Welcome to Infoblox NetMRI Community Sign in | Join | Help
in Search

Applied Infrastructure

Device Modeling

Some network configuration management vendors are building systems that model network devices.  Their intent is that their customers will find that working with a model of a router or switch will be easier than working with a larger number of individual device types.  The alure to the network configuration vendor is that they can write one set of policies and apply it to a wide number of devices. There are at least two reasons why this strategy won't work.

The first reason has to do with training customers on the generic model.  That's the crux of the problem -- it is a new syntax that the customer must learn.  Instead, should the vendor choose a device-independent syntax or design around the leading equipment vendor's syntax?  Both approaches are fraught with problems.

If a new language is used, the customers must be trained on the language, which is a significant burden on the network management vendor as well as the customers.  If a major equipment vendor's language is used, that creates a burden on the customer to remember which syntax to use for the network management system and on the device itself.   I see a large number of human errors occurring in these scenarios.  There are also enough variations between similar hardware of major vendors that this choice has significant problems.

The second problem is that for a given problem, it is practically impossible to use a generic model to match that of real configurations.  My prime example is QoS configuration where the variation between devices is too great to handle.  The underlying hardware, often the ASICs themselves, drive changes in syntax between devices.  Even Cisco's attempt at a single command set, the Modular QoS CLI, has many subtle variations.  It gets even more challenging when other vendors are considered.

The only solution I see that is workable is to use the CLI syntax of each device type, just as network engineers have to do in the real world.  They already deal with the variations and understand them.  Why complicate their lives with yet another variation?

 -Terry

Comments

No Comments

About tslattery

Terry Slattery, CCIE #1026, is a senior network engineer with decades of experience in the internetworking industry. Prior to joining Chesapeake NetCraftsmen as a full time consultant, Terry was the founder and CTO of Netcordia, and inventor of NetMRI, a suite of network management products. Terry started Netcordia as a consulting company in 2000 and transitioned to a network management product company in 2003. During the consulting days, he used his network design and implementation skills to lead a team in the design and implementation of a high availability network at a brokerage clearing house. Terry is the former President and founder of Chesapeake Computer Consultants, Inc., a networking and computer systems training and consulting company. He co-invented and patented the vLab(tm) internet-based remote lab system. He is co-author of the McGraw Hill text Advanced IP Routing in Cisco Networks. Terry led the team that developed the current Cisco IOS user interface under contract to Cisco Systems. Terry is experienced in the design and installation of large TCP/IP based networks and is a successful network protocol instructor. He is the second Cisco Certified Internetworking Expert (CCIE) #1026 and the first outside of Cisco. He enjoys membership on the Vanderbilt University Engineering School’s Industrial Advisory Board and the IEEE.

This Blog

Syndication