Monolith 2014 10

2014-10-31

I still have to cut the acrylic plates for each end to mount it into the rack but the new tier is already being used by John The Ripper runs on sample passwords. The current set of test data are all md5 passwords from the standard linux encryption method.

2014-11-01-002-small.jpg

2014-10-30

I have all 11-nodes of the fifth tier working (I received more micro sd cards and wifi adapters.)

2014-10-25

I have 11-more nodes now so I am building 1-more tier.

When that work is completed the total systems in the cluster will be 54.

I have (4) of the new tier up and running now but ran out of micro sd cards as well as wifi adapters. I have both on order.

2014-10-28-003-small.jpg
I did wire all 11-nodes into the new power harness for the third 12v-5v dc converter but I have still not designed a power-up method as using 3 separate SSR became annoying for me.

I have gone back to a blank page to revisit how to achieve my goal of being able to sequence 11-nodes on at a time to help ease the massive start-up current demand.

I still have to cut the acrylic plates to mount them and I'm also waiting on some new sleeve material that will allow me to ditch the ugly automotive-style wire loom.

I will mount these on the lowest part of the rack just so I can move the master node to the first physical system and install the SSD disk inside the base.

Thanks once again to David and the team at r0ck.me for the incredibly fast shipping.

2014-10-21

You are looking at my first creation with a sheet-metal brake. This particular piece is made from 20-ga aluminum.

2014-10-21-001-small.jpg

I recently obtained a very nice tool for the lab that can bend, shear and roll up to 20-ga steel.

woodward-sp3-1-12.jpg

This is the Woodward Fab SP3-1-12 model. They have up to 30" but that was too large for me and since it still only handles up to 20-ga steel the next sheet-metal brake I buy will be something that can handle 18-ga steel.

I also need to get a small drill press some day. I have only limited area for the lab but I do like being able to fabricate prototypes now.

2014-10-15

I mounted a few things inside the box and shortened the DC power wiring even further. I cut off the AC power cord from the power converter/charger and spliced it onto a heavy-duty power cord (because the original cord was too short.) I also cut a hole in the rear of the box and ran the power cord outside. Getting all this stuff to fit inside the box wasn't easy and repairs (if ever needed) won't be easy without disassembling a lot of stuff.

2014-10-15-003-small.jpg 2014-10-15-002-small.jpg

I did receive the heat sinks for the SSR but I am not sure how I will place them inside as space has become very limited.

2014-10-15-007-small.jpg

2014-10-13

I am waiting on the arrival of the heat sinks for these SSR.

2014-10-13-001-small.jpg

2014-10-11

I shortened the dc power wiring and mounted the three voltage converters onto a strip of aluminum. The strip will be mounted inside the box near the top. As I mentioned earlier each voltage converter will power 22 systems. This will be roughly 40A load (on the 5V-side) when the cluster is busy.

2014-10-11-002-small.jpg

2014-10-08

Updated 2100MDT

I cut the lexan pieces, drilled holes and mounted the 4th tier of 11-systems into the rack. The green leds are the disk in use indicator and not the boot (Android) default.

2014-10-08-008-small.jpg

I need to work on the inside of the box.

1. Strengthen bottom or perhaps attach a battery box that is suspended from the sides removing the weight from the bottom and transferring this to the sides. I can fabricate some sort of mount from aluminum and wood and epoxy glue.

2. Fabricate several internal chassis:
2.1 (3) Solid-State Relay (SSR) with heat sinks to handle the power control for each set of 22-systems. This will allow me to digitally control the power line.

ssr-mounted-heatsink.jpg

2.2 (3) 12V-5V (DC to DC) converters
2.3 SSD (/master) for master control node.
2.4 12V muffin fan (tentative.)

3. Open power converter/charger and replace AC cord with something much longer. Cut hole and install grommet in side of box to feed AC power cord into interior of box.

4. Cut and fit aluminum mesh to top of box/base of Cubical Monolith rack.

5. Glue strips of aluminum with threaded studs (screws epoxied into aluminum plate with thread sticking out. These will act as mounting plates for all the chassis mentioned above

6. Trim all wiring to minimum length.

7. Mount power converter to side of interior box.

8. Move master control node to lowest tier (move sata cables and serial console and micro sd card.) This will allow me to place the SSD drive inside the lower enclosure and out of sight.

Earlier:
I received the first set of wooden boxes. The bottom is a bit understrength to hold the battery over the long term. I will have to strengthen the bottom with something.

2014-10-08-001-small.jpg 2014-10-08-004-small.jpg

As you can see I have not attached the base plate to the rack nor have I placed the 12V-5V converters inside. I still have not had a chance to cut the lexan sheets (I am going to re-use the former side pieces if possible.)

2014-10-07

I had originally planned on powering 33-systems with (1) 12V-5V converter but it seems that when the cluster is busy the current demand was 20.7A which was too much for a single 5V converter.

2014-10-07-001-poster.jpg
I still have to cut and drill the lexan pieces that will mount the "tier" of 11-systems into the rack.

I should get the wooden boxes this week (the first and taller set) so I will be test-fitting the power converter, the (3) 12V-5V converters, the solid-state relays with their heat sinks (assuming I get the heat sinks this week), and most likely the SSD. I might need such tall boxes any way by the time I finish installing everything in there. I still have the time-delay circuits to integrate as well.

2014-10-07-003-poster.jpg
I rewired everything today and have 22-systems per converter now and that will be the final design.

I only checked the input side (12V) which draws 9A but the 5V side was hitting the maximum limit and the manufacturer warned not to use them at 20A for very long at that.

It seems I did not account for the additional current load when the cpu were busy. I am impressed that it held together this long as it has been in use for weeks.

This means I will use (3) 5V converters instead. The current draw for 22 systems was 12.9A when the cpu were at 100% utilization. Only one of those systems has an (SSD) hard drive.

2014-10-05

I was able to get more nylon spacers and (11) more wifi dongles. All 43 nodes in the cluster are operational although I have not yet cut the lexan and mounted them in the rack.

So far the access point is working normally. I did reset everything back to normal without using RTS threshold or short preamble. I am serving 43-clients from my DD-WRT Netgear WNDR3700 v2.

2014-10-05-005-small.jpg

I ordered a few sets of these Dubro threaded inserts (4-40.) I will use them to attach the aluminum plate to the wooden box. If I use epoxy to place the inserts I doubt if they will ever work their way loose no matter how many times I access the fasteners.

dubro-threaded-inserts.jpg

The plan will be to drill slightly undersize holes in the top of the wooden box placing these inside with some epoxy. Let the glue dry for about 24-hours and then I should be able to securely fasten the aluminum plate which will become the base of the cubical monolith rack. This will be a very secure, sturdy method to mount the cubical monolith though I may use more than (4) depending on how strong they are.

I also ordered some aluminum mesh as I will most likely either have to cut a huge hole in the center of the existing aluminum plate for a vent in which case I will use the aluminum mesh to create a nice-looking grill where I've cut OR if the ventilation is not sufficient I may use it for the entire base.

3003 Aluminum Perforated Sheet, Unpolished (Mill) Finish, H14 Temper, Staggered Round 0.25" Holes, 0.063" Thickness, 14 Gauge, 12" Width, 36" Length, 0.375" Center to Center

aluminum-mesh.jpg