The challenge facing many businesses today is how to move from an essentially physical environment to one that is highly virtualized and that enables agile, flexible service orchestration while maintaining and enhancing robust, always-on service availability.
The Software-Defined Data Center: Creating DC-as-a-Service
1. The challenge facing many businesses today is how to
move from an essentially physical environment to one
that is highly virtualized and that enables agile, flexible
service orchestration while maintaining and enhancing
robust, always-on service availability.
Rise of the Virtual Machines
Traditionally, applications were monolithic and were delivered via monolithic
platforms in a series of physical provisioning activities including: procurement
and implementation of an actual server and physical network connections,
installation of operating system and application software, initialization of
physically attached disk subsystems and deployment, connection and
configuration of firewall and load-balancing appliances.
If this all sounds tiresome and slow, that’s because it is. These activities can
consume weeks, even months.
In addition to poor time to service and a clear lack of agility are the high cost of
dedicated hardware and the low level of efficiency it delivers. Rarely achieved,
effective dimensioning of actual hardware components so that they match
application requirements is something of a black art. And then there is the
question of life cycle scale; scaling up or down in an overwhelmingly physical
environment is difficult, costly, and it opens the door to all the risks associated
with human-error-prone manual configuration.
As applications move from the monolithic model to a composite architecture
with elements dispersed across multiple computing platforms, a corresponding
evolution is taking place in compute and storage solutions. The increasing
sophistication of server virtualization solutions (referred to generically as the
hypervisor) along with ever more powerful multi-core x86 compute hardware
are delivering a versatile, powerful, robust solution that can support multiple
virtual machines with each machine offering tenanted hosting to unique
instances of operating system and application software.
At a time of dramatic change in the ways that compute and storage are
dimensioned, implemented, and provisioned in the Data Center, the dynamic,
real-time power of virtual machines is being curtailed by the coordination and
configuration requirement of the network.
avaya.com | 1
The Software-Defined
Data Center: Creating
DC-as-a-Service
Table of Contents
Rise of the Virtual Machines.... 1
The Software-Defined
Data Center strikes back.......... 2
Lifecycle of
Transient Services.......................4
Never send a human to
do a machine’s job.....................4
Fabric Connect: Avaya’s
Software-Definable
Network Fabric............................ 5
A new hope................................... 5
2. avaya.com | 2
Avaya addressed this gap with the introduction of its Fabric Connect
technology. A crucial element of Avaya’s Virtual Enterprise Network
Architecture (VENA) strategy for enabling next-generation networking, Fabric
Connect empowers agile service delivery by optimizing the way that networks
are deployed, implemented, operated, and maintained. Liberated from
topology constraints, Fabric Connect supports versatile placement of
networking components and interconnections, and sets a new standard for
service flexibility.
Fabric Connect features an edge-only provisioning model, empowering service
activation without time-consuming or service-effecting change control thanks
to the abstraction of user services from the network core. This ensures full
compartmentalization of the failure domain and therefore the risk. The business
benefit is pronounced, with service additions and changes enacted in real-time.
Conceptually inserting a ‘virtual Ethernet’ between physical topology and
network service layers, Fabric Connect empowers flexible end-to-end
connectivity.
Abstracting the services from the constraints and limitation of the traditional
two-dimensional design model, that of protocol-applied-to-physical, liberates
the network and empowers a radical shift in network design and service
delivery. At Layer 2, the predominate, although not exclusive requirement
within the Data Center, VLANs (or unique hosts) are simply mapped to the
required ‘Service ID’; this occurs only at the Fabric Connect edge, reducing
time-to-service and the burden and error-prone requirement of end-to-end,
device-by-device, link-by-link configuration. And because the underlying
technology is natively extensible, Fabric Connect integrates support for Layer 3
in the same way; VRFs are mapped to appropriate Service IDs and end-to-end
connectivity is delivered immediately.
Indeed, Avaya is able to optimize the provision of Layer 3 routing functionality
and distribute multiple mutually cooperating gateways throughout the Fabric,
as and where these are most effective. This is particularly relevant in a
distributed, physically dispersed Data Center model.
With the availability of the elements required for a truly end-to-end solution,
highly virtualized compute and storage, and – for the first time – a networking
infrastructure that features real-time configuration propagation, all that’s
needed to deliver an autonomic solution for the Data Center is a common
orchestration framework.
The Software-Defined Data Center strikes back
Having embraced server virtualization as a strategy for the long-term and now
having a network that supports the clear and quantifiable abstraction of
services from infrastructure (or, control plane from data plane, if you will)
business can move forward and create solutions that actually deliver on the
promises of consolidation: agility, performance, resilience, efficiency, and
automation.
3. avaya.com | 3
The breakthrough capability of integrating the many and varied components
of service delivery combined with their universal orchestration is central to
the value proposition of the Software-Defined Data Center (SDDC). When
activating a new service or modifying an existing one, the traditional approach
required considerable planning, time consuming coordination and service-
affecting, error-prone configuration – all of which had a negative impact on
agility and time to service.
