FORUMS: list search recent posts

Soft Raid vs Hard raid. CPU utilization.

COW Forums : RAID Set-Up

<< PREVIOUS   •   VIEW ALL   •   PRINT   •   NEXT >>
Russ D'Arensbourg
Soft Raid vs Hard raid. CPU utilization.
on May 16, 2016 at 1:57:44 am

I'm looking into purchasing a new DAS raid solution again. I'm an indie producer/director that mostly works by myself. So fibrechannel, SAS and other such multituser solutions are more than I need. Instead I'm limiting my search to Thunderbolt connected DAS. I hold out hope of a possibility of a NAS which supports link aggregation playing nice with my (2014)Mac Pro via dual gBE. But though the math makes it seem possible. I am dubious.

My main question is about CPU overhead. I've done a bit of research during this evaluation around 'soft-raid' vs 'hard-raid'. During previous upgrade cycles this question has popped up as well.
To clarify;
A 'soft-raid' (Drobo 5D, OWC Thunderbay 4)is one which uses the host computer to perform raid calculations and then sends the data to a dumb disk controller to route to appropriate disks. If it is raid 1 then it merely sends one copy of the data to two destinations. If however you are using raid 5, 6, 50 or 60, then the host computer has to calculate parity data for all your data it writes to the array. It also has to allocate the parity and data blocks across the raid volume in order to tolerate a disk failure.
A 'hard-raid' (Pegasus R4) has a dedicated controller that handles all of the math for parity data, drive allocation as well as disk host controller duties.
Why this matters to me, is that I have noticed a few reviews which ding products such as the Thunderbay 4 or Drobo 5D for essentially taking over one core of the host to run the raid. There are examples which show the host app running 100% utilization on one core for both Drobo and OWC products. I've not found examples for other soft-raid products. But then I think this may be because Drobo and OWC are two of the more popular DAS products. Hence garnering more reviews.
I am curious if anyone has any wisdom to share on real world experience with CPU utilization, raid arrays and video production.
I've noticed that most of the hard date type websites such as Anandtech and storagereview focus on multiuser scenarios or transactional database types of use. Not so much HD playback and content creation.


Salients: PP cc2016, nMP 6 core 32gb 1tb SSD. Mostly editing with some effecty music video style nonsense.
Pretty much all 29.97 or 23.97 1080p. Though I have been doing some animation in 4k and then bringing it in at HD.


Return to posts index

Jess Hartmann
Re: Soft Raid vs Hard raid. CPU utilization.
on May 16, 2016 at 3:30:55 am

Having a RAID controller (Hard RAID) maintain the stripe and redundancy patterns on the drive is the better of the two choices. Although it will generally be a bit more expensive, the key factor is that these systems have a dedicated CPU on the RAID controller which offloads your primary CPU. This is important mostly because of safety and secondly to performance.

There are conditions (although rare) where your primary CPU (Mac Pro) cannot dedicate the resources when required to maintain the RAID. In these cases software RAIDs are more susceptible to corruption. In addition, when software RAIDs have to wait on CPU resources from your primary CPU, performance can suffer.

Hope that helps.

Jess Hartmann
ProMAX Systems
http://www.promax.com


Return to posts index

Russ D'Arensbourg
Re: Soft Raid vs Hard raid. CPU utilization.
on May 16, 2016 at 4:56:21 am

I'd assume that there is some protocol uitilized by the soft raid app which asserts priority over other applications. Such that you should not have corruption of data written to or read from the array.
I've never written a driver, but as I understand it, hardware level things are not allowed to be starved of CPU attention for the benefit of the software. Not that it doesn't happen, but I haven't seen anything like that since Windows XP.

What I'm more concerned about is whether the raid app/driver is so persistent in asserting priority, that it is impactful of system performance. Most NAS and DAS performance metrics I have seen do not address this aspect. Though one review of the Drobo 5D does note that there are some intermittent throughput slowdowns. It doesn't note any dropped data.


