Aruba MAS series SFP+ & POE+ switches sub-$100

Mine showed up. Packaging was.... meh. It was packed in two PC tablet boxes taped together.

The fans are really noisy while it's starting, and even for a few minutes after it claims it's fully booted, but they settle down eventually.

It showed up with ancient firmware: ArubaOS 7.1.3.2. It is old enough that you have to use IE in compatibility mode to actually access it.



The board is somewhat interesting, if covered in heatsinks. I suspect the same PCB is used for the PoE versions as the non-PoE versions (see all the unused pins in the power connector!).

I'd be curious to see the internals of one of the PoE variants.

 

Here's the innards of the S2500-48P. Main and management PCBs are the same. Didn't feel like removing the PoE module.

 

I finished the fan mod today using the 4-pin version of the Noctua fans that Vicious recommended. I used my sound meter to measure the sound before hand. FYI that from about 3" away, the fans run at ~70-75 dB when running in the lowest mode (I didn't measure full-power mode, but it's ear-splitting) . Even at the low 75 dB, it whines at a pretty high pitch, which was why I opted for the modification. Post modification, it's somewhere < 60 dB, which is as low as my sound meter measures. And probably more useful than the dB reading is to know that it really is silent with the new fans.

@ViciousXUSMC - one thing I noticed in your video is that it appears you reversed the fan direction when you put in the Noctuas. The part where you held the paper to the fans while running went by quickly, but it looks like you installed the new fans to pull cold air in, and the original fans blow the hot air out. Was this intentional or were your stock fans installed differently than mine? (I have the 48 port POE version)

My temperatures are definitely higher post installation (though my ambient was +5 C as well). With that said, the switch doesn't seem worried about them as the fans are still running at low even with the temp increase. This makes me think that Vicious' advice to get the 3-pins that run at high all the time was advice I should have taken. Anything to be worried about below? My worst temp is DPE, and the DPE Critical temp from Aruba is 110, so I'm still a ways off that.

As an aside... what is a DPE?

 

Data Path Element (DPE) traffic management chip
Which, I believe, is the controller for the L3 routing portion of the switch..

 

fancy name for the actual switching ASIC

 

I would not spook until 90c, I mean GPU/CPU get that temp all day long in some systems and Aruba said that its well over 100c that bad things happen.

I think I posted it, or maybe it was in my video but I think critical was actually higher than 110c I think it was 120c.

As for fan direction, that was probably a mistake I made I could have sworn the old fans where pointed inward so that is how I also configured the new ones. I do not think it will make a big difference either way unless the intake from the other side is more near one of those "hot spots" on the DPE resulting in a more direct airflow to them, but the board layout seems pretty symmetrical.

 

Thanks for the answer, I appreciate it. FYI for everyone else, the info re: Aruba temperature thresholds is below (same as in your video) (110 is the shutdown temp).

Also, if anyone else has an unmodified unit and can verify which way the fans blow, can you let the rest of us know please? It'd be good to know for certain.

Lastly, @ViciousXUSMC or anyone else who has taken one of these apart, if I wanted to leave one of the original fans in to hit "DPE-0," any idea where that is on the board and which fan I'd re-install? Not sure if I'll bother, but I might try it to see how bad the noise of 1 vs. 4 of those is if I can hit the right area of the board with the super-fan.

Temp info from Aruba:

Code:

=================
Name               Threshold
-----------------   ----------------

CPU Temp         95 *C
Mid Temp           85 *C
DPE Temp         110 *C

These thresholds values are same for all MAS platforms/models.

System shutdown is triggered when CPU or DPE temperature crosses 
the above mentioned threshold values. Before shutdown there will be 
‘syslog’ and ‘trap’ generated, along with ‘alarm’

Depending on the proximity of the current temperature to the threshold  
value:

Minor alarm:   If temperature lies between (threshold – 15)  
to  (threshold – 5)
Major Alarm:   If temperature lies between (threshold – 5) 
to  threshold.
Critical Alarm: If temperature is above respective threshold.  
Then triggers shutdown.

And when the switch shutdown happens, it will NOT resume/
auto-bootup post cooling. User has to manually cycle 
the AC power.

 

I just checked my 24 port switch and all four fans blow air out of the case.

