The
Platform-as-a-Service (PaaS) layer is becoming the most strategic and
innovative part of the Cloud computing stack. Large data centers that are used
by cloud infrastructure providers such as Amazon are becoming more numerous and
cloud capacity is being built up in every large city of the world. Cloud
infrastructure is becoming less centralized and more distributed on a regional
basis. This new distributed cloud model applies to private clouds, public
clouds, or a hybrid of the two that includes cloud bursting and brokering
capabilities. The PaaS layer provides the glue or federation for the cloud as
application components are distributed across different cloud infrastructure. A
messaging framework known as a service bus enables the application components
to communicate with each other. In the new distributed cloud model the WAN or
Internet is a critical piece of cloud infrastructure that has previously been
“assumed” to be over-provisioned and always available, reliable, and secure. Just as the Internet has transformed the
world as a global network of networks, the “InterCloud” is now evolving as a
“world of many clouds”. In other words, a federation of many clouds that will
be transformational for the next generation of distributed SaaS applications
for cloud services. Technologies such as OpenFlow and Software Defined
Networking (SDN) hold the promise of enabling a new control plane for the Wide
Area Network (WAN). The Federated PaaS will become the next generation
Operational Support System (OSS) to orchestrate a distributed mesh of federated
cloud nodes for cloud scale and high availability. The automated policy system
of the PaaS will respond to events as an OSS and then make changes to the flows
of cloud applications across the Internet to ensure an exceptional cloud user
experience.
The notion of
“federation” is an evolution of Grid and Mesh Computing. A grid architecture is
a computational network infrastructure based on a cooperative use of the
different computing resources connected by the Internet. Mesh networks have
also evolved with grid computing to help connect distributed nodes and enable
automatic reconfiguration when faults occur, broken connections happen, or
nodes disappear. Cloud infrastructure has enabled applications to operate on
the lowest-cost servers and scale up or down additional compute power when
needed. Cloud application developers will still have more specialized
requirements for some of their application components that may require
specialized infrastructure for CPU intensive operations or for greater performance
to reduce latency in user response times.
With cloud providers building data centers in all the major cities
across the world, cloud computing is becoming less centralized and more
regionally available. Cloud infrastructure in the InterCloud model will be
defined to any place you can find compute, storage, and a network: in a central
data center, a regional data center, in future routers/switches in telecom
network, and in mobile devices such as your cell phone or PC. In the future
this will even be in cloud-connected automobiles.
To connect the world
of the InterCloud a “Federated PaaS Model” will be required. This is one of the
three models that CloudAve contributor Krishnan Subramanian discusses
as a trend in the enterprise PaaS space. He distinguishes three models of
service delivery: the
Heroku Model, the Amazon Model, and the Federated PaaS Model. New Federated
PaaS systems will emerge that can enable distributed cloud applications to be placed into a
federated mesh architecture across many different clouds using an automated
policy system. Automated policies will determine how the distributed cloud
scales, how live-live copies of app components will be replicated across
multiple cloud locations for high availability, and multiple layers of policy
will check messages between app components for security and compliance. A federated cloud cloud will understand the location of the user through the GPS on their phone
and uses their location as input to a load balancing algorithm, and like an
amoeba it will shape the geographic distribution of the cloud to respond to the
need for more resource or better performance. The Federated PaaS becomes the
foundation for the next generation of SaaS mobile cloud services. The
intelligence baked into the automated policy system of the Federated PaaS can
move application components and their complementary storage fabric closer to
the user for lower latency, better response times, and improved customer
experience. This not only is cached content as found in the Akamai model, but
could include rendering algorithms or analytics. The PaaS layer can also
respond to events and enable dynamic changes to the cloud to protect the cloud
user experience. This could include the ability to scale up (provision)
additional compute or storage resources to respond to load. In the future the
PaaS will serve as an Operational Support System (OSS) to make adjustments to
the “flows” of cloud services across the Internet.
