Networking Certification for Computational Biologist

Oh man! Ad-blocking software has been detected! :'(

This website is run by the community, for the community... and it needs advertisements in order to keep running. Blocking our ads means your killing our stats!
Please disable your ad-block, or become a premium member to hide all advertisements and this notice.

Hi guys

I have been toying with the idea of learning more about computer networks and their design.

I would like to , in the future, deploy distributed computing architectures for heavy computing work. As the title probably suggests, i deal with biological data such as genetics,etc.

To cut a long story short , i need to learn the intricacies of networking. Of course, my networking knowledge is limited to wireless routers,basic switches and inter connecting pcs using that,but nothing of the advanced features you guys must be knowing.

So any suggestions ? I was looking at the CCNA syllabus and i have brushed upon those topics in my undergrad studies, so its a little familiar. Plus a few google searches revealed i may even need to purchase some test kits/ hardware to actually practice the concepts.

Thanks guys

 

Oh man! Ad-blocking software has been detected! :'(

This website is run by the community, for the community... and it needs advertisements in order to keep running. Blocking our ads means your killing our stats!
Please disable your ad-block, or become a premium member to hide all advertisements and this notice.

Ccna is good for cisco related networking. For a general education I would look at the network + books. Then move to the ccna

 

The more expensive super computers do not use ethernet or TCP/IP for their inter-connects, it does't deliver the peak performance and low latency required.

There are many different ways of doing HPC, some do involve commodity hardware like 10 Gigabit Ethernet to build Grids or Clusters. However these days its most likely easier to rent cloud compute services like AWS.

I do not think a CCNA is the best use of your time if you are a biologist interested in computation. You should probably learn to program one high level language and learn a little about HPC and simulation.

Look at masters degrees that cover scientific computing.

Personally I'd want to see topics like CUDA, OpenCL, Hadoop, BigData, SQL, NoSQL, OLAP, MPI, OpenMP, Multi-threading, R, Matlab, C++, Python, HPC architectures, Simulation, Statistics, Genome Sequencing etc.

You should research the area of HPC, Bioinfomatics, Synthetic Biology etc, and what interests you on the internet.

If all you want to do is deploy commodity hardware then the CCNA may help a little, but you will be competing in a market where there is a lot of existing talent.

Cisco do make some Infiniband and fibre-channel devices but I have no idea if the CCNA mentions them.

 

Aha ... well .. am already a bioinformatician (or my undergrad degree says so :P ) ... and right now working towards a masters degree in computational biology,

have significant experience using java based neural networks, and then programming experience with C/C++, Python, Java, Perl, plsql, mysql, R ( and bio stats ...), bit of Stata.
and dabbled with distributed computing systems.

My main idea was to setup up those actual distributed systems. mainly building or designing custom clusters of computers and if possible, reinforced by GPUs, using either CUDA or openCL for the interfaces.

To be honest, i was looking at the CCNA certification for more indepth knowledge and to show potential supervisors that I am skilled in networking, making Phd applications a little better.

 

I'm not a HPC expert but I suspect that the most important aspects when designing such systems are :-

Performance per watt
Latency
Throughput
Cost per PetaFLOP

As such its most likely that experienced Electronics Engineers get the most input into their design.

Titan cost $100 million to build. The annual power cost is likely $10 million.

It uses a Cray Gemini interconnect made of custom router ASIC's, some systems use Infiniband FDR/EDR, you won't learn about custom interconnects or infiniband on a CCNA.

Gigabit Ethernet was supposed to take over the datacentre and HPC but its yet to happen.



Most applications will likely use MPI which will not use TCP/IP as the underlying transport on these sorts of systems.

Here is a cray document describing Gemini, its the sort of stuff you would need to understand if you wanted to be designing such systems.

 

Interesting development :- New EU based supercomputer to be ARM-based - CPU - News - HEXUS.net

Looks like ARM based processors could make it into the HPC space and change the current performance per watt and price characteristics.