Re: large ide raid system
On Tue, 11 Jan 2000, Thomas Waldmann wrote: Cable length is not so much a pain as the number of cables. Of course with scsi you want multiple channels anyway for performance, so the situation is very similar to ide. A cable mess. Well, it is at least only a half / third / ... of the cable count of "tuned" single-device-on-a-cable EIDE RAID systems (and you don´t have these big problems with cable length). I didn´t try LVD/U2W SCSI yet, but using UW SCSI you can put e.g. 2 .. 3 IBM DNES 9GB on a single UW cable (these are FAST while being affordable each one does ~~15MB/s) without loosing too much performance. Did anybody measure how this is with U2W/LVD ? How is performance when putting e.g. 4, 6 or 8 IBM DNES 9GB LVD on a single U2W channel compared to putting them on multiple U2W channels ? I do not know what kind of RAID level you are discussing, because haven't followed the thread but here are some results for two IBM DMVS09V drives on a single U2W channel with Linux software-RAID1 (with raid1 read balance patch applied). These are tiotest results. Size is MB, BlkSz is Bytes, Read and Write are MB/sec, Seeks are Seeks/sec MachineDirectory Size(MB) BlkSz Threads Read Write Seeks --- --- - --- - -- --- --- icesus-r1p /mnt/ 800 4096 1 25.649 15.296 315.806 icesus-r1p /mnt/ 800 4096 2 33.970 15.528 610.314 icesus-r1p /mnt/ 800 4096 3 37.360 15.684 541.071 icesus-r1p /mnt/ 800 4096 4 35.351 15.447 629.723 icesus-r1p /mnt/ 800 4096 5 41.068 15.285 632.911 icesus-r1p /mnt/ 800 4096 6 40.818 15.131 624.352 icesus-r1p /mnt/ 800 4096 8 37.488 15.157 701.016 Not bad considering the bus was 40Mbytes/sec Here are some info of the hardware: Host: scsi0 Channel: 00 Id: 00 Lun: 00 Vendor: IBM Model: DMVS09V Rev: 0100 Type: Direct-AccessANSI SCSI revision: 03 Host: scsi0 Channel: 00 Id: 02 Lun: 00 Vendor: IBM Model: DMVS09V Rev: 0100 Type: Direct-AccessANSI SCSI revision: 03 Adaptec AIC7xxx driver version: 5.1.20/3.2.4 Compile Options: TCQ Enabled By Default : Enabled AIC7XXX_PROC_STATS : Disabled AIC7XXX_RESET_DELAY: 5 Adapter Configuration: SCSI Adapter: Adaptec AIC-7890/1 Ultra2 SCSI host adapter Ultra-2 LVD/SE Wide Controller PCI MMAPed I/O Base: 0xe300 Adapter SEEPROM Config: SEEPROM found and used. Adaptec SCSI BIOS: Enabled IRQ: 10 SCBs: Active 1, Max Active 24, Allocated 30, HW 32, Page 255 Interrupts: 420588 BIOS Control Word: 0x10a6 Adapter Control Word: 0x1c5e Extended Translation: Enabled Disconnect Enable Flags: 0x Ultra Enable Flags: 0x Tag Queue Enable Flags: 0x0045 Ordered Queue Tag Flags: 0x0045 Default Tag Queue Depth: 8 Tagged Queue By Device array for aic7xxx host instance 0: {0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0} Actual queue depth per device for aic7xxx host instance 0: {8,1,8,1,1,1,8,1,1,1,1,1,1,1,1,1} Statistics: (scsi0:0:0:0) Device using Wide/Sync transfers at 40.0 MByte/sec, offset 31 Transinfo settings: current(12/31/1/0), goal(12/31/1/0), user(12/127/1/0) Total transfers 194067 (98491 reads and 95576 writes) (scsi0:0:2:0) Device using Wide/Sync transfers at 40.0 MByte/sec, offset 31 Transinfo settings: current(12/31/1/0), goal(12/31/1/0), user(10/127/1/0) Total transfers 200068 (103008 reads and 97060 writes) read_ahead 1024 sectors md0 : active raid1 sdb1[1] sda1[0] 4417728 blocks [2/2] [UU] md1 : active raid1 sdb2[1] sda2[0] 4538240 blocks [2/2] [UU] unused devices: none
Re: large ide raid system
On Fri, 14 Jan 2000, Keith Underwood wrote: I also experimented with the master/slave setup, and my recollection is also that it was worse than half the performance of the master only setup. Keith I do not know about performance, but if you build raid array using masters and slaves on same channel, it will lack redudancy because of if master dies, it will take slave with it ? So raid1 or raid5 using masters AND slaves is totally unwise? Some IDE guru could enlighten us on this. Is using slave safe on raid arrays considering redudancy ? -- Mika On Tue, 11 Jan 2000, Jan Edler wrote: On Tue, Jan 11, 2000 at 04:25:27PM +0100, Benno Senoner wrote: Jan Edler wrote: I wasn't advising against IDE, only against the use of slaves. With UDMA-33 or -66, masters work quite well, if you can deal with the other constraints that I mentioned (cable length, PCI slots, etc). Do you have any numbers handy ? Sorry, I can't seem to find any quantitative results on that right now. will the performance of master/slave setup be at least HALF of the master-only setup. I did run some tests, and my recollection is that it was much worse. For some apps cost is really important, and software IDE RAID has a very low price/Megabyte. If the app doesn't need killer performance , then I think it is the best solution. It all depends on your minimum acceptable performance level. I know my master/slave test setup couldn't keep up with fast ethernet (10 MByte/s). I don't remember if it was 1 Mbyte/s or not. I was also wondering about the reliability of using slaves. Does anyone know about the likelihood of a single failed drive bringing down the whole master/slave pair? Since I have tended to stay away from slaves, for performance reasons, I don't know how they influence reliability. Maybe it's ok. Jan Edler NEC Research Institute
Re: large ide raid system
