Infrastructure and operations (I&O) teams are aligning themselves and infrastructure around key workloads to drive greater simplicity and efficiency. In kind, the networking industry has responded by suggesting that networks can provide greater support for this approach using the OpenFlow protocol and Software Defined Networking (SDN) concepts.
I believe the SDN definition today equates to adding closed loop functionality so a network can intelligently orchestrate a set of services. This architecture consists of three components: an automation and orchestration controller, a monitoring and data collection system, and a configuration and management system.
SDN provides the means to automate networks to better support different workloads, but I&O professionals also need to understand how SDN can support turning networks into a virtual network infrastructure.
Whatever technology options networking professionals choose, the value cannot be extracted without preparation. Creating a workload centric infrastructure to serve the business requires the infrastructure to become standardised, self-service and pay-per-use, giving users rapid access to powerful and more flexible IT capabilities.
This, in turn, means I&O teams need to coordinate infrastructure elements, such as switches, firewalls, load balancers and optimisers, to deliver the right set of services to the right user, at the right time and at the right location. Workload centric networks will reconfigure these elements on the fly and monitor the output to ensure that the newly created services are within the bounds of the business policies and rules.
Our recommendations to I&O leaders for what SDN requires follow.
Standardise your process, procedures, roles and responsibilities
I&O need to have a baseline in order to automate infrastructures. With this in place, you
can then make changes to optimise the workload performance and user experience. In our network
assessment engagements with clients, we find that this is consistently an underdeveloped area. A
scant few have standards documents for their config files, products, firmware or architecture. You
need to start on the refining processes to assess the current state of key operational and process
activities and then standardise processes and skills around ITIL.
Invest in tools that empower other I&O teams to utilise the network
Network teams are already stretched too thin to be responsible for every networking decision.
For example, there’s little value in the network team handing out IP addresses every time new apps
are loaded. Advanced DNS, DHCP and IPAM tools providing workflow processes for a set of IP
addresses can be assigned to a server team, who then can grab them as needed. This eliminates waste
and repetitious activities which take away from focusing on higher technical
skills like using SDK kits to hook orchestration systems into network management
Add a software network engineer position
With today's mounting business requirements and escalating technology complexity, manual control
is dead. With all of the variability in users, devices and services, plug-and-play operating
systems must supersede command line interfaces in the network. Network software engineers can help
create these operating systems with C+, Software Development Kits (SDK) and application programming
interfaces (APIs) such as OpenFlow, using them to fuse distributed systems, virtualisation, data
storage etc. This enables developers to build network applications that can integrate and interact
with networking gear by manipulating switch tables or using Layer 2 and Layer 3 protocols such as
Link Aggregation Control Protocol (LACP), spanning tree protocol (STP), Rapid STP, virtual
redundant routing protocol (VRRP) and 802.1x. Infrastructure monitoring is absolutely critical to
support this closed-loop system.
Push the networking teams to start managing the virtual world
Workload centric infrastructure support requires that both physical and virtual infrastructures
work together. Forrester has found that few networking organisations manage virtual switches,
firewalls, application load balancers or wan optimisers. Virtual networking is an extension of the
physical world; all the concepts remain the same. To help your teams overcome the fear of the
unknown, transition the management of virtual switches from hypervisor administrators to networking
personnel. They can choose between managing either the hypervisor virtual switch or a networking
vswitch, like Cisco’s 1000v, IBM’s 5000v or an OpenvSwitch, or the networking administrator can
manage the hypervisor virtual switch.
Deploy more monitoring tools
After standardisation and any automation or adoption of workload centric infrastructure requires visibility to the workloads and processes so the system knows what happens when adds, moves and deletions are made. This is a fundamental requirement in closed loops systems. Monitoring tools and solutions need to move from a ‘nice to have’ to a ‘must have’ before you can derive value from virtual network infrastructures supporting workloads. I&O should be asking themselves how they will be monitor each app, software, and hardware deployment.
Andre Kindness (pictured) is senior analyst at Forrester Research where he serves IT infrastructure and operations professionals. He is a leading expert on network operations and architecture.
This was first published in August 2012