Norco Rackmounts: Issues & Improvements

[PhatalOne]

Bubba,

 

Thanks for posting the link. It's an interesting read and integrates well with what I've been reading but does present new information that I'm not sure how to integrate.  For example, if you were to put two identical fans (25cfm, 5mm H₂O) in a series, would combined performance of these fans be 25cfm 10mm H₂O? 

 

Also, could you give me some input on the "leakage" formula from my above post. Basically, I was trying to quantify the effect on airflow using only the main fan plate with high flow fans. Would the pressure increase in the motherboard area and the resulting leakage through the back panel be great enough that adding additional rear fans would prove fruitless? Would the placement of additional rear fans be beneficial no matter what because they help relieve pressure from the motherboard area and thus increase the flow rate of the main fans?

 

Finally, are there any other good resources I can look at to learn more? Thanks for your help.

[bubbaQ]

(25cfm, 5mm H2O) in a series, would combined performance of these fans be 25cfm 10mm H2O? 

 

Yes, and no.  If you have a fixed system, and did so, no you would not.... you would get higher flowrate and a lower pressure... like 32 cfm at 7 mm H2O.  In order to get the directly additive results, you would have to restrict outflow from the second fan to raise the pressure to 10mm H2O, and then you would have close to 25 cmf assuming no interference and perfectly turbulent flow.

 

Leakage is not the problem - you can fix leakage to the point it is negligable.... the bigger problem is that as pressure in the mobo area increases, you affect performance of the fans exhausting air from the front into the back.

 

This is why the idea to blank the tailplates to prevent air being drawn in to the mobo area is silly.... there is POSITIVE pressure in the mobo area so no air is going to be drawn in through the vent holes in the tailplates.  Absent come form of channeling, you WANT the vent holes in the mobo area.

 

Each chamber will have a profile curve of pressure v. airflow.  For the front chamber, it will vary with every different configuration of drive bays occupied/vacant, and with the positive pressure in chamber 2.  You design the airflow (which will also set the pressure) in the front chamber that you need to achieve the proper cooling of the drives.  Given that airflow, then you select the max positive pressure you will allow in chamber 2.  That will spec the chamber 2 exhaust fans, and tell you the performance your front fans have to deliver (i.e. the positive pressure they have to work against).

 

Fan performance curves are just that -- curves.  Max pressure at zero airflow doesn't mean a thing here.  You have to have the whole curve, or at least the pressure at the flowrate you want.... and you shouldn't use curves of free-air measurements, since you don't have free air!  Where you want fan curves in a non-free-air condition, you have to do your own in the lab since most manufacturers don't ever do those.

 

I've been an engineer for over 20 years (Clemson 1987).  When designing this type of cooling you don't do it on paper.  You can get some very rough ideas of minimums you need, but actual spec and selection of fans will always be based on tests of an actual sample.  This is particularly true here, where there is a WIDE selection of fans with vastly different performance curves that you can chose from.

 

In my mind, the vastly different profiles for the front chamber (not the fans) that you get based on different configurations of occupied/vacant drive bays is a huge complication.  If I was designing this, I would try to make the front chamber profile as consistent as possible, with some sort of blanking plate in the front of the trays that would flip up if the bay is unoccupied, and fold down when the bay is occupied.

 

But the better solution is an active feedback loop... temp sensors on the drives, and vary fan speed to achieve target temperatures.  That solves all the problems of sizing and performance, but you introduce a potential for cascading failure and equipment damage if there is a failure of the fan control subsystem.

[PhatalOne]

But the better solution is an active feedback loop... temp sensors on the drives, and vary fan speed to achieve target temperatures.  That solves all the problems of sizing and performance, but you introduce a potential for cascading failure and equipment damage if there is a failure of the fan control subsystem.[/Quote]

 

Here too, I agree that a active feedback loop is a better solution and I had proposed a variation of that loop in a prior post.  However, as you've stated, the potential of equipment damage from a cascading failure has made this one of the lower ranking options on my list of things to try.