 

Good to know, my temps are very reasonable so I will not bother changing it now that its already racked and setup.
If/when I start to use the PoE switch I'll keep that in mind, and also see if I can find good alternative fans that sit somewhere between the price/performance of the Noctua and the stock fans.

 

Been a lurker for a while but I wanted to make an account and say thank you to Vicious for the videos.

I'm sure this might have been covered by someone else, but another datapoint wouldn't hurt. This is an S2500-48P with the Noctua 3pin fans installed (without any low speed adapters) sitting in my home office with only one port plugged in. My Kill-a-watt was hovering around 64w if I remember right (I'm not home at the moment).

Code:

System Card Slot              : 0
SC Serial #                   : BZ0001307 (Date: 10/07/12)
SC Model Name                 : ArubaS2500-48P
Mgmt Port HW MAC Addr         : 00:1a:1e:0e:bc:81
HW MAC Addr                   : 00:1a:1e:0e:bc:80 to 00:1a:1e:0e:bc:bf
CPLD Version                  : Rev: 2.4 (0x2.4)
PoE Firmware Version          : 4.1.5 (Build: 1)
CPU Assembly #                : 2010110C (Rev: 03.B0)
CPU Serial #                  : MC39000685 (Date: 10/07/12)
Power Supply                  : Present (580W)
                              : 12V System Voltage Ok
                              : 56V PoE Voltage Ok
Power Supply Serial #         : IP058T8203FP (Date: 09/01/12)
Power Supply Model No         : 2510064
Power Supply Vendor Model No  : IPSN-580APA1 (Rev: 48.48)
Console Card Assembly #       : 2010111C (Rev: 01.00)
Console Card Serial #         : CC38000838 (Date: 10/07/12)
PoE Card Assembly #           : 2010106C (Rev: 01.00)
PoE Card Serial #             : PC38001439 (Date: 10/07/12)
System Temperature            : CPU Temp                : 47  C
                              : DPE-0 Temp              : 65  C
                              : DPE-1 Temp              : 54  C
                              : DDR Temp                : 40  C
                              : PSU Temp0               : 36  C
                              : PSU Temp1               : 41  C
                              : Exhaust Temp            : 42  C
                              : Mid Temp                : 46  C
                              : Ambient Temp            : 38  C
                              : Console Card Temp       : 27  C
System Voltages               : 3300mV Sense            : 3360.0 mV
                              : 2500mV Sense            : 2464.0 mV
                              : 1800mV Sense            : 1808.0 mV
                              : 1000mV Sense            : 1012.0 mV
                              : 1100mV MAC              : 1098.0 mV
                              : 1200mV Sense            : 1216.0 mV
                              : 5000mV 5VDD             : 4998.0 mV
                              : 5000mV 5VSB             : 4974.0 mV
                              : 900mV VTT               : 910.0 mV
System Fan Tachometers        : Power Supply Fan 0      : 5648  RPM (Low Speed)
                              : Power Supply Fan 1      : 5625  RPM (Low Speed)
                              : FAN 0                   : 5601  RPM (Low Speed)
                              : FAN 1                   : 5421  RPM (Low Speed)

For what it's worth, I was frustrated for about 20 minutes trying to figure out why I couldn't access it when it was connected to another switch (but would work when I plugged my laptop into it), turns out spanning tree was blocking the uplink port. D'oh!

I haven't updated the firmware yet, but the seller was kind enough to include a console cable in addition to rack ears and a power cord.

I'm looking to replace my top-of-server rack switch which is currently a Dell Powerconnect 5448 with this one, as well as upgrading my servers to 10GB at the same time. My network rack switch is an HP Procurve 2910-al-24G but I don't have the 10G module on that, and it's not worth the high cost on eBay to get one. I don't need 10G on my whole network anyway, just between the servers in the rack for now.

 

Just look at the pics fake-name posted at https://forums.servethehome.com/ind...poe-switches-sub-100.23056/page-5#post-217864. Clearly the back 2 fans are related to the PSUs and the front 2 are for both CPU and memory.

Computer fans are designed to blow towards where they're mounted. Again, among the pics from fake-name shows that portion rests against the chassis.

 

I can confirm all the fans blow outwards on the left-hand side.

Realistically, as long as all 4 are facing the same direction, it'll not likely make a major difference. Theoretically, you want the fans to blow over the most temperature sensitive components (generally electrolytic bypass capacitors) first, because if they run hot it can negatively affect their lifetime, but the power supply design isn't optimized for that (there are caps all over the place, not just on one side or the other), so it's not worth fretting about.

