/dev/chat

I have two raid sets, one for my massive data dump and one for everything else. I can lose a disk in either set and keep going. Everything gets backed up. 10 TB in the case (7 available after RAID takes its share) and 8 TB on the backup array

Braiam

1:08 AM

@Gilles I need something like regex101 but for shell to understand that... wait, got an idea

No Time

isn't RAID 1 slow?

Gilles

@NoTime not particularly, no

No Time

wouldn't 10 be better? or how many disks do you have?

Braiam

@NoTime depends the operation and with what you are comparing

No Time

I am just not sure why you wouldn't use 5 or 10

I just mean if you have the disks

Braiam

1:14 AM

21

A: What are the performance differences between Raid 0,1,5,6,10

One worthwhile location to check out is StorageReview.com's Comparison of RAID Levels But focused on the answer: LEVEL | CAPACITY | STORAGE | FAILURE | RDM READ | RDM WRITE | SEQ READ | SEQ WRITE | 0 | S * N | 100% | 0 | **** | **** | **** | **** | 1 | S ...

No Time

I could see using 0/1 (depending what you are using it for) but for data storage I would go 5/10

It seems that according to the chart linked 10 would be best overall, except for write would be 0, which would lose parity/dependability and amount of fail. Or 6 where you can lose 2 drives. Storage better in 0 too

Is it because of a Raid controller?

@Giles how many disks do you have? and are they same sizes?

* crickets *

@Braiam I understand RAID I just don't see why (if you have disks, don't mind the setup) you wouldn't choose 10, post linked seems to say the same

casey

@NoTime I have 2 disks in raid 1 for most of my stuff and 4 disks in raid 5 for my research data. Its an old setup but it works.

No Time

I mean I can understand not wanting to set it up, or that it has been working (why fix something that isn't broken) but if you had the capability which would you set up first?

casey

and my raid is software raid, so my only performance gain is write/reads spread across multiple disks

No Time

ok that makes sense too

do you have SCSI drives?

I didn't know you could do a 5 with software though

casey

1:26 AM

@NoTime the mirror is 6GB/s sata III, the raid 5 is 3gb/s sata II

No Time

what are you using?

casey

@NoTime mdadm

you can select which raid personalities it knows in kernel config / loadable modules

No Time

im looking through raid.wiki.kernel

for mdadm

@casey ok so you couldn't do a large array with your current hardware

since you have sata 2 and 3, and diff sizes

sorry I am caffeinated

casey

@NoTime I could if I wanted. 8 of my disks (including the e-SATA enclosure) are 2 TB, the other 2 are 1 TB

there is no limitation I'm aware of with mdadm that the disks be the same sata speeds

No Time

ah ok

they will just go to lowest speed then probably

casey

1:39 AM

each disk will go at its native speed

No Time

wouldn't that defeat the purpose if you used a 3/2 combo on RAID 0 then?

I guess if you had a super processor it wouldn't make a huge difference on personal "fun" stuff

casey

my processor is nothing special. i7 990X but it gets the job done

this box does all of my research model and analysis code development and testing

the real work is done by my 512 core allotment on Yellowstone

No Time

sata 3 is 2x faster than sata 2 right? I'm just trying to figure out with the software and processor in between and handling the raid (vs hardware) what the speed difference would be.. like benchmark. Might research a bit :)

casey

@NoTime in theory. in practice who knows

I'd run hardware raid if I could justify the expense on a real battery-backed raid card

but I can't, and I have a UPS between me and the grid, so I'm not worried about losing data if the power goes out

@NoTime I regularly see reads > 400 MB/sec when I'm running my analysis code coming from my raid 5 array

No Time

It's just that drives eventually that worries lots of people, good UPS are needed too

oh damn this article is old sorry

2008

Ill remove that not necessarily relevant now

casey

1:47 AM

# hdparm -tT /dev/md5

/dev/md5:
 Timing cached reads:   18708 MB in  2.00 seconds = 9362.59 MB/sec
 Timing buffered disk reads: 1144 MB in  3.00 seconds = 380.80 MB/sec

No Time

hmm won't let me

casey

which chat line do you want removed?

No Time

the one with the link.. it's only giving me "flag for moderator"

ty

it's fricken hard to find a recent set of benchmarking for software vs hardware RAID

@casey ok I read a bit more on software I see the major upside (much better if you have a bunch of diff drives)

casey

my drives are homogeneous, but yes, that is an advantage. Though if the partitions are differently sized, mdadm will use the smallest size for all of them in the set, afaik

No Time

I can learn still hooray

kinda sounds like how they had.. is it a lot of platters?.. to get around size limitations on IDE drives limitation

like how drive space was spoofed

it's so hard to concentrate my wife is having a loud phone conversation

Braiam

2:11 AM

4

Q: My windsock is erect, what does this mean?

I know that airport windsocks are calibrated to reach full erection at a particular wind velocity. At half that speed, the first half of the windsock should be erect and the second half should droop sorrowfully. At one third the speed it will be one third, and so on. Trouble is that not all win...

No Time

that is great

Ramesh

@Braiam It will end up in the hot questions list.

No Time

measure it and add 3

does anyone work with AD and openLDAP (or something similar) with any regularity

Gilles

8:52 AM

Example of us not downvoting enough @FaheemMitha @MaxVernon @Braiam : 2 upvotes on my comment that says “nonsense”, yet nobody but me downvoted, on an answer that's dangerously wrong about security

Transcript for