APOLLO Build Log - alpenwasser.net
Contents
1. Figure 2 83 Round 2 so far so good Figure 2 84 Hail Hydra 16 Figure 2 86 Different Angle p Jl S E m Figure 2 87 The disks are distributed in the cages for improved airflow and easier cable management i p L v Figure 2 88 Closeup of the connectors All in all though I still wasn t quite happy mainly because 2 8 STORAGE TOPOLOGY AND CABLING 17 Figure 2 89 of this Figure 2 93 and the M B side Much better in my Third time s the charm This time I hid the extra humble opinion cable length behind the triple fan unit y E P v Figure 2 94 The bottom controller had a bit more extra cable length to hide so that part is a bit messier ri Im Figure 2 92 And the cable loom for controller 0 from edis Figure 2 96 Much nicer than before methinks 2 8 STORAGE TOPOLOGY AND CABLING 18 f 4 Aa J a a A ut EO E 0 i y NN el Y 4 d i p Figure 2 102 And the controller LEDs when there s some activity While not completely flawless I am rather happy with the result now Ve AS Ta Li1 32 2 2 P e gt gt l Figure 2 99 And the other side Significant Improve ment I d say Figure 2 100 The cable looms going off the controllers 2 9 STORAGE AND NETWORKING PERFORMANCE 19 2 9 Storage and Networking Performance April 26 2014 Available online at 39 402. A P L Ij FILE SERVER BY ALPENWASSER dem Preface Abstract This document describes in chronological fashion the building process for our home server dubbed APOLLO short for Anonymity Protecting Oblivious Long Legged Ostriches and no that does not need to make sense Most of the content is comprised of transcripts sometimes slightly edited from the build logs posted on linustechtips com themodzoo com overclock3d net and overclock net see 1 2 3 4 which is also why the posts are mostly written in present tense This is not a report on the project written in retrospect but a faithful though not literal transcript of the projet as it developed Thanks George for lending his skill time dedication and machinery The communities on linustechtips com themodzoo com overclock3d net and overclock net for their great support and feedback Revision Current document revision number 1 0 Switzerland November 2014 Contents 1 Prologue 1 1 Hardware 1 2 Purpose 1 3 Modding 2 Chronolical Transcript of Building Process 2 1 First Hardware Tests amp Noctua Coolers a0 ira di OR ee d Ee E ES 22 Temporary eun 05 installed 3 5 2 33 5 2 2 4 08 unu SEEDS SY See GSS 2 AS 2 3 PSU Mounting amp LSI Controller Testing a a 2UE Ane Dus Reus rd Se ee e ee dd ae ae a Ba Chipset COOLE S ue yes ba NU Ae te NADA woe ee ee ete deinde dre Sek y 2 0 DIS Ventilatii ubi
3. 2014 REFERENCES 19 20 21 22 23 24 25 27 28 29 2014 Apr The Disk Racks overclock3d net Online Avail able http forum overclock3d net showthread php p 742203 post 742203 cessed Nov 22 2014 Ac 2014 Apr The Disk Racks over clock net Online Available http www overclock net t 1442386 post__22045804 Ac cessed Nov 22 2014 2014 Apr Chipset Cooling amp Ad ventures in Instability linustechtips com On line Available http linustechtips com main index php act findpost amp pid 1823902 Ac cessed Nov 22 2014 2014 Apr Chipset Cooling amp Adven tures in Instability themodzoo com Online Available http themodzoo com forum index php act findpost amp pid 25297 Accessed Nov 22 2014 2014 Apr Chipset Cooling amp Ad ventures in Instability overclock3d net On line Available http forum overclock3d net showthread php p 743654 post743654 Ac cessed Nov 22 2014 2014 Apr Chipset Cooling amp Ad ventures in Instability overclock net On line Available http www overclock net t 1442386 7 post 22079815 Accessed Nov 22 2014 2014 Apr Disk Ventila tion linustechtips com Online Avail able http linustechtips com main index php act findpost amp pid 1829673 Accessed N
4. 41 42 An update storage topology diagram for this stage of the build can be found in appendix A 2 Beware This section will be of little interest to those who are primarily in it for the physical side of building Instead this update will be about the performance and software side of things So lots of text lots of numbers These results are still somewhat preliminary since I m not yet 100the hardware config will remain like this for an extended period of time I really want to put another 12 GB of RAM in there for example and am considering adding some SSD goodness to my ZFS pools nor am I necessarily done with tuning software parameters but it should give some idea of what performance I m currently getting As you may recall from my previous update I m running three VMs on this machine two of which are pretty much always on the media VM and my personal VM and the third of which is only active when I m pulling a backup of my dad s work machine apollo business The storage setup has been revised somewhat since the last update I now have a mirrored ZFS pool in ZEUS for backing up my dad s business data so in total his data is on six HDDs including the one in his work machine His data is pulled onto the apollo business VM from his work machine and then pulled onto ZEUS The fact that neither the business VM nor ZEUS are online 24 7 ZEUS is turned off phys ically most of the time should provide some
5. After having done that I turned my attention to the side panel making an opening for the ventilation I thought of several ways of doing this but all of them were a bit more complicated than I d have liked them to be Cutting such a big hole with a dremel isn t really practical so I considered doing it with our jigsaw but after doing a few test cuts I didn t really like the result as I couldn t get a straight enough cut And the cut needed to be clean because there s no space to fit a U channel over the edge and I don t really like the idea of covering it up on the outside Anyway the guy just used a nice big angle grinder for the cut and since he s a metal worker by trade it turned out almost perfectly straight not 100but it s still cut by hand after all After that I painted Figure 2 63 The mesh from the inside a B w PIN A x M I NN A WAS I t x E e Yy j i 2 7 FAN UNIT 13 As hinted at earlier the airflow in this build will EHE TTT go from the front compartment through the middle HUTT TTT TTT 4 wall into the M B compartment and out the back MILL 3 This is pretty much how the stock configuration HERE 3 i l i AURORA cM works except in that the air gets in through the front panel not through the side panel Unfortunately I forgot to take pics of the stock con fig but luckily tweaktown com did a review on this iif case and took them for me see 33 L