The SDDC changes the game fundamentally. By totally transforming the
relationships between service components (from that of silos to that of peers)
and by inserting an intelligent real-time middleware, it enables delivery of a
single operational event that utilizes a single administrative interface. And,
leveraging the holistic abstraction of functionality from the underlying
infrastructure, multiple processes are coordinated automatically.
All foundational components remain. Obviously we still provision virtual
machines, server adapters, storage partitions, network appliances, and
interconnect them appropriately. What’s changed is that now we make obsolete
the burden of a series of independent and vaguely associated provisioning
tasks. Orchestration becomes integrated and seamless and the core element
that enables this is the OpenStack open source cloud operating system.
OpenStack’s modular, project-based approach ensures that all aspects of
service activation and delivery are available through a series of integrated
interfaces. For example, the OpenStack ‘Nova’ interface orchestrates virtual
machine resources, the ‘Cinder’ and ‘Swift’ interfaces orchestrate block and
object storage respectively and ‘Quantum’ interconnects everything by
coordinating network connectivity, services, and appliances. Typically these
actions are brought together under the umbrella of OpenStack’s ‘Horizon’
graphical orchestration interface although it’s also probable that major
infrastructure players will wish to express themselves by integrating a similar
capability into their own platforms.
Regardless of the ultimate flavor, this orchestration platform provides key
functionality: one end-to-end view of the service, one execution engine
synchronizing activity, and one point of reference for lifecycle administration.
Individual business applications are visualized in the graphical environment;
components are dragged and dropped together to form the end-to-end
solution and provisioning is initiated via a single user interaction.
Enhancing existing proven infrastructure components with the incremental
addition of an OpenStack capability significantly reduces the costs and risks
associated with a transition to the SDDC. The open sourced framework delivers
those benefits usually associated with proprietary solutions – namely high levels
of integration and functionality – in addition to the cost-competitiveness and
best-of-breed promises of multi-vendor sourcing. Operators can chose their
own level of automation and extend this by simply leveraging incremental
advances in OpenStack interface functionality, introducing additional
OpenStack-compliant products to the solution or even developing custom-
built, mission-specific OpenStack capabilities.
Use-Case Example: A
financial services clearing
house that provides hosted
IT services for hundreds
of regional banking
organizations that support
hundreds of thousands of
employees and millions
of customers.
Coming from a legacy
position that relied upon
connecting together
physical components, the
ability to scale and react
was always limited but,
in a time when server
deployment took weeks of
planning and provisioning,
the gap in coordination
with network and storage
meant little. Now, in the era
when virtual machines can
be spun-up from bare
metal within minutes, the
status quo could not stand.
In their next-generation
Data Center, all constituent
components that form an
‘application’ are drawn
dynamically from resource
pools then combined,
optimized, and
operationalized in real
time and managed from
a single pane-of-glass. This
is progress defined.
4. avaya.com | 4
Never send a human to do a machine’s job
The network must be prepared to take on a new and more pivotal role, one that
necessitates a far higher degree of intelligence, integration, and automation
than ever imagined. In a service model that may operationalize for a few days or
hours, the network’s traditional change implementation methodology – that
required weeks or months – simply would not work. This is where Avaya’s
dynamic, real-time, service-orientated Fabric Connect technology comes into
its own. Based on the Shortest Path Bridging protocol jointly standardized by
the IEEE and the IETF, Fabric Connect delivers the industry’s most software-
definable networking capability. Empowered by edge-only provisioning, Fabric
Connect features seamless orchestration, a full breadth of integrated services –
Layer 2, Layer 3, and IP Multicast – and is natively architected for multi-tenant
operations.
Focusing solely on the interconnectivity required for highly virtualized
application services, it quickly becomes obvious that Fabric Connect possesses
all the attributes necessary to facilitate the solution. A relatively new concept,
that of ‘virtual wires’, can be dynamically spun-up in order to interconnect any
two or more service components -- seamlessly, instantaneously, and
automatically.
Fabric Connect natively supports a ‘VLAN attach’ method where the Edge node
maps an 802.1Q VLAN tag to an 802.1aq Individual Service ID (or I-SID) and,
while mapping is conventionally achieved through manual configuration, it can
also be orchestrated and automated. A feature of Fabric Connect’s underlying
architecture – the Shortest Path Bridging protocol – is high extensibility and
Avaya is developing the capability to leverage the 802.1AB standard for Link
Layer Discovery Protocol (LLDP) to enable end-points (in the SDDC scenario
these would be represented by Virtual Switches within the hypervisor
infrastructure) to make a standardized LLDP request for the network to auto-
provision specific virtual wires (i.e. network connectivity) as an orchestrated
function of the service delivery process.
In addition to leveraging and enhancing existing VLAN methods, Fabric
Connect also supports the seamless integration of VMware’s VXLAN technique
(or similar IP Multicast-based overlay technologies) of utilizing Internet Group
Management Protocol (IGMP) join requests in order to map service end-points
to network virtual wires. Because Fabric Connect supports a fully integrated IP
Multicast capacity without the need for any additional overlay or overhead
complexity, Avaya can uniquely empower VXLAN-based orchestration, VLAN-
based orchestration, or a hybrid combination of the two.