Return to posts index


Russ D'Arensbourg
Re: Soft Raid vs Hard raid. CPU utilization.
on May 16, 2016 at 1:03:25 pm

I primarily work in plain old HD. I only stray into 4k for some bits of animation. I can suffer through a performance hit for the 4k stuff because it does not stay 4k long.

I'm not trying to buy to buy the 'cheapest possible' array. I am trying to balance performance and capacity on the middle tier of products that support TB.

I'd just pick up one of those dual drive raid 0/1/jbod products if cost was the only concern.


Return to posts index

Blaise Douros
Re: Soft Raid vs Hard raid. CPU utilization.
on May 17, 2016 at 6:31:27 pm

Yeah, OP, SCREW YOU for wanting to wring performance out of finite resources. How dare you have a limited budget and try to understand the technology you're going to be working with. Especially with video, where extra CPU usage by background apps NEVER impacts performance.

And next time, don't waste anyone's time here in the discussion forum asking people for help, when all you have to do is just know the information in the first place. NOBODY ASKS QUESTIONS HERE.

Instead, you should learn everything there is to know about a topic, and then berate, belittle, and demean people who know less than you do on a public forum. That'll make you into a big man.


Return to posts index

Bob Zelin
Re: Soft Raid vs Hard raid. CPU utilization.
on May 17, 2016 at 9:27:41 pm

ok Blaise -
here is your answer -
#1 - you need SOME money
#2 - here are your solutions -

1) Promise R4 (8TB storage - $1099)
2) Promise R6
3) Promise R8
4) G-Tech Studio XL (24 TB is $2999)
5) MaxxDigital 8 bay ThunderRaid
6) Highpoint 6324 4 bay
7) Highpoint 6328S (8 bay)
8) Highpoint Rocketstor 6324AS (hardware raid - the AS is hardware raid) - this is a 4 bay
9) Highpoint 6328AS (8 bay)
10) highpoint 6314A
11) Highpoint 6318A
12) Lacie 5 Big Thunderbolt 2
13) Lacie 12 Big Thunderbolt 2
14) Cal Digit T4
15) OWC Thunderbay 4 (there are many models, but the Thunderbay 4 is
$729 for the 4 bay 4TB, and its the entry level model).

of course, there is the Netstor solutions, but this is out of the budget that we are talking about.

SO - here we have LOTS of solutions - they will ALL work, they are all from recognized companies. And I am sure that there are others that I have not listed here.

And if your reply is "well, what if this guy only has $200 - $300 bucks - how can he do 4K editing" - well here is the answer -
YOU CANT and you should NOT be running a company - you should GET A JOB and work for someone, save up some money , and buy the right equipment so you can do your job properly.

If you cannot afford a Toyota Prius, you cannot be an UBER cab driver. If you can only afford a pawn shop $50 guitar, you cannot play in your local bar on the weekends to make $100 bucks. I don't care how much you love music.

Bob Zelin

Bob Zelin
Rescue 1, Inc.
bobzelin@icloud.com


Return to posts index


Blaise Douros
Re: Soft Raid vs Hard raid. CPU utilization.
on May 17, 2016 at 10:25:37 pm

Why change the subject? To make you look smarter? Nobody, especially not OP, said he only had $200-$300 bucks to play with. You made that up.

You still haven't addressed OP's real question, which is CPU performance of softRAID products versus dedicated hardware controllers. You've just named a bunch of midrange RAID products without addressing the actual question about whether different kinds of products will work differently.

Yes, in Zelingrad, the difference in SoftRAID vs hardware RAID performance might be negligible since everyone in that fair city has the budget to just "go buy the right equipment, right now." But the reality is that not every producer can afford to do massive upgrades all at once, and might have to wait another six months to get a more powerful CPU with buckets of processor cycles to spare. Hell, I HAVE a corporate job and I still have to upgrade my equipment incrementally, because budgeting is for smart people.

So lay off, and don't tell people trying to be smart about their options that their behavior "makes you vomit." Yours is the only emetic attitude in this thread.