I do not know about performance, but if you build raid array using masters and slaves on same channel, it will lack redudancy because of if master dies, it will take slave with it ? So raid1 or raid5 using masters AND slaves is totally unwise? I can only speak from experience. I have 3 production raid5 servers on 2 ide channels. Over the last few years there have been 2 deaths on different servers, both involving the master slave channel. In both cases the data reduncancy was fine and no data was lost due to the drive failure. I can see how data loss is possible but have not experienced it in this environment. BTW, moving the systems to 1 disk per controller anyway. Michael [EMAIL PROTECTED]
Re: large ide raid system
Cable length is not so much a pain as the number of cables. Of course with scsi you want multiple channels anyway for performance, so the situation is very similar to ide. A cable mess. Well, it is at least only a half / third / ... of the cable count of "tuned" single-device-on-a-cable EIDE RAID systems (and you don´t have these big problems with cable length). I didn´t try LVD/U2W SCSI yet, but using UW SCSI you can put e.g. 2 .. 3 IBM DNES 9GB on a single UW cable (these are FAST while being affordable each one does ~~15MB/s) without loosing too much performance. Did anybody measure how this is with U2W/LVD ? How is performance when putting e.g. 4, 6 or 8 IBM DNES 9GB LVD on a single U2W channel compared to putting them on multiple U2W channels ? Thomas
Re: large ide raid system
Thomas Davis wrote: JMy 4way IDE based, 2 channels (ie, master/slave, master/slave) built using IBM 16gb Ultra33 drives in RAID0 are capable of about 25mb/sec across the raid. nice to hear :-) not a very big performance degradation Adding a Promise 66 card, changing to all masters, got the numbers up into the 30's range (I don't have them at the moment.. hmm..) I was also wondering about the reliability of using slaves. Does anyone know about the likelihood of a single failed drive bringing down the whole master/slave pair? Since I have tended to stay away from slaves, for performance reasons, I don't know how they influence reliability. Maybe it's ok. When the slave fail, the master goes down. My experience has been, when _ANY_ IDE drive fails, it takes down the whole channel. Master or slave. The kernel just gives fits.. hmm .. strange .. I got an old Pentium box, and disconnected the slave and the raid5 array continued to work after a TON of syslog messages. Anyway, I agree that the master-only configuration is much more reliable from an electrical point of view. I was wondering how much IDE channels linux 2.2 can handle, can it handle 8 channels ? would an Abit with 4 channels + 2 promise ultra 66 cards work ? or a normal BX mainboard (2 channels) + 3 promise ultra 66 ? thanks for infos, Benno.
Re: large ide raid system
[EMAIL PROTECTED] wrote: john b said: Performance is pretty good - these numbers are for a first generation smartcan (spring '99) these numbers are also useless since they are much too close to your ram size, and bonnie only shows how fast your system runs bonnie :) a better benchmark would be to see how this runs with multiple concurrent accesses to even larger files. perhaps something like tiotest? Good point, I'll try re-running it with 500MB - 2GB file size. Just need to keep other processes from doing raid i/o while I'm testing - it *is* a production machine and has been running quite happily since June... but even with bonnie getting more cpu time, the speed did not seem terribly different. this makes me wonder about how fast the smartcan's logic really is... True, but then again, it was a first generation smartcan setup...:-) speech type=rant i cant tell you about the division of responsiblility, but i can tell you i keep closed source, binary modules out of my kernel. i have enough problem with vendors who dont release specs to their equipment, let alone those who ride on the backs of the kernel developers by taking advantage of open code, and keeping theirs closed. vote with your dollars i say. /speech I agree with you about the open source in the kernel... one thing you might want to consider though is trying to motivate companies into developing / releasing products for the Linux environment. In virtually every other market, the *standard* is closed, binary modules drivers. How do you get these companies to open up and release their intellectual property? By refusing to buy their products when they attempt to enter the market (you're basically saying, its my way, or no way)? How about *working* with them, buying their initial products. Once you *prove* there is a market for them, they'll listen to you about the market culture... if they still refuse, don't buy anything else from them... John
Re: large ide raid system
Brian Grossman wrote: RZ RZ Of course this is not the only thing the affects speed. Other issues that RZ make our units fast is the PCI bus which is 133Mbs and DMA directly to RZ drives. It is however, still unclear whether it's safe to run reiserfs on a raidzone. I have a question about that out to Colin. I tried the ext3 alpha code once.. it went BOOM almost on immediately on first boot.. But, that may be the code, and not RZ's fault. -- +-- Thomas Davis| PDSF Project Leader [EMAIL PROTECTED] | (510) 486-4524 | "Only a petabyte of data this year?"