6. OS Installed overclock3d net On line Available http forum overclock3d net showthread php p 701635472post 701635 Ac cessed Nov 22 2014 12 2013 Nov Temporary Ghetto Setup for Testing OS Installed overclock net On line Available http www overclock net t 1442386 4 post 21202531 Accessed Nov 22 2014 2014 Apr PSU Mounting amp LSI Controller Testing linustechtips com On line Available http linustechtips com main index php act findpost amp pid 1789104 Ac cessed Nov 22 2014 2014 Apr PSU Mounting amp LSI Controller Testing themodzoo com Online Available http themodzoo com forum index php act findpost amp pid 25037 Accessed Nov 22 2014 2014 Apr PSU Mounting amp LSI Controller Testing overclock3d net On line Available http forum overclock3d net showthread php p 742061 post742061 Ac cessed Nov 22 2014 2014 Apr PSU Mounting amp LSI Controller Testing overclock net On line Available http www overclock net t 1442386 post__22042818 Accessed Nov 22 2014 2014 Apr The Disk Racks linustechtips com Online Avail able http linustechtips com main index php act findpost amp pid 1792154 Accessed Nov 22 2014 2014 Apr The Disk Racks themodzoo com Online Avail able http themodzoo com forum index php act findpost zpid 25055 Accessed Nov 22
7. but I hope it s somewhat understandable Before doing storage benchmarks across the net work we should of course get a baseline for both the storage setup itself as well as the network The base pipe from dev zero into dev null transfers has a transfer speed of 9 GiB s Noth ing unexpected but it s a quick test to do and I was curious about this 2 9 1 Networking Baseline Perfor mance The networking performance is where I hit my first hickup I failed to specify to the VM which network ing driver it was supposed to use and the default one does not exactly have stellar performance It was an easy fix though and with the new settings I now get pretty much the same networking performance across all my machines except the Windows ones those are stuck at 500 Mbit s t for some reason as you can see above but that s not hugely important to me at the moment TBH When using the default QEMU networking card I got about 520 Mbit s t typical transfer speed when switching to the virtio driver it increased to roughly 930 Mbit s typically PARTNER PARTNER MACHINE O S BANDWIDTH Office PC Windows 7 513 0 Mbit s HELIOS Windows 7 517 0 Mbit s ZEUS Arch Linux 931 0 Mbit s Dell XT2 WiFi APOLLO Host APOLLO VM 2 APOLLO VM 3 S S S Arch Linux 12 7 Mbit s Arch Linux 939 0 Mbit s Arch Linux 885 0 Mbit s Arch Linux 858 0 Mbit s Table 2 2 Example from one run with iperf not av erages Measured fro
8. www overclock net t 1442386 post__21185159 Ac cessed Nov 22 2014 2013 Nov First Hard ware Tests amp The Noctua NH UD9X 1366 linustechtips com Online Avail able http linustechtips com main index php act findpost amp pid 1043570 Accessed Nov 22 2014 16 2013 Nov First Hardware Tests amp The Noctua NH UD9X 1366 themodzoo com On line Available http themodzoo com forum index php act findpost amp pid 19676 Accessed Nov 22 2014 17 2013 Nov First Hardware Tests amp The Noctua NH UD9X 1366 overclock3d net On line Available http forum overclock3d net showthread php p 701017 post701017 Ac cessed Nov 22 2014 2013 Nov First Hardware Tests amp The Noctua NH UD9X 1366 overclock net Online Available http www overclock net t 1442386 post__21188827 Accessed Nov 22 2014 2013 Nov Temporary Ghetto Setup OS Installed linustechtips com On line Available http linustechtips com main index php act findpost amp pid 1053907 Ac cessed Nov 22 2014 33 11 13 H4 H5 16 iia 18 10 2013 Nov Temporary Ghetto Setup OS Installed themodzoo com Online Avail able http themodzoo com forum index php act findpost amp pid 19721 Accessed Nov 22 2014 2013 Nov Temporary Ghetto Setup
9. 0 000 ee eee 17 Cable Management Third Try Cable Loom e 17 Cable Management Third Try Cleaning Up enn 17 Cable Management Third Try Cleaning Up enn 17 Cable Management Final Result 2 oaa a a a a 17 Cable Management Final Result 2 oaa a a a 18 Cable Management Final Result a a a ee 18 Cable Management Final Result 24 4 2 64 a a a A A a 18 Cable Management Final Result ee 18 Cable Management Final Result lt 2 oaoa a a a a a a a ee ee XR 18 Cable Management Final Result a a a 18 Storage Benchmarks CPU Load w Compression Disabled 20 Storage Benchmarks Write Speed w Compression Disabled 20 Storage Benchmarks CPU Load w Compression Enabled 0 20 Sound Dampening Front Panel a 24 a ox Pa om emp eR xS BUR ARES 21 2 5 C aselabs Drive Gace uas xeu uox Re cx Wc NN RS PA eS ee 21 Emal Peones Coad ect i CAR By Sy STR BURN Eod Ae ete RES E S 22 Final Pictures Open CPU Side 4235205523433 44 Coe Ed e erbe pos G 22 LIST OF FIGURES 32 3 9 3 6 3 1 3 8 3 9 3 10 3 11 3 12 3 13 3 14 A l A 2 A 3 Pinal Pictures Fon Galore Au ur bd m8 m e T Benet Rede es Be ee AS 22 Final Pictures Oa DIO a van ae a ds eS eS ek we a 23 Pinal Pietures Cabling Lacu ack d ie dat ds RA A dede o A HY 23 Pinal Pictes S Cate nd os duce Deren Eo E
10. IW PP689 Pedestal Performance Server Chassis tweaktown com Online Available http www tweaktown com reviews 3769 in win iw pp689 pedestal performance _ server chassis index html Accessed Nov 23 2014 alpenwasser 2014 Apr Storage Topol ogy amp Cabling linustechtips com On line Available http linustechtips com main index php act findpost zpid 1853407 Ac cessed Nov 22 2014 2014 Apr Storage Topology amp Cabling themodzoo com Online Avail able http themodzoo com forum index php act findpost amp pid 25412 Accessed Nov 22 2014 2014 Apr Storage Topol ogy amp Cabling overclock3d net On line Available http forum overclock3d net showthread php p 744591 post744591 Ac cessed Nov 22 2014 Apr Storage Topology amp Cabling overclock net Online Avail able http www overclock net t 1442386 post_ 22101845 Accessed Nov 22 2014 2014 wpirobotbuilder 2013 Nov Reducing Sin gle Points of Failure SPoF in Redundant Storage linustechtips com Online Avail able http linustechtips com main index php act findpost amp pid 993585 Accessed Nov 23 2014 alpenwasser 2014 Apr Storage and Net working Performance linustechtips com On line Available http linustechtips com main index php act findpost amp pid 1934179 Ac cessed Nov 22 2014 2014 Apr Sto
11. decent protection against most malheurs the only thing I still need to implement is a proper off site backup plan which I will definitely do in case of unforeseen disasters break ins theft and so on For convenience s sake I was planning on using NFS for sharing data between the server and its var ious clients on our network Unfortunately I was getting some rather disappointing benchmarking re sults initially with only 60 MiB s to 70 MiB s transfer speeds between machines I m not really a storage benchmarking expert and at the moment I definitely don t have the time to be come one so for benchmarking my storage I ve used dd for the time being It s easy to use and is pretty much standard for every Linux install I thought about using other storage benchmarks like Bonnie and FIO and at some point I might still do that but for the time being dd will suffice for my purposes For those not familiar with this dev zero ba sically serves as a data source for lots of zeroes dev null is a sink into which you can write data without it being written to disk So if you want to do writing benchmarks to your storage you can grab data from dev zero without needing to worry about a bottleneck on your data source side and dev null is the equivalent when you wish to do reading bench marks To demonstrate this I did a quick test below directly from dev zero into dev null Basically It s a bit of a simplification
