RAID 5 & 6 are professional forms of RAID for hard drives and SSDs. This brief overview aims to give you a basic understanding of how they work FORUM LINK: linustechtips.c...
I thought only Germans used their thumb, index, and middle finger? XD Seriously though, way more comfortable on the hand to express 3 that way. Holding the pinkie down with the thumb feels... unnecessarily strenuous.
@@matthewb1601 completely the opposite...for me at least. Using the thumb l, pointer and middle finger to express "3" is more on the painful side for my ring finger.
"Rebuilding the array... can be time consuming." That's a huge understatement. It takes upwards of several DAYS to rebuild larger arrays (say, a dozen terabytes or so). I work for a NAS manufacturer in tech support and we always get calls and tickets from people who expect a huge enterprise RAID 5 or 6 to rebuild in a much more timely manner. They usually think our products bottleneck the process somehow when it's simply a limitation of current HDD speeds, which have lagged way behind storage capacity. Also, for the love of God, back up your data. Always. It doesn't matter how many redundant drives you have. If they're all connected to the same motherboard, THEY ARE NOT BACKUPS.
Like I said, read and write performance has not advanced with HDDs the way storage capacity has. If you're rebuilding after losing an 8TB drive you've got way more data to copy than you would have had a few years ago, but it's not going to write much faster than it would have a few years ago.
last month I got a failed disk in a raid 6 array with 15 drives of 6tb. It took just a few hours to rebuild and the server was being used during the process.
I use RAID6, but I am considering upgrading to RAID61, which is a RAID 6 (2 drives can fail) with an entire matching RAID 6 mirroring the first one, so in an 8 drive array up to 10 drives could fail without data loss :p
These vids have so much well explained info that if you watch a few dozen in a row the amount of knowledge gained is pretty phenomenal compared to the efficiency of school or university. It's like getting an extra brain
Answered my question in - as you say - "as fast as possible". I'm working on a new server with 4 drives and had the option to choose between RAID5 and RAID6, so your discussion about impact on systems with smaller# vs. larger# of drives was helpful.
Thank you very much Linus, we have this in our exams and you helped clear it up. I was literally taking notes while you went on explaining about the RAID setups. Again, thank you very much.
So, just so I am clear on this. In RAID 0, you get 1/2 the info in each drive, so failing of one drive is catastrophic, but write speed is increased with each drive. RAID 1 is the opposite of RAID 0, and you have 2 copies of each drive. Raid 5 has 1 "Safe" drive, and has the rest as RAID 0's. So if one drive fails, it can be fixed, and doesn't suffer too much from write speeds. RAID 6 has 2 "Safe" drives with the full information, and the rest are RAID 0's so 2 drives can fail, and it can still be re-built. Finally, RAID 10 is 1&0 so there is one pair of RAID 0, and one pair of RAID 1. This means if one from each category fails, you can still fully recover, and have nice disk speeds.
RAID 10 is not one pair of RAID 0 and one pair of RAID 1. It requires min 4 drives. It forms two pairs of RAID 1 and uses RAID 0 way of stripping (strip+mirror). And people are saying Linus discussed RAID 3 instead of RAID 5 because RAID 5 uses some parity thing in it.
It holds a "check" value of the other drives, which allows you to work backward and figure out what the failed drive had. It's more calculation heavy which is why it takes ages to build the array as well as rebuild it after a failure.
With the magic of editing, you can turn your bad hair day into a good hear day! ˙sɯɐǝɹʇsǝʌıן uı ʞɹoʍ ʇı sǝop ɹou ǝɟıן ןɐǝɹ uı ʞɹoʍ ʇou sǝop sıɥʇ ˙soǝpıʌ uo sʞɹoʍ ʎןuo sıɥʇ :ɹǝɯıɐןɔsıp
question is.. did he get his haircut on set? or did he get his haircut offset a different day.. finished filming and just wore the same shirt to make it look like it was all the same day? regardless... did he stop filming because he didnt like his hair? Is he that guy who goes ape-shit on the lighting guy as they're trying to film? lol
I've love Linus's vid's for a while and still enjoy them I am even more happy to know that Techquickie was created on my birthday friggin' AWESOME! thank you Linus for helping me with computer technology I've learned so much from watching your videos keep up the great work.
linus you have taught me probably 60 percent of what i know about pcs, for that i thank you. i noticed you dont have a video on how raid cards work and how to use them, at least i did not find one if you did. can u plz do one!! im subscribed so ill see it :)
Can you make a video on pros and cons of watercooling. It can be quick, but before I went into WC myself, it was daunting and some reassurance would be great for many others. Cheers
@wheremyjaffa that's easy enough. So for a HD XXXX scheme, the first X denotes the architecture series, the second X denotes the tier of the architecture process (e.g. The 79XX is 'better' in numerous ways as opposed to 78XX or 77XX), the third X denotes the tier within the tier (I.e. a 7950 is a downclocked version of a 7970 with its computational units restricted), and finally the fourth X is for OEM use, most usually seen within mobile computers.
Hey! Dezel here. The most important aspects of a mouse should be comfort and software/drivers. I mostly play at 800 DPI with most mice I've had. The differences will show in durability, sensor reliability, and not in DPI. It's the same with cameras. It's not all about the megapixels.
RAID is good for keeping you up & running and avoiding down time. Backups are what save you from ultimate data loss. Also, completely off-site storage isn't mandatory in the absolute: Just keep an air-gapped backup (meaning: NOT connected 24/7, not connected at any other time of day/week/month you do the actual backup) in a fireproof lockbox or safe - if the premises burns down, gets flooded or your equipment gets otherwise completely trashed, you haven't lost absolutely everything.
Think of raid 5 as this: You have 2 drives for data that splits them in half just like in raid 0 (ex drive A and drive B), and you have third one (drive R) that writes only 1 if both drives (A and B) wrote two 0s or two 1s at the same location, or R writes 0 is one drive wrote 1 and other wrote 0 at same location on drives. So you are basically backing up two drives of data (A and B), with only one extra drive (R), so if one of them fails (A, B, or R), you can use other two to recover data for third one.
This is mostly accurate, but a disk that is dedicated solely to parity data only existed in earlier versions of RAID (3 and 4 specifically), which are now obsolete. RAID 5 and 6 are different because the parity data is distributed across all of the disks, which gets better read performance (especially for random reads). Also, RAID 5, 6, and 10 all have advantages and disadvantages that make the best choice a matter of what's more important to you between performance, storage space, and fault tolerance. If you're a home user the simplest option is just RAID 1 with two disks, plus an external backup of some kind.
@wheremyjaffa also, a XX9X scheme is used to denote a dual-gpu card. The dual-GPU almost invariably being two of the top tier cards in one. (Note, all this information is relative to the current generation of cards and the proposed next generation, the next generation however drops the first X identifier for the architecture as it seeks to consolidate the numerous card series to one. Simply, AMD is still working on the naming in a hopes to make it easier to understand)
Would have been nice to have had the diagram that you had for RAID 5 on RAID 6. Consistency is always nice, and it'd make it easy to understand if you use the diagrams.
@DerfHD Raid 10 simply means that you have a RAID 0 array (which is effectively one disk) with raid 1 redundancy. The 0 means that each disk is made of two or more disks in RAID 0, and the 1, the 5 and the 6 then tell you what RAID configuration those RAID 0 arrays run in. This approach is designed for speed, but you are better off using an SSD to max out a SATA 6GB/s interface if you want speed.
One thing you should mention is the fact that RAID 5 is usually a pretty bad idea with huge drives. During the rebuilding process (which is going to take FOREVER on a software controller) if another drive goes you are pretty screwed. Dual parity in these occasions is really the only safe way to go. At that point for home users you are probably better off going with a software based ZFS based solution over a linux distro for the cost benefit.
i have to say that i run RAID 5 on my home storage pc with 5 3tb HDD in it. this pc (other then the HDD) was my old gaming PC that i have upgraded. it is used to house all the DVDs that i have got over the years and allow me to get to them where ever i go along with programs and files that i want to get to. i found this to be cheaper them going out can getting a NAS that can hold 3-5 drives and because my MB can do a RAID 5 much cheaper (not so much easier) then getting a RAID controller card. $500-$700 USD for the NAS vs i think i spent like $250 USD on the HDD. but what do you all think
no this is more fast as possible info if you or anyone you know has a question that can be answered by these vids. linus tech will still deal with unboxings, benchmarking and more in depth stuff. and ncixcom will still have netlinked weekly as well as other NCIX related content : )
they could add a part 3 for virtualized raid arrays, and similar tech like dynamic disks, as well as added information for comparison information of SSD arrays in these configurations
For most users, we may end up using RAID 10. Linus always goes for RAED 10 swapping out our mid range 256TB SSDs for top end 2TB SSDs making it an array of Expensive drives.
Working on my NAS right now, just got a bunch of sata cables i needed. I used to think raid 6 was useless for home users, even on massive NAS. I mean, when are you gonna have 2 drives fail at once? Dont you want that space? Well, that drive isn't doing redundancy when you are expanding an array. So one error, and your data is gone. Definitely glad im going with a raid 6
I might just be a bit dumb, but it took me awhile to figure out raid 5, and if anybody wants my explanation, here it is. (Correct me if I'm wrong - I don't know a terrible amount.) (TL:DR @ Bottom) You have your 3 drives (A,B,C) A/B Drives are the data, storing it by splitting the data. C drive backs up both using XOR. That is quite easy, but how XOR saves it is something I didn't know, so here it is. Say you have your data split, and it shows up in 3bit (for the explanation) binary as a = 101 b = 011 C saves this by comparing 101 to 011 in XOR. (If you have one 1 after comparing a single slot, you get a "1", if you get 11 or 00, you get a "0") XOR 101 and 011 gives us 110. (1 and 0 = 1 ------ 0 and 1 = 1 ------- 1 and 1 = 0) (Comparing the two sets) So C = 110 If B fails, we can use C to figure out what it was A = 101 we know, and C=110 Using XOR to compare, you get 0 (1 and 1) 1 (0 and 1) 1 (1 and 0) So we determined B = 011 Checking the first B, this is correct This also works in more than three drives and 3bits using the same method A = 101 B = 111 C = 100 D = 000 E = ? compare in sets of 2 101 | 111 = 010 100 | 000 = 100 Compare the two 010 | 100 = 110 E = 110 Say drive C fails Compare E to A,B, and D Acom = 011 Bcom = 001 Dcom = 110 Compare AB 011 | 001 = 010 Compare 010 to Dcom 110 | 010 = 100 CGuess = 100 Real C = 100 (This also works if you compare A - D, B-D, Etc. If you want to try it - go ahead.) (This works with more than 3 bits, but for the purpose of explanation, it makes it easy.) TL:DR = Make the last drive the XOR of the rest, and then compare it to the good ones if one goes down.
It really is going to depend on a lot of factors, but for the amount of drives you have I would try a RAID 10 a RAID 1 will give you half of your drive capacity and honestly no one needs a 23 drive failure redundancy. RAID 0 would be a speed demon, but it would require constant back ups in order to ensure you don't lose the entire array. RAID 10 gives you the best of both worlds especially with the amount of drives you happen to be using.
I never had a hard time replacing a drive in a RAID 5 SCSI array. If hot-swappable, it's easy and the volume can be rebuilt ...for the next day. That was a while ago. Writing/reading to the array, however, is much faster, not slower. Are you speaking of some other type of array technology? I used to use a big tape library for back-up. What is used these days? Another array?
Probably wont read this but can you do a video on how overclocking voltages affects the lifespan of your parts? Or is it too complicated to cover in this small time frame format?
I know Linus many years since I love computer stuff...until I see this, he said five then six, suddenly I pause and look again...actually when he said 3 or 6, it was ASL! I am deaf and I know the ASL. Didn't know he knew ASL until today!!
Professional applications use ZFS to take care of bit rot and system crashes / power fails during writing too. Besides rebuilding the storage is considerably faster.
1:34 True'ish it will give you (2) two separate logical volumes. Volume 1 of RAID 1 [Disk 1 & Disk 2] and Volume 2 of RAID 1 [Disk 3 & Disk 4]. RAID 1 is limited to two (2) drives. You 'can' have a hot spare, but that's three drives.
per the live stream: Seagate gave them the drives for their ces footage backup machine to dump their cameras while they shot -- buying them a tag on the ces videos.
very neat information tnx, thou 2 quick question windows & bios these days easily provide raid software are they practical and which one is better other one for small offices
linus you forgot to mention that for raid 5 and 6 you need either a dedicated controller OR a fully softare solution like linux's mdadm. I'm running mint debian edition on a 4 samsung F3 mdadm raid5 and had none of the problems you mention. Even at the begining when i had to sync the fourth disk iotop showed that the raid was running at maximum speed at about 320MB/s compared to 456MB/s i had in raid0. and the cpu never got above 40% on one core, and i have 6 (1055T).
With a dedicated raid card, you'll never see CPU usage climb above 5% meaning you can use 35% more cycles for actual useful work rather than I/O shuffling. Having tested this myself, before and after using a RAID card with a performance tester, I first thought something was wrong with the setup after installing the card because it was pinging so low while reaching high throughput.
Not sure why he says RAID 5 and 6 are not good for consumers. I use a Synology 1511+ NAS with 5 drives on RAID 6 at home. As if that isn't enough, I back up to a cloud service. My data is better protected than a lot of small businesses and I sleep better at night because of it!
Your description of RAID 5 is actually RAID 3. RAID 5 Stripes the parity across all the disks as well as striping the data. You should really update this.
I am probably pretty late to this but after watching this video I feel I should say something, sorry @Techquickie but this is incorrect :( What you actually described in this video is RAID 3 or 4. Raid 3 and 4 needs a minimum of 3 HDD's, two for striping data across and one for a disk array parity. RAID 3 generates the parity at a byte level whereas RAID 4 does it at a word level. The problem with this approach is that the parity drive becomes a bottleneck because every write needs to access the parity drive. RAID 5 was actually created to remove this problem by stripping the parity throughout all drive arrays, you still have the capacity of N-1 drives because of this, but you are no longer limited to accessing the parity on one drive. RAID 6 is an extension of RAID 5, the common RAID 6 creates 2 parity bits, one for the traditional horizontal array and a vertical one as well (most common, but not only method)
interesting videos, is a vps hosting with raid 10 a good choice ? considering a machine that splits into multiple other vps' and one folk consum a lot will it affect the others on raid 10 ?
That was some seriously wrong info. Just about everything was wrong. For example... RAID 1 would only provide 3TB of usable storage no matter how many drives you have in the array. You didn't even cover the most important risk information for parity RAID.
If it issint too much trouble could you do a video explaining how DDR is different from GDDR and also how is GDDR5 better than GDDR3...Also is there a GDDR4 ?
Dumb question, but he says that the array can sustain failures because of the redundant drive, but what happens when u use more than the 3tb that the redundant drive in the raid 5 array can mirror?
