Welcome to Netcordia Connection Sign in | Join | Help
in Search

Terry's Blog

Cloud Computing Applied to Network Management

One of the new technology rages in the market today is Cloud Computing (and Storage), of which Amazon's Web Services (S3, SQS, SimpleDB, EC2, and HaDoop) is one example.  As I was reading through the example of "Grep the Web" described in the paper Cloud Architectures, I wondered what applicability this technology had to network management.  In just a few minutes, I had a couple of ideas.

  1. Data storage.  Keep large NMS data sets in the S3 storage system.  The rates seem pretty reasonable for bulk storage ($0.15 per GB-Month), plus an additional amount for data transfers in each direction.  But if you're primarily using it for backup storage, you'd only have a one-way trip for most data.  Deletes are free, so when you age out the data, just delete it.  If you never need to go back to it, you're using Amazon's storage system (servers, disks, racks, facilities, and staff) to store your data.  I'd want to encrypt the data before sending it to them, but that's easily done.  Imagine keeping detailed weekly data for an entire year.  If your NMS can create a summary of each week, store the details on AWS S3 and keep the summaries around for your trend analysis.  Then, if you need to go back to research some detail, you can easily retrieve the detailed backup data.
  2. Routing analysis.  Large enterprise or service provider routing infrastructures are pretty complex systems.  If all the routing tables were extracted and stored in AWS S3, then the other components of AWS could be used to perform analysis on the resulting routing tables.  With the large data store provided by AWS S3, a time-series of routing tables could be kept and, using the other AWS components, a trend analysis could be performed on the tables.  I'll be that there are some interesting visualizations that could be generated from a time series of routing tables.  For example, build a topology map of interconnected routers and color the devices according to the number of local route changes.  Show a sequence of these topology maps as a movie.  Imagine being able to watch an EIGRP Stuck-In-Active route in action as a way to help diagnose what's happening.

I'm sure there are other, more detailed analysis that could be performed if the large volumes of NMS data could be kept and analysis routines were developed to mine it.  What suggestions do you have for using AWS for network mangement?




Michael said:

Hey Terry,

You are absolutely correct on the routing visualization. Packet Design and HP have products that do what you describe, cloud computing not required. You can store over a years worth of routing data in about 100G.

I agree with you that the advent of cheap storage and compute cycles will make analyzing the mass of data that can be collected from networks and systems practice and give the opportunity for unique analysis of data, like anomaly detection baselines.


August 29, 2008 9:59 PM

Amazon cloud computing service applied to network management said:

Pingback from  Amazon cloud computing service applied to network management

September 26, 2008 5:02 AM

Cloud Computing Applied to Network Management | IT Management and Cloud Blog said:

Pingback from  Cloud Computing Applied to Network Management | IT Management and Cloud Blog

September 26, 2008 9:38 AM

This Blog