12. reaquired But right before ZEUS was done I found out that ZFS was production ready on Linux having kept a bit of an eye on it since fall 2012 when I dabbled in FreeBSD and ZFS for the first time Using FreeBSD on the server was not an option though since I was nowhere near proficient enough with it to use it for something that important so it had to be Linux that s why I didn t originally plan on ZFS So I deployed ZFS on ZEUS and it s been work ing very nicely so far However that brought with it two major drawbacks Firstly I was now missing 5 TB of space since I had been tempted by ZFS to use those for redundancy even for our media files Sec ondly and more importantly ZEUS is not an ECC memory capable system The reason this might be a problem is that when ZFS verifies the data on the disks a corrupted bit in your RAM could cause a dis crepancy between the data in memory and the data on disk in which case ZFS would correct the data on your disk therefore corrupting it This is not ex actly optimal IMO How severe the consequences of this would be in practice is an ongoing debate in var ious ZFS threads I ve read Optimists estimate that it would merely corrupt the file s with the concerned corrupt bit s pessimists are afraid it might corrupt your entire pool The main focus of this machine will be e room to install more disks over time e ECC RAM capable e not ridiculously expensive e low mainte
13. well Figure 2 30 Side panels for disk racks completed Using some old broken HDDs to test fit e Figure 2 34 Note the cutout for the bolt head to go through when sliding the drive in Obviously 24 HDDs are going to put out some heat so some ventilation is required I m using six Papst fans for that The fans will be bolted onto the panels with some L profiles Because the fans have a hole spacing of 105 mm and the HDDs are roughly 100 mm wide each fan Figure 2 31 And a different angle can only be mounted on one side with two screws 2 4 THE DISK RACKS Figure 2 35 Taking measurements for the fan mounting holes Figure 2 36 and all three and with Since the fan screws need to be tightened rather heavily the screws exert significant pressure on the Figure 2 38 fan frames To reduce the stress on the fan frames we made some brass bushings that take the brunt of the pressure The panels are mounted to the bottom and top of the case with screws To have some leeway in adjust ing things there are slots instead of round holes in some places Figure 2 39 One of the brackets with slots for adjust ment Figure 2 41 Mounted bottom side The Disks just slide into the slots and lock into place in the recesses you can see above Since I can t tighten the screws I m using Loctite to prevent them from falling out due to vibration I tried to get some s
14. 5 3 1 3 2 3 3 3 4 Chipset Coolme ham Mounting s a 2 24 4544 64 ud a BR EE Be ee A 10 Chipset Coolie Fam Mountain a ERE p RES 8 EE 10 Chipset Cooling New Cooling Unit 0 0 0 ee 10 Chipset Cooling New Cooling Unit 2 3 24 Gee RUE xw Oe Bee eee ee Rew wo 8 11 Disk Racks pan Mount BUSRING Se a aa ede rn E ARAS A A 11 Disk Ventilation Vibration Dampening aa a a 11 Disk Ventilation Overview Shot asta aaa 11 Disk Ventilation Front Frame ses ceadan sod Sere A e ux web ow s 11 Disk Ventilation Marking Front Frame for Cutting 0 0 0 000042 2 12 Disk Ventilation Cutting the Inner Front Panel 12 Disk Ventilation New Inner Front Panel aoaaa a a 12 Disk Ventilation Side Opening evo a AE aa a 12 Disk Ventilation Side Opening Mounted aaa 0 a a a aa 12 Disk Ventilation Mesh s 4 8448 66 5 ba ae eee yxp Se Ad IVO lk eS Gee s 12 Disk Ventilation Complete e a dy de ure Soe se A A OS Sq x xus 13 Main Compartment Ventilation Stock Configuration 13 Main Compartment Ventilation Stock Configuration 13 Main Compartment Ventilation Versionl ao a a a a a 13 Main Compartment Ventilation Version 1 2 2 2 2 2 2 2 2 13 Main Compartment Ventilation Versionl ao a a a 13 Main Compartment Ventilation Adhesive Mount 14 Main Compartment Ventilation Adhesive Mount Issue a a a
15. E M RUN A Sepa ternera alios At af ae a s 29 Final Pictures Loonr Fanning Out 24x eu me WW eS eS wie C E dr ex 3 24 Final Pictures Chipset Cooler and Cables e 24 Final Pictures CPU Coolers 4 4 44 43 4 8 4 6 8 ew ome 09 SR Tere wA USE Se Rd d 24 Pinel Pictures SAV tio le tee oe re UE a oe RR TO eg ee ee ERE we a e E da ond 25 Final Pictures Back Side Overview 2 2 2 2 2 2 2 2 25 242 252 2 5 25 Final Pictures Motherboard Compartment 2 2 2 2 2 2 2 2 2 29 Storage Topology Dara 4664 x ek vee eh eee ES EUN E XX SOR RR A ee Be ES 27 Storage Topology Diagram Revised a a a a a a a 28 Storage and Networking Benchmarks Results a a a a 29 1 alpenwasser J 3 4 5 8 9 References 2013 Nov APOLLO Prologue linustechtips com Online Avail able http linustechtips com main index php act findpost amp pid 1040187 Accessed Nov 22 2014 2013 Nov APOLLO Pro logue themodzoo com Online Avail able http themodzoo com forum index php act findpost amp pid 19652 Accessed Nov 22 2014 2013 Nov APOLLO Pro logue overclock3d net Online Avail able http forum overclock3d net showpost php p 700834 amp postcount 1 Accessed Nov 22 2014 2013 Nov APOLLO Prologue overclock net Online Available http
16. EATHER ra In the stock config the 92 mm fans are mounted HVT ENT TTT TTT inside some plastic fan cages that allow quick and HH toolless fan replacement in case of failure Originally I simply bolted the fans to two alu minium L profile bars Figure 2 64 And the end result Drive temperatures hover between 28 C and 35 C at the moment ambient is about 23 C 2 7 Fan Unit for Main Com partment Ventilation April 11 2014 Available online at 29 30 31 32 Figure 2 65 Stock configuration The blue cages can be easily swapped out if needed Credit tweaktown com 33 Figure 2 69 Intake side It was fixed to the middle wall with double sided Figure 2 66 as seen here Credit tweaktown com adhesive tape It s very strong stuff so the fan unit 33 falling off was not a concern Additionally the tape 2 7 FAN UNIT has some thickness to it which should provide some dampening between the fan unit and the middle wall mS XE T et TN gt Figure 2 70 Just a few strips of double sided adhesive tape Unfortunately due to some bumps on the middle wall getting in the way the tape on the rear angle didn t make proper contact with the wall It held but not very well Additinoally I noticed that there were rather strong vibtrations on the middle wall It turned out that the tape did indeed offer some decoupling but it also did not offer any addition