 

This is why the idea to blank the tailplates to prevent air being drawn in to the mobo area is silly.... there is POSITIVE pressure in the mobo area so no air is going to be drawn in through the vent holes in the tailplates.  Absent come form of channeling, you WANT the vent holes in the mobo area.

 

In my extreme test case of the 140mm fan plate, I used 2 80CFM 80mm fans on the rear panel.  In this case, I found that these fans created a negative pressure area that their inflows actually pulled outside air through the back panels and PSU vents.  This was why one of my solutions would be to seal the real panels.  In the stock Norco configuration, I completely agree with you.

 

 

Leakage is not the problem - you can fix leakage to the point it is negligible.... the bigger problem is that as pressure in the mobo area increases, you affect performance of the fans exhausting air from the front into the back.

 

The leakage calculation was not attempting to say that leakage was a bad thing in this case.  I was attempting quantify, under a test circumstance, how much air would exit the rear of the case solely due to the pressure difference in the motherboard area and the outside air.  With your experience, does my "leakage" calculation look plausible for our case or is there a better equation I can use?

 

 

I've been an engineer for over 20 years (Clemson 1987).  When designing this type of cooling you don't do it on paper.  You can get some very rough ideas of minimums you need, but actual spec and selection of fans will always be based on tests of an actual sample.  This is particularly true here, where there is a WIDE selection of fans with vastly different performance curves that you can chose from.

 

My background is in Financial Accounting and Financial Planning & Analysis (Wright State 2005).  As you can probably guess, fluid dynamics was not a prerequisite for my field; I definitely realize that I am working in waters well outside my chosen profession.  I completely agree with your views and prefer to build prototypes to test the effectiveness of my ideas.  However, I have found through practice (mostly with this case) that my very rough ideas need to be based on actual science rather than a SWAG.

 

As I've been researching and discussing this topic with you, I believe I have significantly improved my understanding of the actual physics involved with solving this problem.  I still have a long way to go, but my main goal is to understand the various aspects of this problem so that I can construct an effective solution to achieve the maintain the desired thermal requirements.  This is why I have been trying to apply simple concepts that may not immediately seem useful to the problem at hand. 

 

For instance, the static pressure of an enclosed space seems not to matter to our problem at hand. However, if you cover the rear plate, seal any major opening between the MB zone and the HDD zone, and leave the main fan plate operational, are you not effectively dealing with the static pressure of an enclosed space?

 

From there, I wanted to refine the model to reflect the fact that our case does have vents in the back for air to escape. My next logical question became: Without any fans in the rear of the case, how much air would exit the back vents due solely to the pressure differential between the MB zone and the outside air.  With this amount quantified, I could then say the following:

 

When using the main fan plate with 4 fans rated at 34.5mm H₂O and 80CFM each and MB zone pressure equaling the fan static pressure: the amount of air exiting the rear of the case would only be the 25CFM that is due to the differential in the pressure of the MB area and the outside air.  The main fan plate could only replenish 25CFM regardless of the fans rating.

 

Based on this knowledge, I easily see why you state that fans in the rear of the case will always be beneficial and why a rear mounted fan rated at 50CFM will actually exceed its official rating.  So, while the concepts did not seem applicable to you, they were being used as a foundation for me to continually refine my model into one that more closely resembles the actual situation.  This is why I wanted to know if you thought my application of these concepts were accurate or if the results I was producing were somehow flawed.  I've yet to learn about and apply many more complex concepts that are applicable to the problem at hand and have not had a chance to fully integrate information contained in the presentation you linked to earlier.

 

From what I have been able to integrate so far, I do have a question:

In a 2 dimensional overview of the case, from the fan plate to the rear vents, is the static pressure in the motherboard zone constant at all points or does the static pressure decrease along a gradient (linear, expontential, etc.) as you move away from the fan plate and towards the rear vents?