OP: Anecdotally, when I switched from the OWC Mercury Elite QX2 RAID 5 enclosure with a hardware RAID controller to a Thunderbay 4 with SoftRAID, I saw a marked change in performance after idle periods. The Thunderbay 4 has a very noticeable, something like 10-second delay after being idle. If I jump into AE to do some motion graphics work, for example, and then go back to Premiere, there is a definite moment where Premiere asks for the data, and the drive goes "uhhh...okay, just a second." I did not see the same delay on the hardware-controlled QX2--that drive was connected via USB, not Thunderbolt, so I can't be 100% sure that it's not a difference in the USB connectivity vs Thunderbolt...but my money is on the SoftRAID. I have multiple Thunderbay 4's and they all do the same thing.

It hasn't been quite bothersome enough for me to go watch my processor and really figure it out, let alone return the Thunderbays. But if I had to go back and do it again, I might have tried to find a comparable hardware-controlled RAID to save my eyes the strain of rolling while the Thunderbays get going again.

There's no science in the above opinion, but that's been my experience. I hope it helps.


Return to posts index

Russ D'Arensbourg
Re: Soft Raid vs Hard raid. CPU utilization.
on May 17, 2016 at 11:15:47 pm

thanks, true my primary concern is that I do not want to drop coin on a high performance raid if that is going to impact my cpu's ability to do other tasks. If however the impact is negligble I can overlook this.
I am basically wedded to the Mac Pro that I currently have and the rest of my budget that is not covering a storage solution will be going into other stuff.
Camera stuff mostly.
As far s price vs performance I can make room for a more expensive raid if there is damning evidence that the Thunderybay 4 will make AE or PP fuck up renders or take twice as long. So far I have only seen gushing reviews of the Thunderbay, but I have to wonder about it since it is basically a 3rd party product with OWC branding and licensed software from another company.

I have noticed in my research that Qnap has some interesting NAS/DAS products due to hit the streets soon. They are TB2 DAS with gigabit NAS as well as 10GB on the upmarket units. These are very interesting since they offer expansion chassis in addition to the NAS/DAS funcitons. Not too enthused about the HDMI?

Have to say, Mr.Zelin is not impressing me as someone I would ever reach out to in a professional regard. If his business is a VAR with specialization in DR, I wouldn't want someone that grouchy around when our ISIS, Dulce or Rorke storage is giving us issues.

Yeah, at my day job I work in network sports broadcast.
So I know from storage.


Return to posts index

Scott Thomas
Re: Soft Raid vs Hard raid. CPU utilization.
on May 18, 2016 at 2:25:11 am

Bob already knows he's smart, so I doubt that he says things to make himself feel smarter. To try and summarize what Bob was saying; Time is money.

I personally have spent a lot of time and money on storage solutions for my home system that have never really performed well, but that's my money and time. At work, I'm running two 10 year old Criprico MediaVaults with Atto FibreChannel. They're way too old! I'm working on proposals to get them replaced, but we certainly have gotten our money's worth. My point is, Bob is right. When it matters, buy the best you can afford. I've never had to apologize to my boss for selecting that hardware.


Return to posts index


Russ D'Arensbourg
Re: Soft Raid vs Hard raid. CPU utilization.
on May 18, 2016 at 6:21:05 am

Whether or not to throw money at the problem is not the question. Doubtless I can get a decent solution for lets say $1200. Probably also for $2500.
Now whether to go for a smaller raid array which is computing parity in hardware at that putative price point or a larger array which relies on software is the question.

I've read elsewhere about some TB2 DAS/NAS Qnap products coming out soon. I'm also looking into TB to M.2 storage solutions with softraid as a possibility.

Either way I'm not going to 'just buy any raid from the advertisers'. I've seen Dulce, Rorke and other so called pro storage products really take a dump. And I just dont spend my money rashly. I'm always trying to see the emperor without their clothes on so to speak, before I open my wallet.


