Data Driven Fluid Data Center

software defined network-1Data Centers have changed tremendously since my days at Exodus Communications, and even from the time of my stint at NetScaler (Citrix) and Ankeena (Juniper):

  • Cloud services (SaaS/PaaS/IaaS) have become mainstream
  • Network as a service (NWaaS) is catching on fast (eg., WebSense et al.)
  • Utilization for COTS (common off the shelf) componentry is becoming the norm
  • We have also seen the data center architecture evolve away from monolithic systems to virtualized stacks. Initially the compute stack led the way; then we saw the advent of virtualized storage to serve virtual machines, and now we are witnessing the dawn of truly virtualized data center networks.

    Software-defined Networking (SDN) has joined the fray to further disaggregate data center architectures. The “software-defined” trend has also spread farther, with terms like “software defined storage” and “Software-defined Data Center” starting to find increasing usage.In this scenario, the data center architecture has become a hotbed of change:

  • Evolving from fixed to converged i.e., multi-tenanted
  • Storage and data networks have converged (Converged network adaptor)
  • Huge increase in number nodes due to virtualization
  • Tremendous increase in spikes and east-west traffic due to consumer demand, componentization, and big-data workloads.
  • Such a dynamic data center requires a near-real-time “closed loop” approach to maximize efficiency and assure a great user experience. However the state of the art is not even close:
  • Reactive, non-real time manual operation
  • Driving blind: operational issues are outrunning ability to monitor/diagnose
  • Isolated islands of operations across compute, storage and networks;
  • Massive over-provisioning across the board – negating the very motivation for creating a fluid converged data center in the first place!
  • Is there any doubt that the new data center architecture needs a new solution for monitoring, troubleshooting and automating, datacenter operations?

    I think that the desired automation solution needs to be able to collect data from a variety of data center infrastructure every few seconds, make near real time decisions and automate dynamic changes in settings on data center elements. You could think of this as a key element of a software defined data center. The industry appears to be headed towards such architecture. Elements of these are already in place, especially in compute. Other aspects of converged data center automation are along the way.

    Therefore my strongly felt conclusion is: “Data-driven Fluid Data Center” is a matter of when, not if!

    Interested in hearing more about Data-driven Fluid Data Center? Please check out the upcoming event hosted by The Fabric.

/by