 

To me, it would seem that there is some form of gradient that would be affected by the vents in the back and the static ratings of the fans in the front.  Is this a correct assumption or is it another swing-and-a-miss?

[bubbaQ]

In a 2 dimensional overview of the case, from the fan plate to the rear vents, is the static pressure in the motherboard zone constant at all points or does the static pressure decrease along a gradient (linear, expontential, etc.) as you move away from the fan plate and towards the rear vents?

 

Yes there is *some* miniscule gradient, but you ignore it unless the transit time is short... i.e. the ratio of flow rate to volume and shape of the cavity.

 

[PhatalOne]

BubbaQ,

 

I am going to be working on another 4220 build next week and need some advice.  Assuming zone 1 (HDD area) and zone 2 (MB area) are completely sealed except for the fans, what effect does zone 1 (assume it's full) have on the fan inflows on the stock plate and do you think there would there be an ideal fan curve for this situation? 

 

As I've worked more with my 140mm plate, I have come to the conclusion that the current options I found at 140mm are not particularly suitable for the dynamics present in the case.  Therefore, I'm looking into a variety of "creative" cooling options for a revised build and would like your initial thoughts on the following ideas:

 

1.) Create a fan plate that has 1 120mm fan that has outflow into the motherboard area, and 2 120mm fans with outflows that are immediately ducted to cutouts/exhaust fans in either one or both sides of the case.

 

2.) Same basic concept as above, except, that the 2 exhaust fans are mounted on the side of the case with their inflows drawing air through a duct to zone 1 with a 3rd 120mm fan that creates airflow into the motherboard area.

 

I am also looking at options like multiple 60mm or 70mm fans in parallel in place of a larger fan for better static pressure. Do you think this is something worth investigating?

 

Thank you in advance for your time and knowledge, it is much appreciated.

[bubbaQ]

You are going about it all wrong.

 

If you want to do it right:

 

1) The purpose of the fans is to cool the equipment.... mostly the drives since the CPU will take care of itself later.

2) There is a performance curve for each configuration of occupied/vacant drive bays (as well as species of drive, since some drives are much hotter than others).

 

So TEST it.  Drive temp is the dependant design variable so plot drive temp vs. something.

 

You need to build a half-lid, that covers the front chamber, and leaves the back open (free air).  Then put some ass-kicking fans in the fan plate, on a variable-speed fan controller.  Add a cardboard shroud of about 20" to the front of the case.  Get a good anemometer, to measure air flow across the front of the shroud

 

Then you test with several dozen configurations.  For each config, plot drive temps vs. air flowrate.  Do one plot with drives idle, and another doing a parity check with all active.

 

From that, you will be able to determine what flowrate (X) of air you will accept as a minimum, based on what max drive temp you will accept.  You may also decide on certain configs of drives that you will or will not use because of how they impact temps and airflow.

 

If your fans are too loud, with free air, then you need some different fans, or you need to change your expectations, or tolerances for drive temps.  You can try differnet mfg's fans, but there isn't a lot of difference.

 

Then  size the rear fans to handle the same flowrate in free (zero pressure) air as X.  This should give you approximate neutral pressure in the rear chamber -- and that is the optimum for both sound and economics.

 

[PhatalOne]

Bubba,

 

I had a chance to experiment with my both of my 4220 builds the past week (one stock, one modified). Although I sealed all of the HDD trays and fan plate openings on both with foam inserts, I found that the stock was much noisier than my the 140mm fan plate and the temps were not all that different.  The biggest improvement in temps I saw was going from a 7200rpm drive to a 5400rpm drive for quite obvious reasons.  I'm at the point where I think your original suggestion is the least work and would provide the most benefit.  Since my case is not in the middle of a rack, I'm thinking that having a mesh/grille section in the case lid over the MB area to help reduce the static pressure would be easiest solution.