17. EUR RP ve S T Disko Racka Ventilation ee ss e he Ew ae ee es Sree Wie eo oe eS 8 Disk Tacks Vert lation ad e a ak wl o was Me he cates Ge pos Sue Se eile ae ee Med Se oe d 8 Disk Racks Ventilatio uius ue acuario Rope de aoee eon dee e De che hee eS ow race 8 Disk Racks Fan Mount Bushings 222 es 8 Disk Racks lt MOUnTING Bra ckets cris svs acs o m eT ee ere a UM a es Ae a e tA ACE d 8 Disk Racks Mounting Brackets amp 4 5 0 a ok hs we aso REO A on des es amp SR a 8 Disk Racks Mounting Bracket wu 44 4 Ad Stat anak add wee bee a oS 8 Disk Racks HDD Mounties 3 x 4 oe x Ye Iu Rae E XE died ed Be eoe RE 9 Disks RaGkers MOUDUOG e doh st oh rue Sac oe Ak IR PUE d et A A a to a a aad d 9 Chipset Cooling Improvising uut x dao sE Rede XU uu Re eo ae ss 9 Chipset Cooling Stock Heatsink 2 2 a 9 Chipset Cooling Naked Chipset ee 10 Chipset Colmes t LOW ss wh rd la oput A ee ee ee tds Be id 10 Chipset oolimg oc WS risas qom ECKE m ur SRR RMS db hee ws Se ee eR BAe ERES S 10 Chipset Cooling Fan Mounting usina GE eS OES BS 10 LIST OF FIGURES 31 2 90 2 91 2 92 2 99 2 04 2 99 2 90 2 97 2 98 2 99 2 60 2 61 2 62 2 63 2 64 2 65 2 66 2 67 2 68 2 69 2 70 2 1 22 2 13 2 14 2 75 2 76 2b 2 78 2 79 2 80 2 91 2 92 2 83 2 84 2 89 2 86 2 04 2 88 2 89 2 90 2 91 2 027 2 93 2 94 2 95 2 96 2 97 2 98 2 99 2 100 2 101 2 102 2 103 2 104 2 10
18. Some fins are bent from the drilling Figure 2 51 And the complete chipset cooling package Figure 2 52 For mounting it on the M B it needs to be mounted first without the fan 2 6 DISK VENTILATION y wez y Let u CEU m EZ T ls Me ADM Figure 2 53 then the fan is mounted on top and done 2 6 Disk Ventilation April 09 2014 Available online at 25 26 27 28 Although disks have become quite frugal when it comes to power consumption these days at least some of them and HDD cooling is not really a huge issue for most people packing 24 disks as closely to gether as in this build will cause heat issues without ventilation There is no need for 3k rpm Delta fans though a whiff of cool air breezing over the disks should do the job nicely For this purpose as you may have seen in some pre vious pics I have chosen 6 120 mm Papst fans specif ically the 4412 GLL model and am running them at 7 V The fans draw air in through a vent area and it then gets passed through the M B compartment and out the back Each fan is fixed to a rail riveted to one of the disk rack panels with two screws You ve seen this before but for completeness sake I m adding the pics of the bushings used to prevent the fan frames from being crushed to this update as well ES Figure 2 54 Closeup of the brass bushings Works nicely I ve exchanged the copper screws for some silver ones and in t
19. WWOTTBWONWUWWOWWOUW Yeah Figure 2 11 Ghetto Testing Setup Hardware Validation I ve put the system to gether temporarily to validate the M B CPU and memory so far all seems good A minimal Arch Linux setup has been installed and is successfully running BOINC at the moment I m not running BOINC as a hardware validation tool that s not what it s designed to do I have mostly validated the hardware and am now just running BOINC Gotta love low power CPUs core temps after about an hour of running BOINC on all cores are OPUO 31 C 31 C 35 C 30 C CPU1 32 C 26 C 29 C 31 C Table 2 1 Core temperatures during BOINC testing 2 3 PSU Mounting amp LSI Con troller Testing April 01 2014 Available online at 13 14 15 16 Yeah it s taking a lot longer to finish this than I d initially hoped doesn t it always with these sort of projects But I ve been working on it in the last few weeks and now finally have something to share The PSU slides into this case through an opening from behind and since the case isn t really made for normal ATX sized PSUs but server PSUs instead it s a rather tight fit To be more specific The PSU in its stock config does not fit the screws for the fan erill and the fan grill itself bump up against the case An easy fix though just needed to remove the fan erill on the PSU Figure 2 13 Good Fit Furthermore since normal server PSU
20. a a 14 Main Compartment Ventilation Version 2 2 ss 14 Main Compartment Ventilation Version 2 Fans Mounted Wrong Way 14 Main Compartment Ventilation Version 2 Fans Mounted Correctly 14 Main Compartment Ventilation Dampening Foam 2 ee eee 14 Main Compartment Ventilation New Mount 2 2 2 2 2 2 2 5 15 Main Compartment Ventilation New Mount 2 2 2 2 2 2 2 2 5 15 Cable Management Hat 15 Cable Management Disks lll e 15 Cable Manacemency First TEV s ei sve ats a ede dm eer a ee RUM desit Ae EE d 15 Cable Management First Try Result 2 0 0 ee 16 Cable Management First Try Result 2 0 e 16 Cable Management Second Try es 16 Cable Management Second DES s s soa ea bee a ee RR RR ok elo ae Ron ae ae d 16 Cable Management Second Try dai x Rue OA m RREGOEOEGUROE Wee OER Re eo ae 16 Cable Management Second Try Result 2 0 2 0 0 eee es 16 Cable Management Second Try Result 2 0 2 0 0 e 16 Cable Management Second Try Result 2 0 0 0 0 0 0 0 eee 16 Cable Management Second Try Issue ee 17 Cable Management Third Try Hiding the Mess 0 0004 17 Cable Management Third Try Hiding the Mess 0 02004 17 Cable Management Third Try Cable Loom 0
21. al strength to the mid dle wall i e no additional stiffness which meant the wall could easily vibrate Figure 2 71 Tape did not make proper contact So I took the unit out and while I was at it I also cut out some recesses for the fans which I didn t bother doing before I also put some dampening foam between the fans and the alu angles E E Figure 2 72 Cutouts for fans dampening foam 14 Aaand of course I mounted the fans the wrong way around at first Sigh Figure 2 73 If there s a wrong way to do things they will be done the wrong way at first Disassemble again reassemble X uo a Figure 2 74 Finally Figure 2 75 Also Foam between the alu angles and the wall itself This time I bolted it to the wall with some screws Much more solid now no more vibrations 2 8 STORAGE TOPOLOGY AND CABLING e p PA Figure 2 77 And from the other side 2 8 Storage Topology and Ca bling April 12 2014 Available online at 34 35 36 37 An overview diagram of the storage topology at this stage in the build can be found in appendix A 1 The idea behind the storage topology is based on wpirobotbuilder s post about reducing single points of failure see 38 Any one of the three LSI con trollers can fail and I still have all my data available You ll see below that I haven t yet gotten around to installing the Velociraptor I use coloured zip ties to mark t
