F@H Ghost

[PCTC2]

I haven't had a machine running F@H since the race in December... yet in the past 4 days I've been getting 80k ppd. WHERE IS THIS PHANTOM F@H CRUNCHER? WHO ARE YOU? WHAT MACHINE ARE YOU USING?

 

[petrusbroder]

A forgotten hero-folder?

 

[blckgrffn]

Hopefully you scrubbed the client from all of those machines you were testing

 

[Drsignguy]

Freebie! :thumbsup: So, does this mean they have your account # too? ^_^. Nah!

A thief that does good for all humanity.....................sweet

All in good fun.

 

[PCTC2]

I found the "ghost". Someone rebooted the PXE server and I guess I had F@H set up to start on reboot (I had stopped it at the end of the race). So it's my own.

Well, hopefully in the next few months I'll have a few more servers to play with. I'm hoping to get a full-time "stability and thermals stress test bench" system which should double my PPD when it's not being used.

Also hopefully getting in a cluster with 1536 cores/3072 threads and probably 6-12TB of RAM. Should have a week to stress test that sucker.

 

[theAnimal]

I found the "ghost". Someone rebooted the PXE server and I guess I had F@H set up to start on reboot (I had stopped it at the end of the race). So it's my own.

Well, hopefully in the next few months I'll have a few more servers to play with. I'm hoping to get a full-time "stability and thermals stress test bench" system which should double my PPD when it's not being used.

Also hopefully getting in a cluster with 1536 cores/3072 threads and probably 6-12TB of RAM. Should have a week to stress test that sucker.

Sounds like fun!

 

[somethingsketchy]

I found the "ghost". Someone rebooted the PXE server and I guess I had F@H set up to start on reboot (I had stopped it at the end of the race). So it's my own.

Well, hopefully in the next few months I'll have a few more servers to play with. I'm hoping to get a full-time "stability and thermals stress test bench" system which should double my PPD when it's not being used.

Also hopefully getting in a cluster with 1536 cores/3072 threads and probably 6-12TB of RAM. Should have a week to stress test that sucker.

By chance the cluster isn't something like this?

http://arstechnica.com/information-...-a-cloud-in-a-box-and-he-wants-you-to-buy-it/

 

[PCTC2]

By chance the cluster isn't something like this?

http://arstechnica.com/information-...-a-cloud-in-a-box-and-he-wants-you-to-buy-it/

no. Though I would really like to grab large cloud and virtualization cluster contracts, I generally handle compute clusters using job schedulers, like large ROCKS deployments (well, no. I don't do ROCKS. But payware equivalents.).

A little more dense. Like up to 64 threads and 512GB of RAM per rack unit. So up to 2560 threads and 20TB of RAM per cabinet.

The interesting thing about Nebula is it looks like unbranded Dell servers in their photos. (not their controller, but the other servers they don't sell). OpenStack is cool. OCP is cool. I like where commodity hardware, open source software, and other high-density systems are headed.

 

[somethingsketchy]

...A little more dense. Like up to 64 threads and 512GB of RAM per rack unit. So up to 2560 threads and 20TB of RAM per cabinet.

Me = jealous I won't lie, I always wanted to see VMware vCenter with that kind of computation power. If only there were demo's of this hardware more local to Vermont, than Boston or NYC

Just out of curiosity, how long does it take to source the hardware with the density of what you're prescribing?

 

[PCTC2]

Me = jealous I won't lie, I always wanted to see VMware vCenter with that kind of computation power. If only there were demo's of this hardware more local to Vermont, than Boston or NYC

Just out of curiosity, how long does it take to source the hardware with the density of what you're prescribing?

A few days to a few weeks. Depending on the manufacturer and distributor, age of the product, and the availability. Some CPUs are readily available and others have 3 week lead times. RAM is always quick. High speed interconnects can vary. Motherboards, power supplies, and chassis can be pretty quick unless the manufacturer is revising it.

High density storage, on the other hand, is crazy heavy and a pain in the butt. But a Petabyte is actually pretty small nowadays (physically). You can actually fit over 2PB raw in a rack cabinet. Even with relatively low-density DFS systems, you can fit 1PB raw in a rack... with a lot of compute power and a healthy amount of RAM.

Here's some examples of systems:

12-core 192GB RAM 8xGPU Compute Server

4-socket Intel w/ 768GB RAM