Return to posts index

Bob Zelin
Re: Soft Raid vs Hard raid. CPU utilization.
on May 19, 2016 at 1:32:22 pm

If you have very little money ($1500) you can purchase the incredibly wonderful QNAP TVS-863+ which has two 1G ports, and one 10G port build in (just add drives). And if you want a thunderbolt DAS connection that can be also used as a wonderful hi speed NAS, you get the QNAP TVS-871T, which is about $2700 (just add drives), and you get TWO 10G ports, four 1G ports, 8 slots for drives, and of course the thunderbolt looping ports. The performance on these boxes are incredible.

As for hardware raid, I have historically been using hardware raid products, from ATTO, Highpoint, and Areca. All wonderful. But of course, they cost money. Software RAID products are inexpensive (SoftRAID, owned by OWC) is teriffic as well.

As for your Dulce - that was the first time I ever used an Areca hardware raid controller, and how I discovered Areca. It's a great product, but like countless companies, it's hard to compete, and market your product, and with all the competition out there, they could not keep it afloat. But it was always a great product. As for the other old dogs, like Ciprico, Rorke Data, etc. - these are OLD products, and your boss has to get used to the fact (and I am sure he is used to it if he has been doing this for a while) that all this junk becomes obsolete. Very quickly. And then it becomes a boat anchor - just like all the $50,000 Sony VTR's.

So when we discuss wonderful new products like the QNAP TVS-871T - let me assure you that QNAP will probably no longer manufacture it or support it 5 years from now, and everyone will start to cry "wah, I have all my data on there, and I need to expand it, and no one supports Thunderbolt 2 anymore". That's our business. I am cranky because I deal with this stuff EVERY DAMN DAY - all this expensive stuff (and cheap stuff) becomes obsolete. This is how most of the companies in our industry have collapsed over the years. From buying expensive stuff, not getting their return on investment, and then they are screwed. And companies today support their products for shorter times then ever before. And the companies with "lifetime warrantees" go out of business, so at some point there is no more support.

In our lovely business, you are doomed to KEEP SPENDING MONEY over and over and over again, and keep seeing your investment become obsolete, with new faster cheaper products coming out every year.
You just have to keep ahead of the flood waters. After a while, it gets you angry. I have been doing this for a long time.

Bob Zelin

Bob Zelin
Rescue 1, Inc.
bobzelin@icloud.com


Return to posts index

Scott Thomas
Re: Soft Raid vs Hard raid. CPU utilization.
on May 20, 2016 at 10:35:56 pm

There was someone who did CGI (Lightwave I think) work who also wrote for the trades. His solution was to buy equipment for a production and quickly turn around and sell it after completion of the project. That doesn't really work for post production workflows, but perhaps there's something to that. Sell the system and upgrade while the old hardware still has value. Easier said than done.


Return to posts index


Russ D'Arensbourg
Re: Soft Raid vs Hard raid. CPU utilization.
on May 25, 2016 at 1:29:25 pm

I tend to beat equipment to death when it comes to computer gear. Sound equipment (I used to run a sound reinforcement company) I'd flip every year.

I am just going to go ahead and get the ThunderBay.
It seems that nowwhere that I have looked on the internet is there anyone evaluating disk systems with an eye towards the specs that I am interested in.
For a lot of the consumer/SOHO NAS systems the reviews focus on VM usage, simultaneous IO ability (which is important in a group use scenario) and gee whiz features like mobile apps.

The reviews of DAS disk systems seem to focus solely on throughput figures. Which of course are important numbers to post-production users.

It still bothers me that nobody really looks at cpu utilization. While this may be an insignificant statistic if you are using a single array. If you end up using solutions from different manufacturers simultaneously, this has the potential to impact performance.

I suppose I've developed a bit of anal retentiveness when I was building DAWs in the 90's and 2000's. Back then you had to sweat every cpu stealing process and memory hog. Otherwise you might get audio dropouts in the midst of a multitrack session. Even with high performance, professional grade hardware.


Return to posts index

<< PREVIOUS   •   VIEW ALL   •   PRINT   •   NEXT >>
© 2017 CreativeCOW.net All Rights Reserved
[TOP]