the controller is 3132 based and small (in center of photo red controller). You can get them with internal or external sata ports. The good news is it works with OpenSolaris out of the box. Just plugging the card in and doing a 'devfsadm' we can do some testing before buying another set to fill the second pci-e slot. I wanted to see if adding a terabyte to the box for 250.00 is going to be reasonable and not cost me an arm and a leg to power and still keep up the performance. First test just use bonnie++ on a single drive connected to the controller with a ufs filesystem. -------------------------------------------------------------------------------- Version 1.03 ------Sequential Output------ --Sequential Input- --Random- -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks-- Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP spongebob 8G 50253 42 59354 9 21242 5 81241 72 104454 11 146.6 0 ------Sequential Create------ --------Random Create-------- -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete-- files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP 16 3992 13 +++++ +++ 23363 47 +++++ +++ +++++ +++ 2323 6 spongebob,8G,50253,42,59354,9,21242,5,81241,72,104454,11,146.6,0,16,3992,13,+++++,+++,23363,47,+++++,+++,+++++,+++,2323,6 script done on Sun Dec 02 23:08:21 2007 ------------------------------------------------------------------------------- Not bad but I'm not interested in ufs as the final result is using zfs. So lets combine both Maxtor 500Gig drives in a zfs pool and test. ------------------------------------------------------------------------------- Version 1.03 ------Sequential Output------ --Sequential Input- --Random- -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks-- Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP spongebob 8G 127012 97 121031 17 62793 9 100226 82 140708 9 388.8 0 ------Sequential Create------ --------Random Create-------- -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete-- files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP 16 +++++ +++ +++++ +++ +++++ +++ 27977 99 +++++ +++ +++++ +++ spongebob,8G,127012,97,121031,17,62793,9,100226,82,140708,9,388.8,0,16,+++++,+++,+++++,+++,+++++,+++,27977,99,+++++,+++,+++++,+++ script done on Sun Dec 02 23:23:54 2007 ------------------------------------------------------------------------------ Ok much better average 125MB on the writes But how much is this going to cost me to run and is it going to affect the performance of OpenSolaris running on this box. Just a cheap test I'll run two bonnie++ 8 gig tests on on the raidz (tank pool) and the new (tank2 pool) simultaneous. Note this is not the way you want to do a load test. You want to use something more like filebench which is included with opensolaris with a read/write workload that matches your environment. At this point I've got nothing to lose by checking with simultanious bonnie++ runs. Here is the result of both bonnie++ running and the average load was about 1.2 during the run. ------------------------------------------------------------------------------ tank pool zfs raidz (3 500G drives) Version 1.03 ------Sequential Output------ --Sequential Input- --Random- -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks-- Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP spongebob 8G 78443 63 119858 21 57535 10 92451 80 125629 9 415.9 0 ------Sequential Create------ --------Random Create-------- -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete-- files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP 16 +++++ +++ +++++ +++ +++++ +++ +++++ +++ +++++ +++ +++++ +++ spongebob,8G,78443,63,119858,21,57535,10,92451,80,125629,9,415.9,0,16,+++++,+++,+++++,+++,+++++,+++,+++++,+++,+++++,+++,+++++,+++ palowoda@spongebob:/home/palowoda> exit --------------- tank2 pool zfs raid 0 Version 1.03 ------Sequential Output------ --Sequential Input- --Random- -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks-- Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP spongebob 8G 83453 65 103065 17 65402 11 90547 78 134291 10 354.1 0 ------Sequential Create------ --------Random Create-------- -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete-- files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP 16 +++++ +++ +++++ +++ +++++ +++ +++++ +++ +++++ +++ +++++ +++ spongebob,8G,83453,65,103065,17,65402,11,90547,78,134291,10,354.1,0,16,+++++,+++,+++++,+++,+++++,+++,+++++,+++,+++++,+++,+++++,+++ palowoda@spongebob:/tank2/tmp> exit ------------------------------------------------------------------------------- So yes I did lose some performance on both pools. Not a big lose though. What is this going to cost me in power and time with simple real world results? I recently bought a Kill-O-Watt meter to measure all my systems power consumption. If your a home enthusiast please buy one of these meters and and measure everything in the house and claim to your SO that you are now God and know everything that deals with the power bill. Well claim everything but paying for the power. With four 500G drives in the system power consumption was about 174 watts mostly idle. Adding the extra controller and two 500G drives consumes about 194 watts. But I want to test with with some load so I will use bonnie++ on the tank2 pool in an infinite loop while compiling the GCC compiler suite 4.2.2 on the other zfs pool which would normally take me 23min I'll run all of this for 1hr and use the system also to browse the web and do video's etc while it's loaded. The results. Average wattage 225 1hr usage .21kwh 1day usage 5.04kwh 1month usage 151.2kwh .20 kwh $/month $30.24 The system user interactive performance was usable with the load still below 2. I did not perceive any interactive video lag while using the the Firefox webrowser or using the system when it was loaded. My compile times of gcc went from ~23min to about 28min under this load. The next objective I guess is to add the next one terabyte of storage and repeat the tests and taking a look at how much it's going to cost me. So far some of the things I would do different. I would have purchased a case without the power supply and purchased a highly efficient maybe 500 to 600 watt power supply separate. Maybe buy a lager case but I still can get all drives into the case. I will have to add another fan on the front side of the case to cool all the drives in the end. I'm guessing that it's going to cost me 35.00 a month to run the final workstation and that is 420.00 a year less running the monitor. It's used as a server and the 15 inch lcd monitor is only running ~12watts so I'm not too concerned. One more SATA controller and two drives card and hit the 1500.00 dollar mark for total system cost the on next round. If you think of it you could have a 8 terabyte box for about 2500.00. *Update* I did record a 24hr period of power usage and it turns out to be 4.53kwh. With normal usage throughout the day. This is actually lower than my worse case but was expected. ---Bob