Re: large ide raid system
Benno Senoner wrote: I was wondering how much IDE channels linux 2.2 can handle, can it handle 8 channels ? I think the limit with the later 2.2 kernel ide patches is 10 IDE channels. I have run quite a bit with 4 Promise cards (8 channels), plus the 2 onboard PIIX channels. Jan Edler NEC Research Institute
Re: Ribbon Cabling (was Re: large ide raid system)
[ Tuesday, January 11, 2000 ] Andy Poling wrote: On Tue, 11 Jan 2000, Gregory Leblanc wrote: If you cut the cable lengthwise (no, don't cut the wires) between wires (don't break the insulation on the wires themselves, just the connecting plastic) you can get your cables to be 1/4 the normal width (up until you get to the connector). I don't know about IDE, but I'm pretty sure that's a big no-no for SCSI cables. The alternating conductors in the ribbon cable are sig, gnd, sig, gnd, sig, etc. And it's electrically important (for proper impedance and noise and cross-talk rejection) that they stay that way. I think the same is probably true for the schmancy UDMA66 cables too... So just check with a cable spec and make sure you're not separating a data signal from its ground return path. Throw some mag rings around the thing if you want, but since we're (hopefully) terminated properly (no reflection) the crosstalk issues aren't huge... they suffer more through the LC matrix of connector adaptors than this split would cause :) James -- Miscellaneous Engineer --- IBM Netfinity Performance Development
Re: large ide raid system
Getting back to the discussion of Hardware vs. Software raid... Can someone say *definitively* *where* the raid-5 code is being run on a *current* Raidzone product? Originally, it was an "md" process running on the system cpu. Currently I'm not so sure. The SmartCan *does* have its own BIOS, so there is *some* intelligence there, but what exactly is the division of responsibility here... From a recent email exchange with [EMAIL PROTECTED] of consensys, makers of raidzone: BG Does the raidzone product for linux use hardware or software raid? RZ It is in the firmware of the unit. RZ RZ Everything is our own raid. eg. it is not RAID tools in Linux. BG [please clarify] RZ Our raid is firmware. This means that its both hardware and software. RZ RZ Most people are interested in a RAID5 configuration. The parity is RZ calculated on the CPU of the mother board. Our raid is as fast as anyone RZ else's raid hardware or software. There is a great misnomer regarding RZ raid today. Basically in the old days of 100 MHz CPU's there was a RZ performance issuer with calculating the parity on the CPU. Today that is RZ not true and many of the PC magazines reflect this in their comments. RZ There are lots of left over cycles to calculate the parity. RZ RZ Of course this is not the only thing the affects speed. Other issues that RZ make our units fast is the PCI bus which is 133Mbs and DMA directly to RZ drives. It is however, still unclear whether it's safe to run reiserfs on a raidzone. I have a question about that out to Colin. Brian
Re: Ribbon Cabling (was Re: large ide raid system)
From [EMAIL PROTECTED] Tue Jan 11 21:44:29 2000 On Tue, 11 Jan 2000, Gregory Leblanc wrote: If you cut the cable lengthwise (no, don't cut the wires) between wires (don't break the insulation on the wires themselves, just the connecting plastic) you can get your cables to be 1/4 the normal width (up until you get to the connector). I don't know about IDE, but I'm pretty sure that's a big no-no for SCSI cables. The alternating conductors in the ribbon cable are sig, gnd, sig, gnd, sig, etc. And it's electrically important (for proper impedance and noise and cross-talk rejection) that they stay that way. I think the same is probably true for the schmancy UDMA66 cables too... vent Back in the day 8-) high end SCSI ribbon cables consisted of twisted pairs between the connectors so it was really easy to deform the cable to fit through tight spots. Now, all I seem to find is the cheap ribbon cable that's excreted from nameless companies in developing countries where their ideas of quality control differ vastly from mine. 8-) Either I'm really unlucky or the quality of ribbon cabling in general is in decline...sigh. /vent And I agree with the idea that slicing up the ribbon cable is probably not going to work. Cheers, Chris -- Christopher Mauritz [EMAIL PROTECTED]
Re: Ribbon Cabling (was Re: large ide raid system)
-BEGIN PGP SIGNED MESSAGE- On 11-Jan-2000 James Manning wrote: [ Tuesday, January 11, 2000 ] Andy Poling wrote: On Tue, 11 Jan 2000, Gregory Leblanc wrote: If you cut the cable lengthwise (no, don't cut the wires) between wires (don't break the insulation on the wires themselves, just the connecting plastic) you can get your cables to be 1/4 the normal width (up until you get to the connector). I don't know about IDE, but I'm pretty sure that's a big no-no for SCSI cables. The alternating conductors in the ribbon cable are sig, gnd, sig, gnd, sig, etc. And it's electrically important (for proper impedance and noise and cross-talk rejection) that they stay that way. I think the same is probably true for the schmancy UDMA66 cables too... So just check with a cable spec and make sure you're not separating a data signal from its ground return path. Throw some mag rings around the thing if you want, but since we're (hopefully) terminated properly (no reflection) the crosstalk issues aren't huge... they suffer more through the LC matrix of connector adaptors than this split would cause :) James -- Miscellaneous Engineer --- IBM Netfinity Performance Development Have a look at old 3M cables (used in most old suns and all old decstations). They have all the wires separated. And they work at least up to SCSI2. I also thought that the sig/gnd/sig/gnd was mandatory but these cables prove that there is another way to do it at least in some cases. My $0.02 - -- Anton R. Ivanov IP Engineer Level3 Communications RIPE: ARI2-RIPE E-Mail: Anton Ivanov [EMAIL PROTECTED] @*** Segal's Law *** A man with one watch knows what time it is; a man with two watches is never sure. - -- -BEGIN PGP SIGNATURE- Version: GnuPG v1.0.0 (GNU/Linux) Comment: For info see http://www.gnupg.org iQEVAwUBOHxGyylWAw/bM84zAQE5Swf+PUSlcf0gX+l1gUZTn/fsSN1Q+cO+kA6M Z5v9X/83mD0KOV8IAo5YRY9+E7BAIBaD5+rXgyFSWYdeIvewI0C9mTjSlliwv1ZN 1goBHvL5tqsIz21v/cbx/veW+zoSssHrj/ufm9GI2dXAIzdIA2YQ3BzZ60w6YLdH Pben/18W/KXNKuEqyEkBpRKJyXiLx6NBt2iM9qlMCfJHAd8KWv2ruqDv9v55aJOX e1HCKNxFHHuO951JjV4zzb+rlhD6lqGsw2EtN77228qGs1uKUkktAviTmRduHtzJ kTAr9YOu1T3/apUIFOjmZHtrjWqZWjVZ8lqT/iKF2HvcHsFCwqJJPg== =84Gh -END PGP SIGNATURE-
Re: Ribbon Cabling (was Re: large ide raid system)
On Tue, 11 Jan 2000, James Manning wrote: If you cut the cable lengthwise (no, don't cut the wires) between wires (etc.) I don't know about IDE, but I'm pretty sure that's a big no-no for SCSI cables. The alternating conductors in the ribbon cable are sig, gnd, sig, gnd, sig, etc. And it's electrically important (for proper impedance and noise and cross-talk rejection) that they stay that way. I think the same is probably true for the schmancy UDMA66 cables too... So just check with a cable spec and make sure you're not separating a data signal from its ground return path. Throw some mag rings around the thing if you want, but since we're (hopefully) terminated properly (no reflection) the crosstalk issues aren't huge... they suffer more through the LC matrix of connector adaptors than this split would cause :) ,,Termination`` means nothing else then a resistance at the end of the cable (each pair) that is equivalent to the cable impedance. And the impedance depends on the cable geometry (and material, of course). So IMHO you can't divide the flat cable to the pairs (unless they're twisted pairs) or even single wires without an impedance change of the cable section involved. BoChal.
RE: Ribbon Cabling (was Re: large ide raid system)
Title: RE: Ribbon Cabling (was Re: large ide raid system) You may be thinking of differential SCSI which uses a balanced (and twisted) pair for each data and signal line. In the old days, there was only one flavor of differential, and it was popular at least on Hewlett-Packard 800 series systems (which uses round SCSI cables). Now, in addition to the old differential, there is something called low voltage differential (LVD) also known in marketing hype as Ultra2. LVD cables look like a ribbon cable but have the signal pairs twisted. The last ones I bought from Adaptec were high quality, but expensive! Best I remember, they were almost $100 for a 4 device cable. But then again, they have active terminators built on the end of the cable (LVD drives don't have built-in terminators). -Original Message- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]] Sent: Tuesday, January 11, 2000 10:11 PM To: [EMAIL PROTECTED] Cc: [EMAIL PROTECTED]; [EMAIL PROTECTED] Subject: Re: Ribbon Cabling (was Re: large ide raid system) From [EMAIL PROTECTED] Tue Jan 11 21:44:29 2000 On Tue, 11 Jan 2000, Gregory Leblanc wrote: If you cut the cable lengthwise (no, don't cut the wires) between wires (don't break the insulation on the wires themselves, just the connecting plastic) you can get your cables to be 1/4 the normal width (up until you get to the connector). I don't know about IDE, but I'm pretty sure that's a big no-no for SCSI cables. The alternating conductors in the ribbon cable are sig, gnd, sig, gnd, sig, etc. And it's electrically important (for proper impedance and noise and cross-talk rejection) that they stay that way. I think the same is probably true for the schmancy UDMA66 cables too... vent Back in the day 8-) high end SCSI ribbon cables consisted of twisted pairs between the connectors so it was really easy to deform the cable to fit through tight spots. Now, all I seem to find is the cheap ribbon cable that's excreted from nameless companies in developing countries where their ideas of quality control differ vastly from mine. 8-) Either I'm really unlucky or the quality of ribbon cabling in general is in decline...sigh. /vent And I agree with the idea that slicing up the ribbon cable is probably not going to work. Cheers, Chris -- Christopher Mauritz [EMAIL PROTECTED]
Re: large ide raid system
Jan Edler wrote: It all depends on your minimum acceptable performance level. I know my master/slave test setup couldn't keep up with fast ethernet (10 MByte/s). I don't remember if it was 1 Mbyte/s or not. Fastethernet is 12mb/sec, Ethernet is 1.2mb/sec. My 4way IDE based, 2 channels (ie, master/slave, master/slave) built using IBM 16gb Ultra33 drives in RAID0 are capable of about 25mb/sec across the raid. Adding a Promise 66 card, changing to all masters, got the numbers up into the 30's range (I don't have them at the moment.. hmm..) I was also wondering about the reliability of using slaves. Does anyone know about the likelihood of a single failed drive bringing down the whole master/slave pair? Since I have tended to stay away from slaves, for performance reasons, I don't know how they influence reliability. Maybe it's ok. When the slave fail, the master goes down. My experience has been, when _ANY_ IDE drive fails, it takes down the whole channel. Master or slave. The kernel just gives fits.. -- +-- Thomas Davis| PDSF Project Leader [EMAIL PROTECTED] | (510) 486-4524 | "Only a petabyte of data this year?"
Re: Ribbon Cabling (was Re: large ide raid system)
$horse='dead'; beat($horse); [ Wednesday, January 12, 2000 ] Bohumil Chalupa wrote: ,,Termination`` means nothing else then a resistance at the end of the cable (each pair) that is equivalent to the cable impedance. And the impedance depends on the cable geometry (and material, of course). So IMHO you can't divide the flat cable to the pairs (unless they're twisted pairs) or even single wires without an impedance change of the cable section involved. Impedance depends on the cable the signal is going through and the distance to its return path (which you agree to above). The distributed RLC model of transmission lines (longer than 3 inches, so I'm not trusting lumped :) is based on properties of the cable itself, though. Now, fast signals (GHz signals skimming only along the top of a microstrip, for instance) need a very-close signal return path (PCB traces that need to be closer to the power plane under/over them, as impedance rises with distance from return path) hence the need to keep the pairs together, but with the pairs kept together each cable is equi-distant to its return path both before and after the splitting, so impedance is not affected. If someone splits out the individual wires, you're absolutely right, but that's not what we're advocating :) Anyway, there's enough cables that come off the manufacturing line not meeting spec (and still get used) that even if messing with the ribbon had integrity issues, that doesn't the cable wouldn't still work :) James -- Miscellaneous Engineer --- IBM Netfinity Performance Development
Re: large ide raid system
James Manning wrote: [ Tuesday, January 11, 2000 ] Thomas Davis wrote: ---Sequential Output ---Sequential Input-- --Random-- -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks--- MachineMB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU pdsfdv10 1024 14076 85.1 18487 24.3 12089 35.8 20182 83.0 63064 69.8 344.4 7.1 hmmm ok... Any chance I could talk you into running the tiobench.pl from http://www.iki.fi/miku/tiotest/ (after "make" to build tiotest)? I'd love to see what it puts out vs. bonnie on such a system. [root@pdsfdv10 tiotest-0.16]# ./tiobench.pl Found memory size of 255.94140625 MB Now running ./tiotest -t 1 -f 510 -s 4000 -b 4096 -d . -T -W Size is MB, BlkSz is Bytes, Read and Write are MB/sec, Seeks are Seeks/sec MachineDirectory Size(MB) BlkSz Threads Read Write Seeks --- --- - --- - -- --- --- . 510 4096 1 63.197 16.073 185.185 Now running ./tiotest -t 2 -f 255 -s 2000 -b 4096 -d . -T -W . 510 4096 2 62.347 15.366 304.183 Now running ./tiotest -t 4 -f 127 -s 1000 -b 4096 -d . -T -W . 510 4096 4 67.285 15.070 528.402 Now running ./tiotest -t 8 -f 63 -s 500 -b 4096 -d . -T -W . 510 4096 8 52.610 14.797 803.213 [root@pdsfdv10 tiotest-0.16]# ./tiobench.pl --threads 16 Found memory size of 255.94140625 MB Now running ./tiotest -t 16 -f 31 -s 250 -b 4096 -d . -T -W Size is MB, BlkSz is Bytes, Read and Write are MB/sec, Seeks are Seeks/sec MachineDirectory Size(MB) BlkSz Threads Read Write Seeks --- --- - --- - -- --- --- . 510 4096 16 33.514 14.806 1302.93 [root@pdsfdv10 tiotest-0.16]# ./tiobench.pl --threads 32 Found memory size of 255.94140625 MB Now running ./tiotest -t 32 -f 15 -s 125 -b 4096 -d . -T -W Size is MB, BlkSz is Bytes, Read and Write are MB/sec, Seeks are Seeks/sec MachineDirectory Size(MB) BlkSz Threads Read Write Seeks --- --- - --- - -- --- --- . 510 4096 32 27.491 13.445 1851.85 [root@pdsfdv10 tiotest-0.16]# ./tiobench.pl --threads 64 Found memory size of 255.94140625 MB Now running ./tiotest -t 64 -f 7 -s 62 -b 4096 -d . -T -W Size is MB, BlkSz is Bytes, Read and Write are MB/sec, Seeks are Seeks/sec MachineDirectory Size(MB) BlkSz Threads Read Write Seeks --- --- - --- - -- --- --- . 510 4096 64 42.667 13.211 2110.63 [root@pdsfdv10 tiotest-0.16]# ./tiobench.pl --threads 128 Found memory size of 255.94140625 MB Now running ./tiotest -t 128 -f 3 -s 31 -b 4096 -d . -T -W Size is MB, BlkSz is Bytes, Read and Write are MB/sec, Seeks are Seeks/sec MachineDirectory Size(MB) BlkSz Threads Read Write Seeks --- --- - --- - -- --- --- . 510 4096 128 44.548 13.627 2511.39 [root@pdsfdv10 tiotest-0.16]# ./tiobench.pl --threads 256 Found memory size of 255.94140625 MB Now running ./tiotest -t 256 -f 1 -s 15 -b 4096 -d . -T -W Size is MB, BlkSz is Bytes, Read and Write are MB/sec, Seeks are Seeks/sec MachineDirectory Size(MB) BlkSz Threads Read Write Seeks --- --- - --- - -- --- --- . 510 4096 256 36.941 12.580 4042.10 (I was playing around at the end..) -- +-- Thomas Davis| PDSF Project Leader [EMAIL PROTECTED] | (510) 486-4524 | "Only a petabyte of data this year?"
Re: large ide raid system
Dan Hollis wrote: On Mon, 10 Jan 2000, Jan Edler wrote: Cable length is not so much a pain as the number of cables. Of course with scsi you want multiple channels anyway for performance, so the situation is very similar to ide. A cable mess. There's a (relatively) nice way to get around this, if you make your own IDE cables (or are brave enough to cut some up). If you cut the cable lengthwise (no, don't cut the wires) between wires (don't break the insulation on the wires themselves, just the connecting plastic) you can get your cables to be 1/4 the normal width (up until you get to the connector). This also makes a big difference for airflow, since those big, flat ribbon cables are really bad for that. Greg
Re: large ide raid system
Jan Edler wrote: On Mon, Jan 10, 2000 at 12:49:29PM -0800, Dan Hollis wrote: On Mon, 10 Jan 2000, Jan Edler wrote: - Performance is really horrible if you use IDE slaves. Even though you say you aren't performance-sensitive, I'd recommend against it if possible. My tests indicate UDMA performs favorably with ultrascsi, at about 1/6 the cost. Cost is often a big factor. I wasn't advising against IDE, only against the use of slaves. With UDMA-33 or -66, masters work quite well, if you can deal with the other constraints that I mentioned (cable length, PCI slots, etc). Do you have any numbers handy ? will the performance of master/slave setup be at least HALF of the master-only setup. For some apps cost is really important, and software IDE RAID has a very low price/Megabyte. If the app doesn't need killer performance , then I think it is the best solution. now if we only had soft-RAID + journaled FS + power failure safeness right now ... cheers, Benno.
Re: large ide raid system
Thomas Davis wrote: James Manning wrote: Well, it's kind of on-topic thanks to this post... Has anyone used the systems/racks/appliances/etc from raidzone.com? If you believe their site, it certainly looks like a good possibility. Yes. It's pricey. Not much cheaper that SCSI chassis. You only save money on the drives. Interesting... The 100GB Internal RAID-5 SmartCan I purchased from RaidZone was approx. $5k. The quotes I got for a SCSI equivalent ranged from $10k to $15K. Personally I consider half the cost significantly cheaper. I also was quite impressed with a qoute for a 1TB rackmount system in the $50K range, again SCSI equivalents were significantly higher... Performance is ok. Has a few other problems - your stuck with the kernels they support; the raid code is NOT open sourced. Performance is pretty good - these numbers are for a first generation smartcan (spring '99) ---Sequential Output ---Sequential Input-- --Random-- -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks--- MachineMB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU raidzone 100 6923 89.7 25987 26.6 14230 28.9 7297 89.4 215121 77.7 16407.3 69.7 raidzone 200 6537 86.2 22175 21.5 14297 30.2 7667 92.5 56355 36.0 377.5 3.1 Softraid 100 6598 86.0 43411 36.5 12077 27.4 6180 77.9 54022 46.4 721.4 4.1 Softraid 200 8337 87.9 25373 24.0 9009 18.8 8952 87.1 34413 21.7 301.1 2.2 The two sets of numbers were measured on the same computer hardware setup (500Mhz PIII w/ 128MB, 100GB Smartcan w/ 5 24GB IBM drives). "raidzone" is using Raidzone's most recent pre-release version of their Linux software (BIOS upgrades all). "Softraid" was based on early alpha release of RaidZone's linux support which basically allowed you to access the individual drives. RAID was handled by the Software Raid support available under RedHat Linux 6.0 6.1. Both were set up as RAID-5 Using "top": - With "Softraid" bonnie and the md Raid-5 software were sharing the cpu equally - With "raidzone" bonnie was consuming most (85%) of the cpu and no other processes and "system" 15% Getting back to the discussion of Hardware vs. Software raid... Can someone say *definitively* *where* the raid-5 code is being run on a *current* Raidzone product? Originally, it was an "md" process running on the system cpu. Currently I'm not so sure. The SmartCan *does* have its own BIOS, so there is *some* intelligence there, but what exactly is the division of responsibility here... John -- John Burton, Ph.D. Senior Associate GATS, Inc. [EMAIL PROTECTED] 11864 Canon Blvd - Suite 101 [EMAIL PROTECTED] (personal) Newport News, VA 23606 (757) 873-5920 (voice) (757) 873-5924 (fax)
Re: large ide raid system
SCSI works quite well with many devices connected to the same cable. The PCI bus turns out to be the bottleneck with the faster scsi modes, so it doesn't matter how many channels you have. If performance was the issue, but the original poster wasn't interested in performance, multiple channels would improve performance if the slower (single ended) devices are used. Lance Dan Hollis wrote: Cable length is not so much a pain as the number of cables. Of course with scsi you want multiple channels anyway for performance, so the situation is very similar to ide. A cable mess.
Ribbon Cabling (was Re: large ide raid system)
On Tue, 11 Jan 2000, Gregory Leblanc wrote: If you cut the cable lengthwise (no, don't cut the wires) between wires (don't break the insulation on the wires themselves, just the connecting plastic) you can get your cables to be 1/4 the normal width (up until you get to the connector). I don't know about IDE, but I'm pretty sure that's a big no-no for SCSI cables. The alternating conductors in the ribbon cable are sig, gnd, sig, gnd, sig, etc. And it's electrically important (for proper impedance and noise and cross-talk rejection) that they stay that way. I think the same is probably true for the schmancy UDMA66 cables too... -Andy
Re: large ide raid system
John Burton wrote: Thomas Davis wrote: James Manning wrote: Well, it's kind of on-topic thanks to this post... Has anyone used the systems/racks/appliances/etc from raidzone.com? If you believe their site, it certainly looks like a good possibility. Yes. It's pricey. Not much cheaper that SCSI chassis. You only save money on the drives. Interesting... The 100GB Internal RAID-5 SmartCan I purchased from RaidZone was approx. $5k. The quotes I got for a SCSI equivalent ranged from $10k to $15K. Personally I consider half the cost significantly cheaper. I also was quite impressed with a qoute for a 1TB rackmount system in the $50K range, again SCSI equivalents were significantly higher... We paid $25k x 4, for: 2x450mhz cpu 256mb ram 15x37gb IBM 5400 drives (550 gb of drive space) Intel system board, w/eepro tulip card (channel bonded into cisco5500) Performance is pretty good - these numbers are for a first generation smartcan (spring '99) ---Sequential Output ---Sequential Input-- --Random-- -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks--- MachineMB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU raidzone 100 6923 89.7 25987 26.6 14230 28.9 7297 89.4 215121 77.7 16407.3 69.7 raidzone 200 6537 86.2 22175 21.5 14297 30.2 7667 92.5 56355 36.0 377.5 3.1 Softraid 100 6598 86.0 43411 36.5 12077 27.4 6180 77.9 54022 46.4 721.4 4.1 Softraid 200 8337 87.9 25373 24.0 9009 18.8 8952 87.1 34413 21.7 301.1 2.2 You made a mistake. :-) Your bonnie size is smaller than the amount of memory in the machine your tested on - so you tested the memory, NOT the drive system. Our current large machine(s) (15x37gb IBM drives, 500gb file system, 4kb blocks, v2.2.13 kernel, fixed knfsd, channel bonding, raidzone 1.2.0b3) does: ---Sequential Output ---Sequential Input-- --Random-- -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks--- MachineMB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU pdsfdv10 1024 14076 85.1 18487 24.3 12089 35.8 20182 83.0 63064 69.8 344.4 7.1 I've also hit it with 8 machines, doing an NFS copy of about 60gb onto it, and it sustained about a 20mb/sec write rate. Using "top": - With "Softraid" bonnie and the md Raid-5 software were sharing the cpu equally - With "raidzone" bonnie was consuming most (85%) of the cpu and no other processes and "system" 15% I've seen load averages in the 5's and 6's. This is on a dual processor machine w/256mb of ram. My biggest complaint is the raid rebuild code runs as the highest priority, so on a crash/reboot, it takes _forever_ for fsck to complete (because the rebuild thread is taking all of the CPU and disk bandwidth). The raidzone code also appears to be single threaded - it doesn't take advantage of multiple CPU's. (although, user space code benefits from having a second CPU then) Getting back to the discussion of Hardware vs. Software raid... Can someone say *definitively* *where* the raid-5 code is being run on a *current* Raidzone product? Originally, it was an "md" process running on the system cpu. Currently I'm not so sure. The SmartCan *does* have its own BIOS, so there is *some* intelligence there, but what exactly is the division of responsibility here... None of the RAID code runs in the smartcan, or the controller. It all runs in the kernel. the current code has several kernel threads, and a user space thread: root 6 0.0 0.0 00 ?SW Jan04 0:02 [rzft-syncd] root 7 0.0 0.0 00 ?SW Jan04 0:00 [rzft-rcvryd] root 8 0.1 0.0 00 ?SW Jan04 14:41 [rzft-dpcd] root 620 0.0 0.0 5640 ?SW Jan04 0:00 [rzmpd] root 621 0.0 0.1 2080 296 ?SJan04 3:30 rzmpd root 3372 0.0 0.0 00 ?ZJan10 0:00 [rzmpd defunct] root 3806 0.0 0.1 1240 492 pts/1S09:57 0:00 grep rz -- +-- Thomas Davis| PDSF Project Leader [EMAIL PROTECTED] | (510) 486-4524 | "Only a petabyte of data this year?"
Re: large ide raid system
Benno Senoner wrote: Jan Edler wrote: On Mon, Jan 10, 2000 at 12:49:29PM -0800, Dan Hollis wrote: On Mon, 10 Jan 2000, Jan Edler wrote: - Performance is really horrible if you use IDE slaves. Even though you say you aren't performance-sensitive, I'd recommend against it if possible. My tests indicate UDMA performs favorably with ultrascsi, at about 1/6 the cost. Cost is often a big factor. I wasn't advising against IDE, only against the use of slaves. With UDMA-33 or -66, masters work quite well, if you can deal with the other constraints that I mentioned (cable length, PCI slots, etc). Do you have any numbers handy ? will the performance of master/slave setup be at least HALF of the master-only setup. Well, this depends on how it's used. If you were saturating your I/O bus, then things would be REALLY ugly. Say you've got a controller running in UDMA/33 mode, with two disks attached. If you have drives that are reasonably fast, say recent 5400 RPM UDMA drives, then this will actually hinder performance compared to having just one drive. If you're doing 16MB/sec of I/O, then your performance will be slightly less than half the performance of having just one drive on that channel (consider overhead, IDE controller context switches, etc). If you only need the space, then this is an accptable solution, for low throughput applications. I don't know jack schitt about ext2, the linux ide drivers (patches or old ones), or about the RAID code, except that they work. For some apps cost is really important, and software IDE RAID has a very low price/Megabyte. If the app doesn't need killer performance , then I think it is the best solution. It's a very good solution for a small number of disks, where you can keep everything in a small case. It may actually be superior to SCSI for situations where you have 4 or fewer disks and can put just a single disk on a controller. now if we only had soft-RAID + journaled FS + power failure safeness right now ... I'll be happy as long as it gets there relatively soon, I'll be happy. fsck'ing is the only thing that really bugs me... Greg
Re: large ide raid system
On Tue, Jan 11, 2000 at 04:25:27PM +0100, Benno Senoner wrote: Jan Edler wrote: I wasn't advising against IDE, only against the use of slaves. With UDMA-33 or -66, masters work quite well, if you can deal with the other constraints that I mentioned (cable length, PCI slots, etc). Do you have any numbers handy ? Sorry, I can't seem to find any quantitative results on that right now. will the performance of master/slave setup be at least HALF of the master-only setup. I did run some tests, and my recollection is that it was much worse. For some apps cost is really important, and software IDE RAID has a very low price/Megabyte. If the app doesn't need killer performance , then I think it is the best solution. It all depends on your minimum acceptable performance level. I know my master/slave test setup couldn't keep up with fast ethernet (10 MByte/s). I don't remember if it was 1 Mbyte/s or not. I was also wondering about the reliability of using slaves. Does anyone know about the likelihood of a single failed drive bringing down the whole master/slave pair? Since I have tended to stay away from slaves, for performance reasons, I don't know how they influence reliability. Maybe it's ok. Jan Edler NEC Research Institute
Re: large ide raid system
[ Tuesday, January 11, 2000 ] John Burton wrote: Performance is pretty good - these numbers are for a first generation smartcan (spring '99) Could you re-run the raidzone and softraid with a size of 512MB or larger? Could you run the tiobench.pl from http://www.iki.fi/miku/tiotest (after "make" to build tiotest) Those would be great results to see. Thanks, James -- Miscellaneous Engineer --- IBM Netfinity Performance Development
Re: large ide raid system
[ Sunday, January 9, 2000 ] Franc Carter wrote: I am planning to set up a large ide raid5 system. From reading the archives of the list it looks like the way to go is with promise ultra66 cards, making sure that I have good cables. I am hopeing to get a minimum of 8 drives into a machine. My current plan is for the following config:- 37gig IBM ide drives 2.2.14 kernel (or may be a 2.3 series) software raid5 Promise Ultra66 cards Good quality cabling extra fans Any comments or suggestions ? I don't care about performance (it's only competing against tape drives), however I do care about dollars per gigabyte and reliablity. Well, it's kind of on-topic thanks to this post... Has anyone used the systems/racks/appliances/etc from raidzone.com? If you believe their site, it certainly looks like a good possibility. James -- Miscellaneous Engineer --- IBM Netfinity Performance Development
Re: large ide raid system
Franc Carter wrote: I am planning to set up a large ide raid5 system. From reading the archives of the list it looks like the way to go is with promise ultra66 cards, making sure that I have good cables. I am hopeing to get a minimum of 8 drives into a machine. My current plan is for the following config:- 37gig IBM ide drives 2.2.14 kernel (or may be a 2.3 series) software raid5 Promise Ultra66 cards Good quality cabling extra fans Any comments or suggestions ? I don't care about performance (it's only competing against tape drives), however I do care about dollars per gigabyte and reliablity. Personally I'd recomend getting 8-channels for your drives. (1 per channel) This will make things a bit more expensive (not too much though), and will make your whole system a lot happier. Of course, if you really don't care about performance, then you should be getting cheaper IDE controllers, and cheaper drives. The two that you've got picked out are high end for IDE components. What kind of a case are you looking to house this in? Greg