22. co pesas ob Wwe pou pu qae He ds deg bdo qnd go aui DAMNIS DP PCM II ON 2 9 Storace Topology and Cablifig xc deu e hp o A ON Re E ER RUP Xd OX xu 2 0 Storage and Networking Performance voices a se Ewa ea wae VO x ee hh Y s 3 Result and Closing Thoughts Appendix A Diagrams List of Figures References ii 21 25 26 30 33 Chapter 1 Prologue Available online at 1 2 3 4 1 1 Hardware e CASE InWin PP689 e PSU Enermax Platimax 600 W e MB Supermicro X8DT3 LN4F e CPU 2 x Intel Xeon L5630 quadcore hyper threaded e HS Noctua NH U9DX Socket LGA1366 e RAM 24 GB Hynix DDR3 1333 MHz ECC e HBA CARDS 3 x LSI 9211 8i flashed to IT mode e SSD Intel 520 120 GB e HDD 0 WD VelociRaptor 150 GB 2 5 e HDD 1 3 Samsung HD103UJ 1 TB F1 x 3 e HDD 4 7 WD RE4 2 TB x 4 e HDD 8 13 WD Red 3 TB x 6 e Total Raw Capacity 29 TB 1 2 Purpose So yeah another build Another server to be pre cise Why Well as nice of a system ZEUS is it does have two major shortcomings for its use as a server When I originally conceived ZEUS I did not plan on using ZFS since it was not yet production ready on Linux at that point The plan was to use ZEUS HDDs as single disks backing up the important stuff In case of a disk failure the loss of non backed up data would have been acceptable since it s mostly media files As long as there s an index of what was on the disk that data could easily be
23. crews similar to those Lian Li use for their HDD mounting but the only ones I could find were so ex pensive that they d have cost me more than 100 USD So yeah nope 2 5 CHIPSET COOLING Figure 2 42 Simple reliable no need to make things more complicated than needed Figure 2 43 Finally mounted in the case Fits nicely Very happy with the result 2 5 Chipset Cooling and Ad ventures in Instability April 08 2014 Available online at 21 22 23 24 As some may be aware I originally had some is sues when trying to get this machine to run stable While stress testing with mprime it repeatedly and reproduceably crashed after less than an hour some times even already after a few minutes Each time after crashing it took me several tries and about 10 to 20 minutes to get the board to POST again After some troubleshooting and running a few di agnostics it turned out that the 5520 chipset was running really hot It s temperature threshold as in dicated by the system is 95 C and when I was last able to check on it before a crash it had already passed 85 C so I suspected that it was bumping up against the threshold upon which the board did an emergency shutoff and mandated a cooldown period until it would run again As an emergency fix I took the 80 mm San Ace fan that came with the case and mounted it to the chipset heatsink with some waxed cotton cord and voil somewhere slightly abov
24. ctually is it s pretty packed now with everything that s in there The impression is even stronger in person than on the pics Appendix A Diagrams 26 Intel ICH10R M B Business Server ZFS Pool mirror 3 way LSI SAS2008 LSI 9211 8i Personal Server ZFS Pool RAIDZ2 LSI SAS2008 LSI 9211 8i Media Server ZFS Pool LSI SAS2008 LSI 9211 8i Figure A 1 Diagram of APOLLO s storage topology LG Intel ICH10R M B Business Server ZFS Pool mirror 3 way LSI SAS2008 LSI 9211 8i ZEUS Business ZFS Pool Personal Server ZFS Pool RAIDZ2 JBOD 3 TB WD Red LSI SAS2008 LSI 9211 8i Media Server ZFS Pool RAIDZ2 oO O LSI SAS2008 LSI 9211 BEN A wh Figure A 2 Diagram of APOLLO s storage topology revised version 8G 84 MB s CIFS 105 MB s Figure A 3 Results of Storage and Networking Benchmarks APOLLO MEDIA VM APOLLO HOST SOURCE dev zero SINK dev null WRITING TO ZFS POOL FROM dev zero SOURCE dev zero SINK dev null COPYING FROM ONE ZFS POOL TO ANOTHER NFS 59 MB s CIFS 98 MB s NFS 101 MB s CIFS 93 MB s NFS 98 MB s CIFS 102 MB s CIFS 101 MB s CIFS 106 MB s NFS 109 MB s CIFS 95 MB s APOLLO PERSONAL VM SOURCE All tests performed with one single 10 GB dev zero file of zeroes Compression on S pools SINK disabled Values given are appro
25. e 70 C maximum Figure 2 44 My temporary solution to stability issues Apologies for the horrid picture quality this is a blowup out of a larger picture Figure 2 45 The stock heatsink Unfortunately I forgot to take pictures of that rather ghetto setup before dismantling it again and replacing it with something more solid but I have 2 5 CHIPSET COOLING managed to blow up some sections from another pic ture that should at least give you an idea of how it looked III JUE t zTTfU B SP ed Figure 2 46 The naked chipset after cleaning off the TIM rather annoying as it had dried Since the 80 mm fan is quite a bit larger than the chipset heatsink itself I needed to either replace the heatsink or modify it in order to be able to mount the fan to it I took a rather crude but very effective approach I took an L piece of aluminium drilled two holes across the heatsink cut some M4 threads on those two holes which worked despite the holes only going through the fins and not being continu ous then bolted the L piece to the heatsink with two M4 screws Works like a charm Cc 2 o LJ Sy Ohi LINE WR Figure 2 47 The holes are a bit unclean but it s not possible to debur them unfortunately and from the other side Figure 2 48 10 Figure 2 49 The fan is held down by three screws two in the L piece Figure 2 50 and one in the corner of the heatsink itself
26. e a nice and sturdy backplate too in fact it s among the most solid backplates I ve come across yet This does however require a slightly dif ferent mounting system You just have four screws which you bolt directly into the plate Aside from that the cooler is identical to its desk top counterpart as far as I know Why the 92 mm version For one thing it was in stock unlike the 120 mm version of this cooler Also the CPUs only produce 40 W TDP each so there really is no need for high end cooling And as a bonus I got supplied some awesome San Ace fans with my case which also happen to be 92 mm The Noctua fans which come with the cooler are just 3 pin fans the newer models of this cooler for LGA2011 come with a PWM fan I think but the San Ace fans I got with my case are actually PWM controlled Since the M B has a full set of PWM headers 8 to be exact how awesome is that I will try the San Ace fans and see how they play on lower rpm s they run at 4 800 rpm on full speed This does not need to be a super silent machine since it will be in its own room and since I really like the San Ace fans with regards to build quality and Pm a total sucker for build quality l d love to use them for this The Noctuas would admitteldy be better suited but I ll see how things go with the SAs first IN AUSTRIA mui Figure 2 1 The Noctua box Figure 2 2 the contents Thank you very much for choosing a Noc
