb/s Fibre Channel from a drive interconnect to HBAs and switched SAN (Storage Area Network) fabrics. LSI Logic says its 4Gb/s HBAs are expected to be available in mid 2004, and will be bundled with MyStorage management software, which is designed to simplify SAN management and reduce SAN installation time. They will be available in single and dual port versions with a PCI-X bus interface, will support 1-2-4GB/s operations, and will support Mac OS X.

    http://www.lsilogic.com/products/hos...ers/index.html

    Fibre Channel PCI-X Host Connect

    LSI Logic has been on the leading edge of Fibre Channel development for over 10 years. Today, our Fibre Channel Host Bus Adapters (HBAs) provide the performance, reliability and ease-of-use simplicity to satisfy the modest one or two server SAN or the most demanding enterprise SAN. These Host Bus Adapters support all major topologies and operating systems with full duplex communications on one, two, or four channels.

  18. #18
    Join Date
    Feb 2001
    Location
    on the landline, Mr. Smith
    Posts
    7,787

    Default

    Just a note:

    The Fibre Channel Host PCI card that was installed by Apple in or G5 Xserve is branded LSI. The optional PCI RAID 5 SATA controller is reported to be from LSI as well.
    "Imagine if every Thursday your shoes exploded if you tied them the usual way. This happens to us all the time with computers, and nobody thinks of complaining." -- Jef Raskin

  19. #19
    Join Date
    Jan 2001
    Location
    Mobius Strip
    Posts
    13,045

    Default

    RAIDKing (MacRAID is the same company) 600 series unit.

    When I first bought the unit we had some trouble with it when attached to a Windows NT box. We moved it to a Mac (G4 dual 500 with an ATTO ultra160 adapter) and it has been a fantastic performer from the outset. Absolutely no problems despite heavy day-to-day use (configured with 12 120GB drives - 10 as RAID 5 and 2 hot spares).

    The RAIDKing folks are responsive and the equipment is definitely a good value for the money.

    They have just released a new model based on SATA drives and SCSI Ultra 320 host connection which I am looking at for expansion.

    MacInTouch: OS X Server RAID Report

  20. #20
    Join Date
    Jan 2001
    Location
    Mobius Strip
    Posts
    13,045

    Default iSCSI Makes SANs More Affordable

    May 12, 2004

    iSCSI Makes SANs More Affordable


    By Eric Hall, Network Magazine

    http://www.storagepipeline.com/20300506


    Storage area networks have proven to be a promising technology, simplifying the management of large and complex storage systems. But they come at a high cost: First-generation storage area networks (SANs) depend on Fibre Channel (FC) networks, which require laying new cables, learning new skills, and buying specialized switches. Unfortunately, this has made SANs hard to justify for all but the largest installations.

    Internet SCSI (iSCSI) could change all of that. Because iSCSI runs over standard TCP/IP links, it's easier to build, manage, and justify than FC, making SAN technology more affordable and accessible for everyone. A small iSCSI SAN can be built on top of an existing network infrastructure and even use ordinary Windows or Linux servers as remote storage arrays. Larger networks can also benefit since iSCSI turns the SAN into a low-cost commodity.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •