Building NAS, Differences between software raid, onboard RAID controller, PCIe RAID controller (16x), sas and scsi?

MDF

Reputable
Sep 3, 2014
7
0
4,510
It has been a while since I tried to build a system, so I'm kind of outdated. As an indication, I was very surprised to see that the ribbon connectors are replaced...

Now I'm trying to build a small home server (NAS, file, email, media, ftp and sql dbase).

With this server I believe a few point are most important.
1. About 9 clients will need direct and simultaneously access (with the least delay).
2. These clients will read write large files (about 100gb to 2tb per file / time) to and from the NAS over a GB network.
3. There will be heavy SQL dbase analyzes on the data stored on the NAS.
4. There should be around 16 to 20 TB of storage.
4. The read write speed from the NAS should be between 500 MB and 1 GB/s.
5. There should be some form of redundancy.

Sadly my knowledge seems to be very outdated. In the past you would have used a scsi controller, so that multiple disks could work in unison and the data would be divided across the multiple disks in an array, which increased the speed and stability and added some redundancy. I think this will not have changed that much.

But nowadays it seems that everyone is talking about Raid controllers on the internet. Do these raid controllers do the same thing as the old scsi controllers and what’s the difference between a scsi controller, sas controller and raid controller?

Furthermore I noticed that there are a lot of options with these raid systems (controllers (PCI PCIe (4x, 8x, 16x), software and even onboard?). Is it still necessary to have such a controller, or are the software based and onboard systems strong enough?

So what kind of hardware should I be looking for? I already looked around on the internet, but there seems to be a lot of discussion.

But I came up with sort of a list:
- mother board with a fixed 8core Atom CPU (2750) with probably two PCIe (16x) slots (those for the raid controller and network adapter as listed below). Unfortunately I can’t find one;
- presumably 4GB or 8 GB internal;
- a good raid controller (PCIe 16x) with approximately six to eight ports (6GB/s. I coudn’t find anything faster);
- a four-port GB network adapter (PCIe 16x presumably);
- 4 to 8 3 or 4 TB drives;
- silent fans, quiet power supply;
- small housing.

In connection with the discs, there appears to some discussion about whether you should choose the Seagate NAS drives or Western Digital Red's. What is the discussion about? Is it just the brand?

And what kind of read /write speeds can be achieved with such a system with at least four constant connections (hence the four GB ports, intended load balancing aggregation) with a large amount of moving data (say first computer streaming a movie, second computer listening to music, third computer making a file backup to the NAS, fourth computer performing an analysis on the SQL database).

If anyone can help me with my questions (for example, the motherboard, or disks, or in general notes about possible bottlenecks / overkill that / I do not see) that would be very nice to me.
 
Solution
before i even start i should probably point out that 99% of the users here build home storage just for their pirated music and pr0n collection. so i don't think you will get much useful information on these boards.

having said that, i am one of the few people that has experience with this sort of junk. I work for a IT consulting company. i will only address the things i know very well.

-Use WD Red drives. they are far more reliable. while all modern consumer grade drives are garbage compared to what they were just a few years ago, the WD Red drives have almost acceptable reliability for head duty applications like you are trying to achieve. Several of our clients have small dell file servers. the seagate drives die constantly. it's...

terroralpha

Distinguished
before i even start i should probably point out that 99% of the users here build home storage just for their pirated music and pr0n collection. so i don't think you will get much useful information on these boards.

having said that, i am one of the few people that has experience with this sort of junk. I work for a IT consulting company. i will only address the things i know very well.

-Use WD Red drives. they are far more reliable. while all modern consumer grade drives are garbage compared to what they were just a few years ago, the WD Red drives have almost acceptable reliability for head duty applications like you are trying to achieve. Several of our clients have small dell file servers. the seagate drives die constantly. it's just ridiculous. on average we swap like 5 seagates out a month as opposed to like 1 WD drive every few months while more of our servers are actually equipped with WD drives. go figure.

-software RAID is junk. stay away unless you want to lose data. it's geared for home desktop users who want a false sense of security

-use RAID 6. you lose 2 drives worth of storage as opposed to 1 drive in RAID 5 but is actually a little faster. it is also safer since you can lose 2 drives without losing the entire volume. it is also less susceptible to the "write hole" phenomenon than RAID 5 and RAID 1.

-You can forget about read/write speeds of over 350MB. if you want 500MB/s + read/write you need a full fledged server with 10k/15k RPM SAS drives and with your storage requirements you are looking at a $10,000 invoice. the theoretical MAX of the SATA III (aka SATA 6Gb/s) is 600MB/s but you will be lucky if you hit 500. using a RAID array with parity will slow your write speeds down a bit. but the other problem is the speed of the drives themselves. they top out at 140MB/s.

that's really all i can comment on. but as i look at your other requirements, i think you will be better off purchasing an already built diskless NAS system.
i think your best bet is the Synology DS1513+
it can take 5 drives of up to 4TB each
it has 4 gigabit lan ports that support teaming
it does all kinds of funky RAID arrays
it has a laundry list of other advanced features you may like, like bandwidth control, Windows ACL, VPN support, etc, etc, etc
 
Solution

MDF

Reputable
Sep 3, 2014
7
0
4,510
Terroralpha, thank for your answer, especially regarding the Seagate and WD Red drives. I looked at the Synology DS1513 and I believe it's excellent. I'll keep it in mind, but I believe buying a home NAS kinda takes the fun out of it. Maybe it would be better to skip the atom board an processor and go for a multi CPU Xeon system (trying to keep the electricity bill down). But I won’t get in to that right now.

I don’t know if it’s allowed to add comments on to my first post, so let me know if I should start a new one. But I find your comment on the drives speeds very interesting. If I understand you correctly you are saying that adding on drives might even slow the read/write speeds down a bit.

I was under the impression that by having all the drives controlled by a scsi card/raid controller, they could al operate at the same time. Meaning that the drives are always running/turning and you wouldn’t los speed by accessing a new drive.

Furthermore I was under the impression that having more drives containing exact copy's of the same data meant that you could read part of the requested data on different drives. This would mean that the read speed should be calculated by adding the read speeds of all the separate drives. Say 100MB/s per drive, five drives, meaning a total accessed reading of 500MB/s.

While I’m writing this post I suddenly realise that the formentioned would mean that writing to the drives means that every drive should get the same data and therefore wouldn’t function faster than a single drive, unless there would be some sort of smart way of rebuilding the distribution of the data while drives are idle. I can theorise that this could be done by the two redundant drives in the raid 6 configuration, switching active drives into the redundancy pool and redundant drives to the active pool. This should create a speedup, but this would probably only work for a short while and the speed would eventually decrease while continually reading and writing large amounts of data. But who am I.

I’m going to think about your answer. If you or anyone else has more information or comments I’m very interested.


 

TyrOd

Honorable
Aug 16, 2013
527
0
11,160
Good god. Just don't do it. If you don't know the basics of RAID you have no business building this kind of system for anyone.

Also sadly most IT people will give you the advice to go with something like RAID6 because they believe it offers particularly good data protection.

Unfortunately, RAID is NOT IN ANY WAY A BACKUP. it is redundancy built to eliminate downtime and will not replace a proper separate dedicated backup device.

I don't know how many times I've said that RAID isn't backup on these forums, but it's worth saying again. Assuming RAID is like backup is one of the BIGGEST reasons why data recovery companies get tons of Enterprise storage cases in all the time.
 

MDF

Reputable
Sep 3, 2014
7
0
4,510
TyrOd, thanks for the tip. But I'm aware that raid isn't a BU and I'm learning about Raid. I seem to be kinda outdated, but i'll get there before i'll start buying an building. You guys ar very helpfull and these boards are excellent. I'm looking for for some practial information. Pcie 16x needend for 4 port nic? Pcie 16x needed for a good raid controller or are the drives, or the bus on the M-board always the bottelneck? Any suggestions regarding the 4port nic and/or controller. If the 16x Pcie doe matter, do you know a mini itx board with two ports?
 

terroralpha

Distinguished
@MDF
That synology unit is not a home unit. it's geared to small business that need a sizable storage solution at low cost. i've seen $10+ million revenue business run on these. you just plug in one or two external USB drives into it to do daily back ups but otherwise they work great.

