Greetings,
We design and build server farms for clients already through our IT consulting business. We have access to pretty good discounts on servers, and have both in house and colocation space for running them.
The in house facility, while small, is secured by the same systems covering our forensics lab. Security at the colocation facilities can be, at times, better.
The network security is more … interesting. You need a secure job submission and results delivery system. Screwing that up would be quite costly.
And you'd need to vet your clients. Cracking stolen IP or personnel files would be "bad".
-David
kovar
vetting definitely. but ppl seem locked into the idea of a PC as a Base Unit to crack. More PC's = Better .
Why use PC's? surely wasted resource with all that I/O capability built in.
whats the possibility of having a pico farm. And could such things be employed to generate rainbow tables given the raw processing power.
The DNA approach is great if you are "In company" but the hardware gains from dedicated boards may be the way to go if starting from scratch.
This is something i've been interested in for a number of yrs, but simply haven't had the resource to get above a hobbyist level.It's certainly worth a brainstorm.
kern
Greetings,
Both rainbow tables and things like the Tableau can eliminate a lot of PCs but you're right, most people still think of PCs as the smallest unit for building computing farms. There are some notable exceptions - SETI research using the processing power of graphics cards for example.
Tableau is 16 (?) FPGAs in one unit. Cascading four of those together gets you a 64 FPGA server farm in a single chassis, not too shabby. But at $5K a unit, that's an expensive server farm still.
Now, get a Pico Computing supercluster ….
http//
Don't know what their price is, though.
-David
Greetings,
Elcosoft is releasing a password cracking system that uses the GPUs in GeForce 8 graphics cards to get a 25x increase in speed.
http//
And
http//
A $150 graphics card vs a $5,000 Tableau. Hmmm.
-David
Sorry for the tangent here, but I can't believe they are trying to patent cracking passwords on GPUs.. it's not as if they even invented a new way of running code on the GPU, as they are just using Nvidia's CUDA SDK. It would be interesting to see for sure, but it's hardly "revolutionary" and patent-worthy.
Thanks for the link though, I read it before it appeared on Slashdot -) The technology gets more interesting when you expand it out using either multiple graphics cards or one of Nvidia's Tesla "Desktop Supercomputers" (http//
For a sub-tangent, check out Golden and Vassil's DFRWS presentation on using CUDA to speed up Foremost carving - I was very interested when I saw it (http//
Ok, back on topic -)
re GFX,
i think they are just making use of retail available fast chip technology, and its readily available built onto gfx cards.
Nice idea tho, use whats available and benefit from the economies of scale.
Kern
The reason they are using GPUs is that they are designed to perform maths-like tasks and can be heavily optimised to doing similar functions (like calculating password hashes) because they don't have overheads associated with multi-purpose CPUs. It's basically a cheap (but not as efficient) version of an FPGA.
Sure, it's probably a nice implementation and handy to have available, but heardly revolutionary or patent-worthy considering the amount of people who do essentially the same thing. I just hope it doesn't stifle similar development is all.
A lot of reaching around for this one post. Thanks guys, It must be great to have cubic$ for hardware. D . I started a GPU project for transcodeing video at my last company. I don't think they will get a patent.
http//
Swaping out my IBM dual proc for a Quad dual Mac xServe.
Greetings,
If you can string a couple NVidea cards, or PS3s, together and outperform Access Data's rainbow tables or Tableau cards, you will not need cubic$ for hardware.
Password cracking looks like it'll be pretty affordable soon.
-David