[bubbaQ]

Don't just put mesh... mount a 200mm exhaust fan on the lid.  Put a speed controller on it, and dial it to the audible level you want (but make sure you don't pull negative pressure inside the case.  That's easy to test with a punk or cigarette next to a case vent.

[nate1749]

How are people mounting these?  I bought this mount rack  http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&item=270421343063

but the handles on the norco case are in the way of the screw holes for the rack mount (so it can't be mounted).  Did I buy the wrong type of rack mount or is it normal to remove the handles?

 

Nate

[jk2587]

nate1749,

 

The rack you bought is not correct. It is intended mainly for networking gear and is only 18" deep, while the Norco case is 25.5" deep. You need a much deeper case, and I can't say for sure as I don't have mine racked but I think you might need the rails too.

[NAS]

The rails require back rail attachments and are flimsy and a pig to fit at that.

[PhatalOne]

Update on my experiences with the 4220 with my 140mm fan plate.  After having taken a few months off and tinkering with the server occasionally, I've found the best results using the following combination of hardware.

 

Added 2 rear mount 120mm exhaust fans with all rear exhaust fans controlled by a manual rheostat:

http://www.svc.com/y720dcd-25t1-gp.html

5900 or 5400 RPM drives... the extra few degrees C helps significantly

 

I focused on doing something relatively mundane so that others could replicate what I did.  My array is now up to 9 drives with the highest temps on my 7200rpm drives being about 37 or so during parity checks (fans are not turned up all the way either).  The new 5900rpm 1.5TB Seagate drives run much cooler than the existing 7200rpm Seagate 1.5TB drives. Eventually I'd like to swap them out, but the operating temperatures are now acceptable.

 

I'll post some updated pics once I get my new drive controllers installed and working.

[smino]

Of course one option is to dump the Norco case and use someone else's case that has been proven in a production business  environment:

Clean power and sata cable lines. No drive cages to trap heat. Just drop your hard drive on the sata backplane and go.

Large fans to cool the drives. They have been running with this setup for a while, so it is well proven. Now they use software raid, instead of unraid.

 

http://blog.backblaze.com/2009/09/01/petabytes-on-a-budget-how-to-build-cheap-cloud-storage/

 

45 Drives of 1.5 TB for 8-9K. Now you can run unraid on this if the Port multipliers are supported in unraid.

What is the maximum number of drives support in unraid now? Is there a link to all the stats which get updated?

 

Now the case above needs to have a better power supply. Why they did not chose a redundant power supply setup, is not clear.

I also think they could have done better than use a pci card in the mix.

 

[WeeboTech]

They have been running with this setup for a while, so it is

well proven. Now they use software raid, instead of unraid.

 

I didn't see anywhere they were using unRAID. Besides that would be a big performance hit in itself.

 

 

45 Drives of 1.5 TB for 8-9K. Now you can run unraid on this if the Port multipliers are supported in unraid.

What is the maximum number of drives support in unraid now? Is there a link to all the stats which get updated?

 

unRAID officially supports 16 data drives.

This was expanded to 20 drives in the beta version, however it does not work yet.

 

I also think they could have done better than use a pci card in the mix.

I was surprised they used PCI in the mix also.

[PhatalOne]

Of course one option is to dump the Norco case and use someone else's case that has been proven in a production business  environment:

Clean power and sata cable lines. No drive cages to trap heat. Just drop your hard drive on the sata backplane and go.

Large fans to cool the drives. They have been running with this setup for a while, so it is well proven. Now they use software raid, instead of unraid.

 

That is an interesting link but is off topic.

 

I chose the Norco considering the following:

1.) My home is not a "production business environment"

2.) This is my hobby, I like to tinker. Please see number #1

3.) This case is significantly outside the budget for this project... I could buy 2 norco's for this same price.

4.) unRAID does not support 45 drives, even if I wanted it to do so.

 

That said, you do know what unRAID is and its limitations, right? It seems like you're astroturfing.