27. he cables that go to the different controllers BLUE controller 0 GREEN controller 2 There isn t really any space to hide the cables so this was rather tricky and required three attempts until I was satisfied with the result In the end I hid the extra cable behind the triple fan unit good thing they re 38 mm fans which makes the space behind them just about large enough to fit the extra cable bits 15 The power cables for the disks are two cables that came with the PSU and onto which I just put a lot more connectors while taking off the stock connec tors because those were neither placed in the correct locations nor facing in the right direction 1 A y nnnm 98 M m TA j ye UN ay E amm r3 de Vim any YN f i K d WE d EM Nt j 3 iN Figure 2 80 OK then first try I soon realized that this wasn t going to work The problem was that I had the disks arranged in the same way as they will be set up in the storage pool layout so the disks which go into the same storage pool were also mounted below each other Sounds nice in theory but if you want to have disk from each pool distributed among the different controllers you ll get quite the cable mess 2 8 STORAGE TOPOLOGY AND CABLING Figure 2 82 Different angle On the second try I arranged the disks to that the cables to the controllers could be better laid out E UE ier own
28. he process added some dampening foam between the mouning rails and the fan frame 11 Figure 2 55 Just some foam The complete fan panel assembly Figure 2 56 Looks neat doesn t it While doing some test runs I noticed that a rather large amount of air was being expelled through the front of the case instead of going into the M B com partment and out the back I wasn t really surprised by this seeing as how open the front was Obviously this was not optimal So I took a 1 5 mm panel of alu and bolted it to the front Because the existing front has a few folds in it I needed to do some cutting on the case first Md Figure 2 57 Can t mount a panel to the front of the frame with these folds 2 6 DISK VENTILATION 12 the bare edge with some model paint to not have the blank metal staring at me I thought about painting the mesh but at the mo ment I don t really have the time plus I kind of like the look of this bare piece of alu so I ve left it as is LA Figure 2 58 Making some marks for where I need to cut Figure 2 61 The opening for the mesh The mesh doesn t cover the entire fan area nor is it very open with those rather narrow slots but there is no need for high power ventilation here so this is not a big deal Figure 2 62 Checking if it all fits It s fixed to the inside of the panel with some double sided adhesive tape Figure 2 60 Cut and mounted
29. ive cage I mentioned the top drive is the WD Velociraptor Other than that I could have gone with some cheaper CPU coolers as the 40 W CPUs btw core voltage is 0 9 V nice don t really require much in that area but the rest is pretty much what I want need for an acceptable price JL TA I Figure 3 4 And open on the CPU side uni inn T NIU ul ul Wi MU till MTT TT TI 000099100000 itt Figure 3 3 The box in its closed state Figure 3 5 Quite a few fans in there 22 23 Figure 3 6 Some more cable shots because why not Figure 3 7 and more The power connectors look crooked because they re not all connected to a drive so the ones which are floating freely in the air don t align perfectly Once the build is filled up with drives this will look much better 24 Figure 3 9 One of the cable looms fanning out to the drives Figure 3 10 The Chipset cooler and data ca bles Not too much room in there now Figure 3 11 Switched the Noctua fans back on the coolers since I m using the San Ace ones for compartment ventilation Figure 3 13 The back side q um ll And that s it we re done Thanks for reading see you next time 20 Figure 3 12 Looks better with six sticks in there I d say Figure 3 14 And the motherboard compart ment It s kinda funny Considering how large the M B compartment a
30. ly offers four in its stock form which you can upgrade to a maximum of thirteen drives You would need to buy another four disk enclosure which btw I could not find anywhere to buy and a five disk enclosure for the 5 25 bays Figure 2 25 Progress Since 13 drives aren t really that many and since these enclosures aren t exactly cheap I decided to go another route It took me a while to figure out how to do it but in the end this is what I came up with I had very generous help from one of my neighbours who has a mill and a lathe at his disposal as well as plenty of time he s a pensioneer The Mill can also serve as a drill press The drill chuck he looks ridiculously huge when you put a small drill bit into it he said they didn t have the smaller model in stock when he needed to buy his so he went with the large one Figure 2 26 It s pretty massive Figure 2 27 Lots of work 2 4 THE DISK RACKS Lots of holes needed to be drilled for the pop rivets that were going to hold it all together o Y Im Ter eer 7 Figure 2 32 The last one of these This is how the construct looks on the side where you slide in the disks You can see the pop rives I used to assemble it the slots which are pictured being milled above for the screwheads and the screws on the disks You can also see the recesses into which the screws mounted on the HDDs lock The system works very
31. m a KVM QEMU virtual machine running Arch Linux 2 9 2 Storage Baseline Performance I had a similar issue with the storage subsystem at first the default parameters for caching were not very conducive to high performance and resulted in some pretty bad results Writing toa ZPOOL of 4 x 2 TB WD RE4 resulted in speeds of a miseable 50 MiB s at first with option flag cache writethrough Read ing from the pool was possible at a nice 240 MiB s though Once I fixed that though setting the writethrough option to none much better and sufficient to saturate a gigabit networking connection 2 9 STORAGE AND NETWORKING PERFORMANCE 20 QEMU READ WRITE CACHE FLAG ZPOOL MiB s MiB s writethrough 4x2 TB REA 50 240 none 4x2 TB REA 155 210 none 6x3 TB Red 240 240 Table 2 3 Read Write speeds depending on QEMU s cache flag for my two main ZFS pools 2 9 3 Actual Performance over Net work The results can be found in appendix A 3 Initially I got only around 60 MiB s for NFS after that the next plateau was somewhere between 75 MiB s and 80 MiB s and lastly this is the current situation I must say I find the results to be slightly peculiar Pretty much everything I ve ever read says that NFS should offer better performance than CIFS and yet for some reason in many cases that was not the result I got I m not yet sure if I ll be going with NFS or CIFS in the end to be honest On one hand CIFS does give
32. my better performance for the most part but I have found NFS more convenient to configure and use and NFS performance at this point is decent enough for most of my purposes In general I find the NFS results just rather weird to be honest But they have been reproducible over different runs on several days so for the time being I ll accept them as what I can get As an alternative I ve also tried FTP on recom mendation of Whaler 99 but results were not really very satisfying only yielding slightly more than 50 MiB s on average 2 9 4 ZFS Compression Also for those curious about ZFS compression which was usually disabled in the above tests be cause zeroes are very compressible and would there fore skew the benchmarks I did a quick test to com pare writing zeroes to a ZFS pool with and without compression This is CPU utilization without compression the grey bars are CPU time spent waiting for I O not actual work the CPU is doing Figure 2 103 CPU load without compression Figure 2 104 And the write speed for that specific test run With lz4 compression enabled the CPU does quite a bit more work as expected though it still seems that you don t really need a very powerful CPU to make use of this Figure 2 105 CPU load with compression The transfer speed for a zero file goes up to almost 1 GiBE s with lz4 compression enabled pretty neat if you ask me Side note ZFS lz4 comp