Leveraging proven commodity techniques such as LLDP and IGMP helps ensure
that development can be accelerated and deployment free of risk. Policy-based
managed access can be delivered using technologies such as the 802.1X
Extensible Authentication Protocol (EAP) and/or RADIUS and confidentiality
can be assured using the MD5 message-digest algorithm.
Lifecycle of
Transient Services
The following scenario outlines
the flow of orchestration and
automation delivered by the
Software-Defined Data Center
solution. Compare and contrast
the single interaction – on a single
console – by a single operator
with the cross-silo burden of the
planning, coordination, and
configuration required by today’s
disaggregated approach.
To support a new or expanding
business application, service
resources need to be spun-up.
1. Leveraging the SDDC
orchestration platform,
an Operator selects the
appropriate compute,
memory, and storage profile
from pre-defined
standard service options
2. Wizards guide the Operator
through the process of
combining the
necessary service resource
components
3. Once the end-to-end service
has been graphically created,
provisioning can be executed
in real-time or scheduled
4. OpenStack interfaces
propagate provisioning
instructions to the relevant
resource controllers,
operationalizing individual
components
5. As required, temporary virtual
wires are established between
the virtual machine and
software distribution servers.
In the case of a bare metal
machine, an additional phase
would see virtual wires
temporarily connect
to an OS imaging server
6. The fully imaged virtual
machine is spun-up and
operationalized
7. Layer 2 and/or Layer 3 virtual
wires deliver optimized private
and/or shared connectivity
between virtual adaptors,
storage arrays, end-user
networks, and network
services and appliances such
as Firewalls or ADCs
5. avaya.com | 5
In addition to providing this bottom-up mechanism to empower service delivery
orchestration to pull network connectivity resources on an as-required basis,
Avaya is also delivering a top-down push capability in the form of a ‘Quantum’-
enabled SDN Controller. This Controller will seamlessly integrate with
northbound OpenStack orchestration platforms and southbound with Avaya
Networking platforms. Its role is to deliver incremental networking services
such as Layer 3 routing optimization that may not necessarily be on the critical
path of service creation and delivery but would enhance specific or generic
application performance and robustness.
The abstraction of services from topology enables Fabric Connect to transform
the network into a resource pool that allows arbitrary assignment and
deployment of Layer 2 or Layer 3 virtual wires, the corresponding network
addressing, Layer 3 route engines, and other virtualized networking services
(e.g. firewalls, load-balancers, and application delivery controllers). Gone is
the concept of complex and time-consuming pre-planning and manual
provisioning; Fabric Connect delivers the industry’s first autonomic capability
that seamlessly meshes with service delivery orchestration.
A new hope
What businesses seek are solutions that reduce burden and increase agility.
Orchestration, automation, abstraction of the network control function from the
network infrastructure and integration with feature-rich middleware are the
fundamental elements required to meet these business needs.
Far from being driven by dumbed-down generic silicon that sports an esoteric
programmable interface, the future of software defined networking will
gravitate toward those implementations that can deliver the most mainstream
business value. It is difficult to visualize the benefit to the enterprise of having
to learn how to instruct a networking platform on how to bridge, switch, and
route – especially in light of the fact that these particular wheels were invented
decades ago.
Fabric Connect is part of the VENA framework of next-generation networking
solutions and with it Avaya is uniquely positioned to support businesses as they
venture on to greater service agility and reduced operational burden. For
example, Fabric Connect is based on a fully standardized technology and can
interoperate with the massive installed-base of existing Ethernet products,
thereby reducing exposure and costs. It can be deployed harmoniously
alongside legacy technologies – running as ‘ships-in-the-night’ – and services
can be selectively and progressively migrated to avoid any big bang risks. It also
fully supports traditional VLAN connectivity attachment techniques in addition
to emerging alternatives such as VXLAN – even offering integrated hybrid
connectivity between both schemes for maximum flexibility.
Fabric Connect:
Avaya’s Software-
Definable Network
Fabric
The characteristics that have
made Avaya VENA’s Fabric
Connect the technology of
choice for end-to-end, next-
generation networking are the
same ones that elevate it to a
unique position in terms of the
Software-Defined Data Center.
Quite simply, no other existing or
proposed technology comes
close to the levels of abstraction,
orchestration, and automation
that Fabric Connect is delivering
today as more and more
businesses implement this
genuinely evolutionary
technology.
Fabric Connect’s checklist of
services and functionality shows
how different it is:
• Layer 2, or VLAN, extensions;
‘L2 Virtual Service Networks’
• Layer 3, or VRF, extensions; ‘L3
Virtual Service Networks’
• Inter-VSN Routing; native
interworking of independent
VSNs
• IP Shortcut Routing; native
internetworking of
independent end-points
• Integrated IP Multicast;
seamless coalescence of
Unicast and Multicast flows
• Abstracted, highly scalable,
service-orientated provisioning
model
• Real-time propagation of
provisioning control
• Highly extensible TLV-based
architecture
• Support for VLAN, VXLAN, or
hybrid mapping of end-points
to virtual wires
• Ready integration with third-
party orchestration and
automation platforms
• And, crucially, a readily
software-definable
architecture