[Oxidd]

Suggested Future Improvements

 

Cooling

 

6.) Blanking plates or foam inserts for empty slots to stop air being drawn through them as path of least resistance

 

 

 

Sorry to rez. this post, but I noticed in my 4220 that I got a few weeks ago, for each tray I can slide the little metal-grill to block the vent holes on empty trays.

So no need for blanking plates.

[Rajahal]

Sorry to rez. this post, but I noticed in my 4220 that I got a few weeks ago, for each tray I can slide the little metal-grill to block the vent holes on empty trays.

So no need for blanking plates.

 

I never would have noticed that if you didn't point it out.  Thank you!

[EMKO]

just wondering whats safe temps for hard drives? mine where at 18C-20C but slowly over few months they are going to mid 20's during parity check. Just want to know if they go any higher when is it bad?

[Joe L.]

just wondering whats safe temps for hard drives? mine where at 18C-20C but slowly over few months they are going to mid 20's during parity check. Just want to know if they go any higher when is it bad?

You are fine.

 

Under 40C is best.    Over 45C is bad... Many disks will shut themselves down at 55C as then consider it dangerous.

 

It sounds as if you have your server in a cool place and have great air-flow across the disks.

 

My server is in the basement, in heated/air-conditioned space.  The disk temps when idle are high teens to mid-20's.  When all spinning during a parity check they average mid 30's.  My highest disk temp is typically 37C.  I have my unMENU screen set to alert me if the temperature goes above 40C.

[EMKO]

yea i have 5400rpm hard drives plus i have it in the basement in a closet. The low temps are due to the low temps here i Canada but as summer comes its going to get hot here. Winter can get to -40C and summer can reach 30C+ so i hope this doesn't cause the heat to rise to much as i can see it already has risen a bit. I will have to see how hot they get this summer to find out if i will have to do some mods, hopefully they will stay below 40C.

[DocBlock]

I'm considering the Norco 4220 for a build, though initially only populated with perhaps 10 disks. I'm a little unclear about the fans that come with this unit, though. Are they undersized for 20 disks? Or is it just that they are too noisy?

 

I have a data closet in my basement that is consistently around 19C, and noise is not an issue. I wonder if I'd be better off paying for a better designed chassis, as I'm not dying to do metalwork. I have two 19" two-post racks in the closet, and I can support the unit with shelves if necessary, but I definitely prefer a rackmount case to a tower.

[Rajahal]

I haven't filled all 20 bays on my 4220, so I can't really say if the fans are sufficient or not.  It probably depends on the environment quite a bit as well.  If you are able to keep your server closet at 19 C year round, then I expect the 4220's included fans are sufficient.  Also be sure to close off the unused bays so that air doesn't flow through them (there are small sliding latches on the inside of each bay to close them off).

 

I can verify that the 4220's fans are loud.  Really loud.

 

There are a lot of custom backplane designs floating around, but check out the latest from Senjo:

 

Custom 4220 Backplane design by Senjo

[electron286]

This looks like a good thread for a question...

 

  I have a RPC-4116, with the SATA backplanes, not SAS.  Has anyone else got one of the NORCO racks with SATA backplanes?  If so, does anyone else have the situation where the drive trays do not match up with the actual LEDs on the backplane?

 

  In my case;

 

  1.  The green drive power LED shows through the lower light pipe to the hard drive activity front bay indicator.

 

  2.  The blue drive activity LED does not show at all on the front!  It is lined up with the masked out middle light pipe.

 

  3.  The upper light pipe which lines up with the drive power front bay indicator, has NO LED behind it on the backplane.

 

  Does this mean there is some version number of bays and/or SATA backplanes that are correct, and match up?  Did I just get shipped a bogus combination of engineering change rev parts that do not match up?

 

[ratmice]

For my 4220 build, I did get a 120mm fan plate, which I have yet to install. The 2 80mm fans on the back of the case were the real culprits (in my case) for horrible noise. After switching them out I haven't really felt the need to make it more quiet.