Re: large ide raid system
From my experience, it works fairly well, but there are some constraints: - Performance is really horrible if you use IDE slaves. Even though you say you aren't performance-sensitive, I'd recommend against it if possible. - Thus, to get 8 drives in a machine, you not only need mounting, power, and cooling for 8 drives, but also 4 available PCI slots for the Promise cards (or maybe 3 if you can make use of onboard ATA channels). - Cable length can be a problem. I've had good luck with the 24 inch cables, although they exceed the length specified in the spec. Even so, it can be tough to route the cables from the promise cards to the drives. I think it would be completely hopeless for 8 drives with 18 inch cables. - It may be worth getting hot-swap drive boxes, although it will add significantly to your per-drive cost. Be careful to get ones that support udma-66 (or at least udma-33). This allows you to recover rather more quickly from a drive failure, assuming you buy at least 1 extra hot-swap box and drive. Even if you don't mind rebooting to deal with a failure, it sure beats tearing open the machine. Good luck, Jan Edler NEC Research Institute On Mon, Jan 10, 2000 at 03:26:26PM +1100, Franc Carter wrote: I am planning to set up a large ide raid5 system. From reading the archives of the list it looks like the way to go is with promise ultra66 cards, making sure that I have good cables. I am hopeing to get a minimum of 8 drives into a machine. My current plan is for the following config:- 37gig IBM ide drives 2.2.14 kernel (or may be a 2.3 series) software raid5 Promise Ultra66 cards Good quality cabling extra fans Any comments or suggestions ? I don't care about performance (it's only competing against tape drives), however I do care about dollars per gigabyte and reliablity. thanks -- Franc CarterMEMLab, University of Sydney Ph: 61-2-9351-7819 Fax: 9351-6461
Re: large ide raid system
On Mon, Jan 10, 2000 at 12:49:29PM -0800, Dan Hollis wrote: On Mon, 10 Jan 2000, Jan Edler wrote: - Performance is really horrible if you use IDE slaves. Even though you say you aren't performance-sensitive, I'd recommend against it if possible. My tests indicate UDMA performs favorably with ultrascsi, at about 1/6 the cost. Cost is often a big factor. I wasn't advising against IDE, only against the use of slaves. With UDMA-33 or -66, masters work quite well, if you can deal with the other constraints that I mentioned (cable length, PCI slots, etc). Jan Edler NEC Research Institute
Re: large ide raid system
On Mon, 10 Jan 2000, Jan Edler wrote: - Performance is really horrible if you use IDE slaves. Even though you say you aren't performance-sensitive, I'd recommend against it if possible. My tests indicate UDMA performs favorably with ultrascsi, at about 1/6 the cost. Cost is often a big factor. -Dan
Re: large ide raid system
On Mon, Jan 10, 2000 at 02:03:14AM -0500, James Manning wrote: Has anyone used the systems/racks/appliances/etc from raidzone.com? If you believe their site, it certainly looks like a good possibility. The raidzone stuff works, and the packaging is nice. They provide much more scalability than a roll-your-own ATA-based solution, so you can have many more than 8 drives. In terms of software, their raid layer lives within the driver, which has some advantages (and some disadvantages) also. The disadvantages are mainly that it's much more expensive than a roll-your-own approach and they are taking a fairly closed attitude towards the software. The driver is distributed in binary form. You apply a bunch of kernel patches, and link in their driver. This causes all sorts of problems. Jan Edler NEC Research Institute
Re: large ide raid system
On Mon, 10 Jan 2000, Jan Edler wrote: My tests indicate UDMA performs favorably with ultrascsi, at about 1/6 the cost. Cost is often a big factor. I wasn't advising against IDE, only against the use of slaves. Here we agree :D 1 device per channel. (When will any vendors implement IDE disconnect? the spec has existed for ages.) With UDMA-33 or -66, masters work quite well, if you can deal with the other constraints that I mentioned (cable length, PCI slots, etc). Get an Abit BP6 and you have 4 onboard udma channels, and 6 PCI slots. :D Cable length is not so much a pain as the number of cables. Of course with scsi you want multiple channels anyway for performance, so the situation is very similar to ide. A cable mess. -Dan
Re: large ide raid system
James Manning wrote: Well, it's kind of on-topic thanks to this post... Has anyone used the systems/racks/appliances/etc from raidzone.com? If you believe their site, it certainly looks like a good possibility. Yes. It's pricey. Not much cheaper that SCSI chassis. You only save money on the drives. Performance is ok. Has a few other problems - your stuck with the kernels they support; the raid code is NOT open sourced. -- +-- Thomas Davis| PDSF Project Leader [EMAIL PROTECTED] | (510) 486-4524 | "Only a petabyte of data this year?"
large ide raid system
I am planning to set up a large ide raid5 system. From reading the archives of the list it looks like the way to go is with promise ultra66 cards, making sure that I have good cables. I am hopeing to get a minimum of 8 drives into a machine. My current plan is for the following config:- 37gig IBM ide drives 2.2.14 kernel (or may be a 2.3 series) software raid5 Promise Ultra66 cards Good quality cabling extra fans Any comments or suggestions ? I don't care about performance (it's only competing against tape drives), however I do care about dollars per gigabyte and reliablity. thanks -- Franc CarterMEMLab, University of Sydney Ph: 61-2-9351-7819 Fax: 9351-6461