33. n Figure 2 7 I might still go with the stock Noctua fans but it s not the plan at the moment I had to improvise a bit with mounting the San Ace fans to the tower The clips which you d use with the Noctua fans rely on the fan having open corners which the San Ace s do not Ah well nothing a bit of cotton cord can t fix Figure 2 8 The San Ace fan mounted on Noctua heatsink Figure 2 9 The San Ace fan mounted on Noctua heatsink Side note The coolers were actually more expen sive than the CPUs 2 2 TEMPORARY SETUP OS INSTALLED 2 2 Temporary Ghetto Setup OS Installed November 16 2013 Available online at 9 10 11 12 1 J 1 1 18 16 00 16 70 15 48 1 E DOS nono nro rro 7 projects mi lkyway cs rpi edu_mi lkyway ni lkyway_separation__modified_fit_1 28_x66_64 pc linu 4 22 gU Li ul MA xB6 64 pc _rosetta ninirosetta_3 46_ _64 pc linux gnu abinitio u 7 7pro jects mi lkyway cs rp a_i Ikywaym i lkyway_separat ion__modified_fit_1 28_x66_64 pc 1 inu tex 7 7pre jects boinc baker lab htop usr bin boinc client dir u 7 pro jects boinc baker lab bake 35 74 t 37 b rn 98 39 1 82 78 30 70 E 39 06 10 2 E 28 83 89 75 b 79 98 SETE AN 33 eo oooooooooocoscoo a EEEESENEE dU AAA EAT ETE TE Ww uuu uuu SSSTSSSSTSSSSSSSSSRKEHRHSTRSSEEE 438 24999999999999999899090999998099999 GRY E WWWWWOWOESSWUNSS
34. nance high reliability and availabil ity within reason it s still a home and small business server 1 9 Modding Instead of some uber expensive W C setup the main part of actually building this rig will be in modify ing the PP689 for fitting as many HDDs as halfway reasonable as neatly as possible I have not yet de cided if there will be painting and or sleeving and or a window A window is unlikely the rest depends mostly on how much time I ll have in the next few weeks this is not a long term project aim is to have it done way before HELIOS Also since costs for this build should not spiral out of control I will be trying to reuse as many scrap and spare parts I have laying around Chapter 2 Chronolical Transcript of Building Process 2 1 First Hardware Tests amp Noctua NH U9DX 1366 Novemer 14 2013 Available online at 5 6 7 8 M B CPUs and memory have all arrived The CPUs and M B seem to be working OK One of the memory modules seems to be having a bit of trouble being recognized the other five work fine I ll see if it s really defective or if it s just the IT gods screwing with me a bit 2 1 1 The Noctua NH U9DX The Noctua NH U9DX 1366 is a cooler from Noc tua s series specifically made for Xeon sockets For those who don t know LGA1366 sockets have an in tegrated backplate just like LGA2011 which makes them much more convenient than their desktop coun terparts It s quit
35. ov 22 2014 2014 Apr Disk Ventila tion themodzoo com Online Avail able http themodzoo com forum index php act findpost amp pid 25321 Accessed Nov 22 2014 2014 Apr Disk Ventila tion overclock3d net Online Avail able http forum overclock3d net showthread php p 743951 post743951 cessed Nov 22 2014 Ac 2014 Apr Disk Ventilation over clock net Online Available http www overclock net t 1442386 post__22085853 Ac cessed Nov 22 2014 2014 Apr Fan Unit for Main Com partment Ventilation linustechtips com On line Available http linustechtips com main index php act findpost amp pid 1844788 Ac cessed Nov 22 2014 30 31 32 34 35 36 97 38 LL 39 40 34 2014 Apr Fan Unit for Main Compart ment Ventilation themodzoo com Online Available http themodzoo com forum index php act findpost amp pid 25379 Accessed Nov 22 2014 2014 Apr Fan Unit for Main Com partment Ventilation overclock3d net On line Available http forum overclock3d net showthread php p 7444524 post744452 Ac cessed Nov 22 2014 2014 Apr Fan Unit for Main Compartment Ventilation overclock net On line Available http www overclock net t 1442386 post__22096775 Accessed Nov 22 2014 C Sebring 2011 Jan IN WIN
36. rage and Networking Performance themodzoo com Online Avail able http themodzoo com forum index php REFERENCES 41 42 43 44 act findpost amp pid 26162 Accessed Nov 22 2014 2014 Apr Storage and Net working Performance overclock3d net On line Available http forum overclock3d net showthread php p 750157 fpost 750157 Ac cessed Nov 22 2014 2014 Apr Storage and Networking Performance overclock net Online Avail able http www overclock net t 1442386 post 22177739 Accessed Nov 22 2014 2014 May Sound Dampening and Final Pictures linustechtips com On line Available http linustechtips com main index php act findpost amp pid 2008737 Ac cessed Nov 22 2014 2014 May Sound Dampening and Final Pictures themodzoo com Online Avail able http themodzoo com forum index php act findpost amp pid 26875 Accessed Nov 22 2014 45 2014 May Sound Dampening and Final Pictures overclock3d net On line Available http forum overclock3d net showthread php p 755148 post755148 Ac cessed Nov 22 2014 46 2014 May Sound Dampening and Final Pictures overclock net Online Avail able http www overclock net t 1442386 post_ 22245033 Accessed Nov 22 2014 39
37. ression is allegedly smart enough not to try to compress incompressible data such as media files which are already compressed which should prevent such writes from being slowed down Very nice IMHO Chapter 3 Result and Closing Thoughts June 23 2013 Available online at 43 44 45 46 As mentioned previously the 92 mm fans are rather noisy but I didn t want to replace them For one thing I do actually need some powerful fans to move air from the HDD compartment into the M B com partment on the other hand I didn t feel like spend ing more money on expensive fans For this purpose I ordered some AcoustiPack foam in various thicknesses 12 mm 7 mm and 4 mm and lined parts of the case with them I wasn t quite sure how well they would work as my past experiences with acoustic dampening materials weren t all that impressive but to my surprise they re actually pretty damn effective I have also put in another 12 GB or RAM I was lucky enough to get six 2 GB sticks of the exact same RAM I already had for 70 USD plus shipping and fees but still a pretty good price IMHO from eBay 24 GB should easily suffice for my purposes Lastly ve repurposed the 2 5 drive cage from my Caselabs SMH10 cleaner than the rather improvised mount from before For the time being the build is now pretty much complete One of the original goals was to not have this be come ridiculously expensive Uhm yeah
