Debugging Software using Virtual Networks

gunnk writes "In today's large intertwined networks it is awfully easy for a bug in one piece of software to have far-flung ramifications that may be horribly difficult to track down. For example, think about the chaos that could ensue if Cisco had a subtle routing bug in it's next generation of internet switches. Unfortunately, it is too expensive to build a test environment anywhere near the scale of the real environment, so the final "testing" occurs in the real world. A new idea has come along, however, that might just help: creating a virtual world in the real world network to test the software (with the added bonus of being able to step backwards through the process to analyze bugs). Sciencedaily is running an article on this method entitled "New Computer System Solves Problems by Tricking Computers"."

Avoid monocultures

[ka9dgx]

The simple answer is to avoid monocultures. They are very bad in agriculture, and society, and networks. Many vendors competing with a variety of implementations of a standard are a better way to achieve long term stability.

The internet is too much a brittle tree, instead of the mesh it was intended to be. If I have internet from two vendors, and you have internet from a different vendor, why shouldn't we be able run a wire, and add redundancy to the internet?

--Mike--

Re:Avoid monocultures

[Bookwyrm]

Indeed. People seem to be rushing towards making everything networked purely a monoculture -- "Everything on IP".

Re:Avoid monocultures

[PD]

That's backwards. The statement is "IP on Everything".

If you're thinking of marking this as a flame, take the time to research the T-shirt that the inventor of IP likes to wear. Guess what it says?

Virtual Networks for Testing?

[NetFu]

It's interesting that I see this article at the exact same time I'm actually testing some server software in three virtual Windows 2000 Advanced Server machines networked in a virtual network on my PC ... using VMWare. It allows me to test configurations and operations without actually using hardware and it's faster since I'm using backed-up VM's in case the software I'm testing destroys a VM (if it does, I toss the destroyed VM and unzip the original from a zip file in about 5 minutes). And, I can test virtually any combination of Intel-based OS's in my virtual network this way.

Anyway, back to work...

This isn't new.

[Mordant]

See http://www.isi.edu/nsnam/ns/ .

Heisenburg Uncertainty Principle

[FortKnox]

You are adding a new environment to a test environment. First of all, how tested is the virtual network virtual machine?
What's the odds of a bug occuring in the virtual network and not the real network?
What's the odds of a bug NOT occuring in the virtual network and in the real network?

There is no clear-cut way to be absolutely sure...