The last missing
component to this evolution of the Intercloud is the network resource component
of the Internet. In the data center, server capacity was being over-provisioned
for peak loads. Cloud computing solved that problem with cloud scaling. Today the
WAN connection and Internet network pipes are still being overprovisioned to
handle peak traffic loads. Service providers over-provision their network
capacity for unpredictable spikes in traffic loads. This is the next problem
for cloud computing to solve. The Federated PaaS will become part of the next
generation of Operational Support Systems (OSS) to not only federate
application nodes across the Internet, but serve as a control plane for the
end-to-end network connections (flows) between federated cloud nodes and to
mobile end users of cloud services. In other words the Platform-as-a-Service
(PaaS) layer of cloud computing will understand the application requirements
for cloud services and provide additional control over the wide area network (WAN)
connection between federated data centers, to branch offices, and to connected
mobile users. The Federated PaaS system will be the control plane for the
InterCloud.
In the OpenFlow model, the Federated PaaS will
become what is known as a “controller” for critical WAN network points. These control points will typically be
network entry points known as ingress or egress points. These can be at the edge of the data center
where new federated nodes are created in the InterCloud. They can also be at
the other end of the connection at the edge of the last mile of user
connectivity. Behind the mobile cell tower base stations, at aggregation points
for fixed high speed broadband or where enterprise branch office connections
enter the network. In Software
Defined Networks which use the OpenFlow protocol, the controller interacts
with an OpenFlow-enabled switch or router to identify packets that are
associated with a “flow” (a connection) and perform operations on those
packets. An OpenFlow operation may be to change the destination (IP Address of
destination app server) of the flow or to reprioritize the TOS bits to give the
flow higher priority in the processing queues of edge routers. OpenFlow can
also be used to configure a L3 tunnel or GRE tunnel and then direct packets
into the tunnel. The automated policy system of a federated PaaS will scale out
and replicate application nodes across the InterCloud. When the PaaS provisions
a host in a cloud provider as a new federated node, it will understand the
functional requirements of the application component in the cloud node
(storage, analytics, processing, ingest) and the connection (WAN) requirements.
The PaaS will create a new cloud node, add it to the federated cloud, and then
use OpenFLow to configure the connection (or flow) properties for that node.
This could include building a secure tunnel for the cloud services to flow
through. The PaaS as an Operational Support System (OSS) can also monitor the
cloud and ensure that those connections are operating within the thresholds
required for an exceptional user experience. If the WAN connection is not
meeting the needs of the cloud application, the PaaS will be able to use
OpenFlow to modify a flow at a critical point in the network either by changing
its path or increasing the cloud flow’s packet priority. Another option is that
the Federated PaaS may determine the cloud node is not in an optimal location
and clone a copy in a different cloud somewhere else, begin using that app component in the
federated cloud mesh, then kill (scale down) the first node that is not
performing well. The automated policy system of the PaaS will be a critical
Operational Support System (OSS) and foundational layer of the cloud stack to
enable the cloud to reconfigure and relocate to ensure a secure, reliable,
available, and responsive user experience for cloud services.
The next generation of
cloud Software-as-a-Service (SaaS) applications will operate over a world of
many clouds. Cloud SaaS applications will become more distributed as in the
Service-Oriented Architecture (SOA) model to take advantage of the world of many
clouds (the InterCloud). The Federated PaaS layer will sit underneath the
distributed SaaS application in the cloud stack to ensure cloud scale, fault
tolerance and high availability, and to manage secure and reliable network
connections (cloud flow management). The Federated PaaS layer will become the
control plane for software defined networks and leverage the OpenFlow protocol
as an enabling technology for the next generation network. The automated policy
system of the PaaS will orchestrate the federation of distributed cloud nodes,
including the management of cloud flows across the network. The PaaS as an OSS will monitor and respond
to events such as threshold crossing alarms to make adjustments to cloud flows
across the network or even relocate cloud nodes to locations with better
connections to protect the end user experience of future cloud SaaS
applications.
No comments:
Post a Comment