This document discusses just-in-time IT infrastructure provisioning using commodity resource pools. It advocates allocating resources on demand to accomplish production tasks by leveraging massive investments in compute, storage, and network infrastructure from various service providers. This approach aims to reduce costs and spread risks across production participants by only paying for consumed resources. The document also describes how abstraction of infrastructure through virtualization and programmatic interfaces can facilitate heterogeneous, multi-vendor provisioning driven by workflow tools through a centralized exchange point.
2. Key Characteristics of JIT Provisioning
• Allocation of resources as needed to accomplish a
production task.
• Utilizing commodity resource pools to fulfill production
demands.
• Pre-defined methods of integrating different resource
pools together to appear as one production
architecture.
• Programmatically triggered upon demand or policy.
3. Why JIT Provisioning
• The economics of commodity resource pools are
too disruptive to ignore.
• M&E Productions need to effectively leverage
massive Service Provider Capital Investments
across compute, storage, and network.
• Production costs and risks should be spread
amongst production participants as resources are
consumed.
4. Reduce Costs: Abstract, then
Commoditize
• Server
• Storage
• Circuit
Physical
• Virtual Server
• Volumes on
NAS and SAN
• Virtual
Circuits
Virtual
• PaaS
• NFV
• SaaS
• IaaS
• Object Stores
Abstracted
OS Console Frameworks API
Efficiency Pressure
5. Cloud Disruption:
Abstract, then Commoditize
• The Economics of Virtualization are
inescapable – cost management and
flexibility requirements demand
more granular controls.
• The development of flexible
Provisioning Frameworks have
spawned a new age of multi-vendor,
heterogeneous infrastructure to
expose these controls.
• Management systems have become
increasingly more complex and
capable, allowing for policy
application across heterogeneous
infrastructure by using these
frameworks.
• The search for new, abstracted
Control Points has become critical to
these new Management Systems.
Nascent:
Network
(SDN and
NFV)
Mature:
Storage
(SaaS)
Mature:
Compute
(IaaS)
Mature: Apps
(PaaS)
6. Leverage Wireline Infrastructure
Investment
$25
$26
$27
$28
$29
2010 2011 2012 2013
US Wireline Capex 2010-2013
Spend in Blns
Source: US Telecom
The wireline portion of the $1.3
trillion in broadband provider capital
expenditures from 1996 through
2013 was 52 percent. The wireline
segment continued to contribute a
significant portion of industry capital
in 2013, 37 percent of total spend.
7. Network Commoditization =
Opportunity
• Metro Ethernet capacity glut will require new
provisioning strategies to achieve scale
• Excess long-haul capacity as CSP Intra-regional
replication reduces utilization
• SDN Automation is a must to reduce opex and
attract Provisioning Engine Integration
8. Today: Compute and Storage
Infrastructure Provisioning
Abstracted
• Manual Processes and Extended
Timeframes required to provision
core network and some CSP
infrastructure.
• Workflow Integration to
Management Frameworks
beginning, but missing network
resource allocation capabilities.
• Workflow systems operate on
statically defined topologies.
9. Next: Full Infrastructure
Provisioning Abstraction
• Manual Processes and Extended
Timeframes eliminated for
Orchestrated API calls.
• Workflow Integration to
Frameworks allows users to
provision ALL resource pool
infrastructure – Servers, Storage,
Network.
• Workflow systems operate
dynamically to infrastructure.
10. Abstraction drives Evolution of Tools
• User View: Workflow tools now become mechanism to provision infrastructure
– Example Tools: 5th Kind, Front Porch Digital
• Administrator View: Resource Management and Workload Migration tools
– Example Provisioning Engines: Migration Accelerator for Azure, Docker for GCP, Cisco
Intercloud, AWS Console VM Import/Export, VMware vRealize, Right Scale, HotLink,
OpenStack
• PaaS Frameworks allow for Applications to be readily provisioned in either private
or public infrastructure as needed:
– Example PaaS Frameworks: Azure Cloud Services, Google App Engines, AWS Elastic Beanstalk,
Apprenda, AppFog,
• Network Provisioning becomes virtualized and programmatically provisioned:
– Examples of SDN Frameworks: Equinix EPN, Cisco Intercloud
11. Abstracted tools need Exchange Points
• Need a centralized, neutral provisioning point to
apply policy and signal topology changes.
• This ‘Exchange Point’ becomes a ‘provisioning
hub’ for all physical and virtual infrastructure in
the topology.
• This point needs to access all resource pools via
high-speed, low latency interconnection
methods.
12. Public Cloud Private Cloud
Workflow
Production Partners
Remote
Production
Partners
On-Location Media On-
Ramp
SDN Control
Central Storage and Policy
13. Workflow-Driven Infrastructure
Pre-Production
Bid Package
Award Shot to
VFX House
Award Trailer to Vendor
Production
Post-Production
RAW
Capture
Meta Data
Video Capture Dirty Pass
Validate
Color / Meta Data
Clean Pass
Trailer Cut
Ingested
VFX Shots
Status Updates
Compare Cut to VFX Status to
catch any cost impacts
Editorial
VFX HouseTrailer House
Media & Entertainment Exchange
On Demand Software Defined Networking
14. In Summary
• Leverage Commodity Resource Pools – let someone
else do long-term capacity planning
• Provision via abstracted interfaces to allow for vendor
diversity
• Select providers based upon neutrality and pre-
integration to tools and frameworks
• Exchange providers should provide programmatic, low
latency, high bandwidth access to majority of required
resource pools