as far as your understanding of RAID functionality and just storage technology goes, well, all i can suggest is that you read up on it. I can't really type a whole computer science lesson up here but all I'll say is that you are way off.



woah, slow down there. this guys appears to somewhat know what he is doing. i think he knows that RAID is not a backup. he just needs a large storage volume with some redundancy. I suggested RAID 6 because consumer grade drives, like WD RED and Seagate NAS series, have approximately ZERO quality control and will need to replaced often with the workload he is planning to put on them. I've lost an entire volume due to back to back drive loses in a 12 hour span on a single array. where i work we use tape drives (Dell PowerVault 124T LTO6 units) to back up our servers nightly, but the down time caused by an inoperable volume can't be patched by any back up solution. RAID 6 would have saved my bacon that time.

on a small storage solution RAID 6 is useful because it takes like a week to get a manufacturer to replace stand alone drives. RAID 6 gives you some breathing room when it comes to waiting on replacement on dead drives.
 

MDF

Reputable
Sep 3, 2014
7
0
4,510
Guys thanks for your help, but don't argue. Hahaha.

All your help is welkom. I know small businesses could function on the suggested NAS. But the fun is in building it and i need read and write speeds for sql analyses on multiple tables of 20mio lines and more, without freezing the system. I'm goimg to readup on the raid systems and some redundancy is exactly what's needed. It's just that it's been a long time and I'm trying to catchup.
 

TyrOd

Honorable
Aug 16, 2013
527
0
11,160



First I'd like to apologize for being harsh, i hope the OP can get the help he needs here.
Secondly, we're not disagreeing actually, just a misunderstanding.
Yes he needs redundancy for exactly the reasons you described and he ALSO needs backup in the way you described for separate reasons.
RAID6 on a mid-range synology system plus backup in the form of direct attached drives is a solid solution.A tape library/autoloader as backup is a solid solution as well. really anything is better than RAID alone.
I actually work for a major data recovery lab that gets in enterprise RAID with inadequate backups fail every day, so I always remind people of that first.

tl;dr: RAID + Backup = minimal downtime and safe data.
 

MDF

Reputable
Sep 3, 2014
7
0
4,510
I've read up on the way Raid Array's work. My conclusion is that a Raid Array won't really increase the read/write speeds. Especially regarding big data analyses it won't be a solution. The ammount of data is to large to read into the memory of the system, so it will always keep using the drives. At the moment a singel query takes between 15 minutes and two hours to run from an 840 EVO 1TB. Seeing that the 1Tb isn't large enough the only way to speed things up will probably be building an array of ssd's. I found a very funny youtube movie of a group of people that build a Raid Array out of 24 ssd's.

So I'm going to ad two more ssd's to my desktop so that I can build a Raid 0 Array of 3TB and use two normals 3TB Harddrives as BU, and switch these out each day. If one of you could advise me on the best Raid controler for this system I'd be very greatfull.

Furthermore I'm still going to build a NAS as a central file storage system. I'm tinking about a system comparable to the Synology DS1513 (it's really great). I've you have any advise on the brand of the 4 port NIC and the Raid controller I'd also be very greatfull.

Thanks,
 

Benteknik

Reputable
Oct 21, 2014
19
0
4,520
Great post , i just wanna add and did some searches , where can i setup NAS RAID 1 with dual 15000 rpm HD , been use raptor in the past for gaming , and i realized , it is still one of appealing product till these days... around 4 TB less than 500$
 

MDF

Reputable
Sep 3, 2014
7
0
4,510


Update: build the first part of the NAS. Didn't add the nic and raid controller yet. First need cash :)
At the moment i'm fiddeling around with freenas and trying to get that to work properly. Strange thing is that you can't add drives to an already existing raid. This means that you can actually only build a raid out of disks without data on it, and the drives that already have data on it will never become part of the raid. Seems to me that this must be a corporate thing that's aimed at selling more drives...

Anyway, thank you guys for all your help!

@Bentenik

I don't really understand your question. But maybe I'm not really the person your asking. Maybe the guys that answerd my questions can help with your problem.