38. s 4d deed ee LER ap oe ee RW BR cn eee He edem E a 5 pst Bracket Comision aeuum A A ES We segs ee a ee ee eee ee ae aei d 5 PSU Bracket Colisior wA Plue seai ek cb uoti Ned iet ee eee ok ee eR td d 5 PSUS COOL OR TUS o donde de iced medi dedo e ug hdl ote ete she se edel iiy de eee d 5 PSU Mounted variada AAS Be M YU Fist Son oves SS UU iod doe dd 5 Pee MOUNT ae RECIBE ah he Sle eh ca a See de de Ree UN A Hecate a ha ee ae S 5 Lh Controller Testo nu hee ok Ao heed eee ee eo Oe ae She SES 5 EsiController Testit mon ded de ahi Ge eae DR a ee he o e e eem d 6 Disk Racks Dbhe Ville 2 us sete eet Oe en tte REREAD ee ee oS 6 Disk Racks First SUCpS ral RM BPE Me IA E Sb eee Shh ee eS ee s 6 Disi racks Fro iOS cede lA BS Spe are ea a eee Se SEIS S 6 Disk Rick rill COIUCE S utat ws we 49 9 5 2 Ut wid uet wo OE e eee f sce ES 6 Disk Racks Predrllns Or RIVS uuo ud o uoa aa WR REC eed Ae ew XE RR s 6 Disk Racks More Diino 650 as px of CS ee Re OE EC oe eur dedos rr hp ES S 7 Disk Racks MINS e united ir due BE Se a Re he as a E dt Rae R x S i Disk Racks Cage Panels 4 4 4 e S9 RP etd A ox one dy x eg ER Ee OR RU Wee os T Disk Racks Cage Panels ur aseo xm do ou BS oe EEE UE qe eS SS ace EE T Disk Racks Cace Panels oe oe eee hb ow Sus v See he eS oe aa Bie 4 T Disk Racks Ral Detail i teo dba aeti tie Bo amp BLL ee a Be XC XE ia ee EA ecc f T Disk Racks Rail Detail Y 442 ie ed gt 6 206 IRE Pt ode de 9 s a Aa
39. s usually blow air through along their longitudinal axis there is no ventilation hole on the case for a fan on the top of the PSU which most of today s PSUs have Not to worry I still had an old Aquacomputer rad grill laying around A bit of dremeling should be able to fix this problem Marking for cutting 2 3 PSU MOUNTING amp LSI CONTROLLER TESTING 5 Figure 2 18 like so o E a enough air R sitili ITEMS iy Platimax Figure 2 15 And with the grill mounted Since this case was not designed with standard ATX PSUs in mind some cutting was required to make it all fit nicely Native cables AP Mainboard 4P 444p CPU Y Figure 2 20 Fits nicely now Needed to hook up some HDDs to test the LSI controller Looks very ghetto worked like a charm Figure 2 21 Yeah I won t be keeping things like this Figure 2 17 This will be fixed naturally 2 4 THE DISK RACKS Figure 2 22 And the HDD package 2 4 The Disk Racks April 02 2014 Available online at 17 18 19 20 The main part of this undertaking As mentioned elsewhere one of the two main prob lem of our current server is that it only has seven HDD slots and they re already all filled up The only way to get more storage is to install larger disks which isn t really all that oru a Figure 2 24 First steps One of the main objectives of this build is to have more disk slots The PP689 on
40. tua cooler for betel Kaon veu Please install the cocker accordieg to this moral Should you encounter omy difficulties please check the FAQs ot ou emba end don heutyta fo conto out support teem and unpacked Figure 2 3 2 1 FIRST HARDWARE TESTS amp NOCTUA COOLERS 3 The Box Unlike its shiny desktop counterparts the NH U9DX comes in a nice and subtle but sturdy cardbord box with a simple sticker on it I must admit I like this box more than the shiny ones Figure 2 4 Some glory shots Figure 2 5 does look nice doesn t it There is no info about this fan on the web I m presuming it s something San Ace makes specifically for InWin in an OEM deal I ve hooked it up to a fan controller and got a max reading of 4 800 rpm and the Supermicro board turns them down to 2 200 rpm on idle They seem to be very good fans you can only really hear the sound of the air moving no bearing or motor noises so far Also they are heavy 200 g per piece which is al ways nice for a build quality fetishist such as myself Figure 2 6 One of the San Ace 92 mm fans Note Hooking such a fan up to a desktop board as its power source would not be advisable they are rated for 1 1 A and might burn out the circuits on a desktop board Server boards usually have better fan power circuitry since they are desinged with high performance fans in mind Just as a side note and compared with a Noctua fa
41. ximate dev null averages over multiple runs SOURCE dev zero SINK dev null UV List of Figures 2 1 2 2 2 3 2 4 2 9 2 6 2 2 8 2 9 2 10 2 11 2 12 2 19 2 14 2 195 2 16 2i 218 2 19 2 20 2 21 2 22 2 23 2 24 2 25 2 26 ELSA 2220 2 29 2 30 2 31 2 32 2 33 2 34 2 39 2 36 2 37 2 38 2 39 2 40 2 41 2 42 2 43 2 44 2 45 2 46 2 47 2 48 2 49 NOC ADOR ue i descen nde E uda e deiude uer a e A RE 2 Noctua Box C ONOS i e acute ue re ee Sk Se Ss a Joder E ood ut AE S 2 Nod Uar BOX COmMent o adus Moe wes os eS erit e BEC de Yee e dede m e ede E 2 Noctua Cooler lo acia BO weh Bo Nhe ote eA AA Ros Roue d x Xon s 3 Noctua Cooler Glory 24 44 uve ead v we wende vu igudexgedxX YS wae Bee hee s 3 Dat T OZ MIO BOM Los Gee oe gs ee hg Ege oe Beh TII 77 1T 2 ew e a 3 San AGS 09 MIRAD 0d Hoe e a et eae ER ea ee de Vg hase Ne SEVA RC RP Ge 3 Noctua San ACS Editio utu eo ibo X Rode Se BS ee EEE OX qe AERE SSS aue x ESE 3 Noctua san Ace Edition score age xmi oe Ses v See ha eS oe aa Bie oS SE 3 BO AV OP ese weet ieee dao sg Se aired de Pie Rs LIES ae RUMP ee HC HAGO ETE he E A d 4 Ghetto Testing Setup 4 4 4 acer y edes x ewe ee Eee EOE PhO ew 24 Bod eG 4 pst Dad beso So pe et a e SOR OAOGee hu xe VEA ok ede MUSEUM NOR Ee A Ae 3 4 SE OO lie Bs are ae ote eae ee Bw a a O ee ete ae Be ee Boe 4 PSU Cutout Markings lt e und doped eo pede bos ee wa bee eee agr pope 5 POU Ventilation
42. you know how these things usually go Total system cost 5 000 USD of which were HDDs 2 500 USD My share of the total cost is 42 the remainder was on my dad which is pretty fair I think In the long run my share will probably rise as I ll most likely be the one paying for most future storage expansions at the moment I ve paid for 54 of the storage cost and 31 of the remaining components One thing to keep in mind though is that some of these costs go back a while as not all HDDs were bought for this server but have been migrated into it from other machines So the actual project costs were less by about 1 300 USD Overall Pm still pretty happy with the price performance ratio There aren t really that many areas where I could have saved a lot of money without also taking noticeably hits in performance or features 21 I could have gone with a single socket mother board or a dual socket one with fewer features say fewer onboard SAS SATA ports as P m not using nearly all of the ones this one has due to the 2 TB disk limit but most of the features this one has I wouldn t want to miss TBH the four LAN ports are very handy and IPMI is just freaking awesome And let s be honest A dual socket board just looks freak ing awesome OK I ll concede that that s not the best argument bit still it does Figure 3 1 The inner side of the front panel with sound dampening Figure 3 2 That Caselabs dr
Download Pdf Manuals
Related Search