Tuesday, October 16, 2012
Networking from A to PI
Often with technology it is difficult to quickly identify the root cause of a problem. Complex systems present multiple possible points of failure, especially when they have no single point of failure - a strong argument for the KISS principle.
In enterprise IT environments, this challenge is compounded by the complexity of the IT organization. Multiple silos of expertise and responsibility lend themselves to a problem triage approach whereby each constituent group in the IT organization separately goes through an internal assessment of a technical issue. Frequently this results in all groups saying "it's not me" when asked to identify the source of a problem - and no immediate isolation of root cause or consensus definition of a likely solution.
I like to describe this challenge with an old joke:
There's a problem with the performance of a mission critical application. A huge number of resources assemble - server and storage administrators, application developers, DBAs, QA and Performance Test teams, business analysts, virtualization administrators, line of business executives, program and project managers. An auditorium is used as the "war room", and this huge group rapidly identifies the likely cause of the issue. It's the networking team - they weren't in the room.
Whatever humor that joke has stems from two things - human nature and an unspoken IT cultural divide that places networking outside of other technology disciplines. Networking has turned into an essential, "always on" service that has different characteristics from other IT disciplines and service offerings. It is a unique culture with strong barriers to entry. As implied in the phrase "ping, power and pipe" It's the air we breathe, it's the dial tone on our phone...
Dial tone is not as good an analogy as it used to be, and old conceptions about networking are being challenged as well. Software Defined Networking (SDN) is headed rapidly towards the peak of its hype cycle (or is well past it, depending on who you are talking to). It promises to deliver the same agility, CapEx and OpEx reductions that companies have attained with servers leveraging VMware (the Band-Aid of hypervisors).
Also like VMware, SDN allows for abstraction of the underlying hardware components and supports automation and orchestration via APIs. The benefits to this are clear - agile, nondisruptive configuration of network resources with the ability to provision in seconds rather than days or weeks. No more waiting for someone to assign an IP address from a spreadsheet-managed "pool". New ability to tie network management and monitoring together with application and other infrastructure components, presenting a "single pane of glass" for all and an immediate and automated root cause identification should problems arise. An end to finger pointing - Nirvana!
While the Promised Land has not yet arrived for most enterprises, the allure has been driving significant interest, media coverage and market transformation. VMware's acquisition of Nicira in July 2012 for 1.26B USD is one example of this. Multiple standards still vie for acceptance and the traditional networking vendors are placing their bets.
The end state goals, if not the winning path to those goals, are clear - simplify, automate and orchestrate networking. Move to policy-based management. Find a way to abstract networking infrastructure in the same way that VMware abstracts server infrastructure - as well as storage infrastructure via VAAI and VASA APIs. In doing these things, cut costs and eliminate provisioning delays. Make it easier to dynamically move workloads across geographies and into and out of the cloud. Make networking just another pool of resources that can be managed and monitored with compute and storage through a single dashboard.
Does all of this sound as attainable as world peace? Perhaps it would if we weren't in the midst of rapid cloud computing adoption, or hadn't already seen this happen with server virtualization. The technology, while new to market on the whole, is already commercially available and rapidly expanding in capabilities.
At some point, perhaps not too far off, a gathering of all critical IT resources to identify the root cause of an issue may become a SWAT team of one cloud administrator.