 

How about 2 intake on the cpu/memory side and the other 2 in exhaust mode on the psu side?

 

You would lose your positive/negative pressure and really kill the airflow.

 

Neutral pressure != zero airflow

 

No it does not, but it also depends a lot on the conditions.
If you put two fans in a closet, one at the top and one at the bottom of the door and had them opposite one another that would work fine (especially intake at the bottom as cool air goes down and hot air goes up)

But try the same thing in a door at the end of a hallway, unless you are right infront of the fan you wont get any air down the hall.

So you have two options. Put both fans in the same direction and open a door on the other end of the hallway that has access to outside air and the pressure will cause airflow through the entire hall. The other option have the fan on the opposite end of the hall running the reverse direction where you are more pressure neutral but the pressure differential between intake and exhaust will cause airflow.

These little fans in a big switch mounted on the same side are very much like having fans mounted on a single doorway for a big hall. Trust me it wont work the way you are expecting. What ever is right by the fans will be fine but the rest of the switch is going to get hot, and without air circulation those temperatures may continue to get higher and higher until an issue occurs.

I would much like you to try it yourself if you cant believe me and let us know.

 

Thanks for the heads up on these. Got my 48P yesterday and even though the seller claimed used it came brand new in box (but with a foreign power cable?). Really solid unit. Didn't like my 10Gbe DAC cable so I ordered some transceivers and LC cables. Found the compatibility chart here for ref -- https://www.arubanetworks.com/support-services/interoperability/#SFP-XFP . I was concerned about the fan noise from the ongoing discussion, but I found that after boot it cranked down to a very reasonable level. A lot quieter than one of my Hp Gen8 servers.

 

I also want to verify that the Finisar modules (FTLX8571D3BNL-E5) referenced by ViciousXUSMC work and connect to Mellanox connectx3 (MCX312A-XCBT i think) and the Chelsio T440 nics (https://forums.servethehome.com/index.php?threads/10x-chelsio-t440-quad-port-10g-sfp.22337/). Havent done any "real" testing with throughput but at least they were recognized and I could transfer files, etc.

 

ConnectX-3 with the Finistar modules give me 9gb/s with 1500MTU and almost the full 10gb/s with 9000MTU between my VM's and Workstation.

 

2 fans in and 2 out will make for excellent airflow; it's referred to as 'balanced draft'. Just because the internal pressure is held near a zero delta to the outside doesn't mean there's no air flow...

Power generating station scenario: petroleum coke fired, top-supported watertube steam generators producing superheated steam which is expanded through steam turbines to generate power. Process control software holds the furnace pressure to -0.5", while twin, steam-turbine driven forced air fans pressurize the unit at the burner deck and a slip stream gets directed to secondary air fans then to ball mills where it mixes with/ carries pulverized coke to the burners. At the exhaust plenum, a single induced-draft fans take suction from the furnace and delivers it to the electrostatic precipitators for ash removal then to the stack. All fans have adjustable inlet vanes to vary airflow.

The furnace is held to a slightly negative pressure so that air infiltration is inward, not outward. Internal restrictions to air flow include a feedwater pre-heater (economizer) and both primary and secondary superheaters (steam must be heated beyond its saturation temperature to prevent condensation on blades as the steam is expanded through turbines. Condensation causes pitting and can lead to deposits on the blades resulting in unbalanced conditions and power loss. Blade washing is a thing.)

This scenario is common to all fired boilers, whether they have just forced air fans or both forced & induced air systems, due to a condition known as 'complete combustion'. Target combustion is a slight excess O2 and zero CO in the exhaust: too much excess O2 and you're wasting fuel by heating too much air and sending it up the stack; too little O2 and the fuel is partially un-burnt and creating CO as a product of incomplete combustion. CO is still capable of being burned, and in large enough % is explosive, and can lead to a 'furnace pressure excursion'. Redundant instrumentation at the stack continuously measures exhaust O2, CO2, CO, NOx, SO2 and particulate matter.

Zero pressure differential does not necessarily indicate zero air flow.

I worked 16 years in this environment; 10 years on the floor and 6 in the control room. 450MW generating station attached to a heavy oil upgrader.