IRC Logs for #crux-devel Wednesday, 2012-11-14

frinnstRomster: i have two of those01:03
frinnstthey are ok01:03
Romsterrmull, add || true to the end of command01:24
Romsterjaeger, the only thing i have is keeping away from OCZ after hearing the warranty issues with them.01:31
Romsterhttp://www.scorptec.com.au/computer/45423-ssdsc2ct120a3k5 intel 33001:33
Romsterhttp://www.pccasegear.com/index.php?main_page=product_info&cPath=210_902_909&products_id=19790 sandisk http://hexus.net/tech/reviews/storage/36529-sandisk-extreme-ssd-120gb/01:34
Romsterand that samsung  830 Series http://www.scorptec.com.au/computer/45507-mz-7pc128b/ww which seems to have a xonsiderbly lower write speed. 320mb/s01:35
Romsteri prefer to pay a little more if need be for a solid drive that wont degrade over time01:36
Romsterhmm I also forgot Seagate owns Samsung now too.01:50
frinnstRomster: then wait for the new dc-series01:53
frinnstmore expensive but probably worth it01:53
Romsteri can wait awhile longer.01:54
frinnsthttp://www.anandtech.com/show/6433/intel-ssd-dc-s3700-200gb-review01:54
frinnsti *REALLY* want one of those :D01:54
frinnstmore expensive, but if you stay around 100gb it should be a nice buy01:55
RomsterSATA Express... lovely so how long before sata is obsolete -_-01:57
Romster'd imagine it'll be backwards compatible01:58
Romsteri'd*01:58
Romsterincluding the shift from a B-tree indirection table to a direct mapped flat indirection table which helped enable this increase in performance consistency01:59
Romsteri thought b-trees were fast.01:59
Romster100GB $235 by there guide02:01
Romster100GB is more than enough for root02:01
Romsterdamn if i'm reading that comparison chart correctly the write iops of that S3700 crap over the other models02:05
frinnstthen youo're not02:56
frinnst:)02:56
frinnstsome models might have higher peaks in some places02:56
frinnstbut i doubt you'll find more even and predictable performance02:57
frinnstmy 330 is great here at work running on ext402:57
frinnstat home with btrfs i get horrible lag sometimes02:57
frinnstrunning lilo can take a couple of mins if there are other semi-intensive i/o tasks running02:58
RomsterO_O with lilo that seems ridiculous02:58
frinnstdunno if its btrfs or something with the drive.. or if i've done something else wrong02:59
Romsterwonder how long i have to wait for the newer SSD than getting the 330. weeks months next quarter03:03
frinnstseems a 335 model is available too03:28
frinnstoh, only smaller nand03:29
prologicalright we're in business03:32
prologicnginx, haproxy or lighttpd for reverse proxying? :)03:32
Romstervisitors ugh04:40
Romsterwhen do i get time to myself04:40
Romsterreverse proxy for what purpose?04:41
prologicweb applications04:44
*** diverse has joined #crux-devel04:55
*** horrorSt1uck has joined #crux-devel06:19
*** horrorStruck has quit IRC06:21
*** horrorSt1uck is now known as horrorStruck06:50
*** horrorStruck has quit IRC07:01
*** horrorStruck has joined #crux-devel07:30
jaegerRomster: sustained write speed isn't a useful metric, anyway. Look for the small random read/write speeds08:19
jaegerThe VAST majority of your daily workload is going to be tiny random reads/writes08:19
jaegerfrinnst: I'd guess that's btrfs, I've never seen that on any of my SSDs, even the shit OCZ ones08:20
frinnstyeah must be something. though btrfs on spinning drives seem a lot faster08:33
frinnstoh well :)08:33
jaegerI just guessed that because I haven't used btrfs but I do have a few SSDs :)08:35
horrorStruckno one mentioned crucial, i'm very happy with my m4 but it may be considered old by today's standards08:44
jaegerI don't have personal experience with the M4 but they seem to be good from reviews and anecdotal evidence, though you have to be careful about your firmware version sometimes08:47
jaegerIt was the M4 that had the 5000 hour crash bug if I recall correctly08:47
diversejaeger: what do you use currently?08:47
jaegerI've got 2 Samsung 830 256GB, 1 Intel X25-m 128GB, 1 Intel 520 240GB, 1 OCZ Vertex 2 120GB, and 1 OCZ Vertex 60GB08:48
diversedid you raid0 them?08:50
jaegerno, bad idea :)08:50
jaegerIt's only very recently that any RAID configuration of SSDs has supported TRIM and that only on a specific chipset08:50
diverseah08:51
jaegerSo if you put them in RAID on some other configuration TRIM isn't happening and your performance degrades badly over time08:51
jaegerYou can combat it to a small extent with SSDs using controllers that have aggressive garbage collection but it's still far from ideal08:52
Romsteralso note that kernel now calls TRIM discard08:53
jaegerThat isn't new, actually08:53
diverseTRIM seems to be a mandatory feature08:54
jaegerIf you want good performance, it is08:55
jaegerThe drive works without it, certainly, just not nearly as well08:55
diverseso do you use the SSDs to store your OS core and configurations while putting your personal data on HDD or is everything you store on SSD?08:57
jaegerDepends on the machine. On my main desktop I use it for the OS and a couple games while large data is stored on platters. My laptop only has one drive so everything is on the SSD08:59
jaegerOne of them is used as a vsphere cache09:01
diverseoh, what is vsphere?09:01
jaegervmware's virtualization stuff09:02
diverseso you improve the emulation drastically?09:02
jaegerThe difference isn't particularly noticeable for my small installation, it would likely be far better in a big deployment09:03
*** sepen has joined #crux-devel09:05
sepenjue: ping09:05
diversejaeger: thanks for the insight. I might end up partitioning / for SSD and /home for HDD on my next installation.09:07
jaegernp09:07
sepenwho created 2.8 branch on contrib.git?09:07
jaegerfor some reason he really doesn't want to give specific information09:11
sepenlast 3 commits on contrib.git/2.7 are not in 2.8 and I cannot merge these commits http://46397d29d58e2a11.paste.se/ it said that branch is uptodate09:11
sepenyeah jaeger09:11
jaegersepen: if they can't be merged, cherry-pick them?09:12
sepenjue told me about subversion-perl, but I remember that I updated it to 2.709:12
sepenyep, but just I wanted to be sure that the branch 2.8 was created fine09:12
sepenI'll do that, thanks09:12
jaegerok :)09:13
jaegerjue, frinnst: this conversation in #crux has me thinking we should make a wiki page or web page similar to FreeBSD's UPDATING doc09:20
jaegerwhen things change like udev requiring new kernel config options, in addition to whatever ML posts we normally make, a handy reference of those things on the website would be nice09:21
*** sepen has quit IRC10:38
*** mike_k has joined #crux-devel12:50
*** mike_k has quit IRC13:30
*** joe9 has joined #crux-devel16:53
horrorStruckjue: wouldn't it be better to have ldns a separate port instead of building it in unbound?19:05
horrorStruckjue: https://github.com/destructure/for-jue20:02
*** __mavrick61 has quit IRC21:37
*** joe9 has quit IRC21:37
*** _mavrick61 has joined #crux-devel21:38

Generated by irclog2html.py 2.11.0 by Marius Gedminas - find it at mg.pov.lt!