Changing IT Demands on Network Infrastructure

Reviewing the evolving IT care-abouts in cloud computing deployments

Last week I received an email from Julie Bort, editor of this site, asking me a very simple question, "Where have you been?  Are you going to write again soon???"A good question!  It made me pause and reflect for a second, where have I been? - besides 'busy' which really isn't a place but seems to be a common destination all the same.  What seems to have happened in the past six months or so is that I moved from pontificating on cloud computing and the changing role of the network to actually executing on it a good bit  The good news is that IT infrastructure spending is back on the table, people are rolling out networks, servers, and storage to support their business growth  The bad news, for some people, is that in an increasing number of these engagements the 'care abouts' are changing.Our jobs as systems engineers, network architects, or CTOs in the infrastructure world used to be easier than it is today.  To evaluate a network vendor we would look at some of the following criteria:

  • Price- Capital outlay and recurring maintenance costs
  • # of ports
  • Speed of those ports
  • Ability to move packets from port to port
  • QoS and security Features on those ports
  • # of IP Routes
  • Features to construct a topology (i.e. routing protocols, rapid Spanning Tree, etc)
  • Level of vendor support

I am increasingly finding in this cloudy world that there are several more criteria, and that because of the way people are wanting to build clouds and virtualized environments the scaling metrics have changed or at least new pressures have been placed upon existing tables in a way that was somewhat unforeseen by the vendor community.Some of the questions I have increasingly heard in the past six months, in addiiton to the traditional ones, are:

I am sure there are a few that I have missed, but in short, it wasn't TOO hard to evaluate several vendors, pick the product that met your technical requirements, justify it to the purchasing department, and head towards implementation.

  • What is the revenue per square foot I can derive from this architecture?
  • What is the power cost per port, and per wirespeed port?
  • What is the three year OPEX on this deployment?
  • How many MAC/ARP/IP entries can you support? Can you handle a large/flat network of N thousand hosts? and NN thousand VMs?
  • How can we program the infrastructure to create a differentiated service level?
  • Do you have programmable APIs in the operating system?  What are they based on? XML-RPC, SOAP, NETCONF, XMPP? 
  • What distributed message bus are you building between your systems?
  • Can you support 24/7/365 operations with zero planned downtime? 
  • How do I automate network operations and tie them into my OSS/BSS or Cloud Provisioning system? 

In short, the world is changing a bit.  The cutting edge customers are applying pressure to networking vendors to create new classes of capability, more operational focus, and a generation of products not just defined by speeds and feeds, but by the capability to be managed, by a machine.

Join the Network World communities on Facebook and LinkedIn to comment on topics that are top of mind.
Now read: Getting grounded in IoT