Flash Tutorial on - Weather Forecasting for Cloud Computing - Moving "AAS-As A Service" Out of the Fog

The tutorial is available at: http://www.techtionary.com/ocsforum/cloudcomputing/
Thirty years ago it was distributed computing, then timesharing, and then a decade ago there were ASP-applications services providers. Today all these are now generally called "Cloud Computing" aks-also known as WaaS-Whatever as a Service like CAAS-Communications As A Service and others. It's the same old story of computing-centric strategy where the functionality is in the device versus communications-centric strategy where the functionality is in the network. I love my iPhone and love its ability to have a lot of movies, songs, TECHtionary animations, photos and nearly everything else I need to take with me. I also love iTunes where I can go buy nearly everything I need at one place. I really don't care what cloud or clouds iTunes is located in - here, there anywhere. How I am tethered can be many things - wired or wireless. How they aggregate all the songs, movies, content is not my concern. However, if I was the architect for iTunes I would be very concerned about every aspect of my cloud and all the clouds that connect together. In addition, I would be concerned about storms like tornados, hurricanes and all the types of weather conditions that can cause chaos. Planning and designing "cloud computing" networks is going to be more complex than the any of the vendors or providers will ever tell you. Before attempting to "forecast the weather" for cloud computing, let's review some of the key benefits otherwise we might as well stay in port.

Here are a few key points: - No CAPEX - no financing required - good benefit but like leasing long-term MRC-monthly recurring costs can end up costing more than the benefit. This is listed first but maybe the least valuable benefit. - Ondemand model - flat-rate per user pricing - its more than price its applications "by the slice" with any or all the "toppings" you want. - Scalable globally - allows distribution of labor costs to their lowest place-point. This becomes the most significant benefit as it gives corporations access to "priceless" labor costs - moving labor anywhere where the price is less. - Server Virtualization using Abstraction Architecture moving data to anypoint-anyone where - this is the most abstract concept as virtualization means so many different things.

In planning and designing cloud computing networks, we recommend taking a time-tested proven approach using the OSI-Open Systems Interconnection model. In each OSI layer provides unique functions:
- Divide issues into distinct activities
- Simplify processes
- Create the means to attempt standards
- Provide the means for interoperability between systems and services
- Support the idea that you don't have to do everything at once
- Allow for security and error checks at each layer

In other words, OSI keeps it simple. If you prefer to put in cloud computing terminology, it might look like this (see the animated tutorial for a pictorial):
- CAAS-Communications As A Service - User Application Level - OSI-Layer 7
- SAAS-Software As A Service - Development Level - OSI-Layer 6
- PAAS-Platform as A Service - Operating System Level - OSI-Layer 3-4-5
- IAAS-Infrastructure As A Service - Packet Level - OSI-Layer 2
- NAAS-Network As A Service - Device Level OSI-Layer 1

Without going on and on, the following section explains just some of the complexity for cloud computing. As any airplane pilot knows, stay out of the cloud and be careful where you fly.

The following details are provided in the detailed animation available at:
http://www.techtionary.com/ocsforum/cloudcomputing/

If you are still interested in cloud computing, the animated tutorial provides an introductory tutorial into many of the cloud computing concepts.

MSDP-Multicast Source Discovery Protocol is a process that connects PIM SP-Sparse Mode (pull down the data when asked for) domains and allows RP-Rendezvous Points or Shared Root to share information about active sources or shared trees. RP-Rendezvous Points in a rooted Shared Tree distribution system provide sub-optimal routing but are created as a result of user receiver densities such as rural areas or sub-continental geographic limitations. Content is distributed first to RP and then to receivers. RP-Rendezvous Points in a Bidirectional Tree distributes traffic content at any "point" in the network reducing data delays. SSM-Source Specific Multicast is a single-source-rooted distribution tree with no RP-Rendezvous Points using PIM-Protocol Independent Multicasting and IGMP-Internet Group Management Protocol. Receivers are explicitly-joined existing members which reduces complexity, billing, security, etc. MBGP-Multi-protocol Border Gateway Protocol is used to carry unicast prefixes of source networks across AS-Autonomous Systems boundaries. RP in different AS establishes peer arrangement to exchange MBPG updates called MSDP-Multicast Source Discovery Protocol SA-Source Address AM-Active Messages. AM are used to know about active sources in other AS-Autonomous Systems to build inter-domain (AS) Source trees. IP Anycast is defined, by Cisco, as the point-to-point flow of IP-Internet Protocol packets between a single client and the "nearest" destination server identified by any IP Anycast address. Criteria for selecting "nearest" server determine the type of anycast content. Also referred to a Virtual IP address with mirrored (duplicated or load-balanced disk) servers sharing same IP address. That is, the DNS-Domain Name Server maps (identifies) unique IP address which then distributes content to local or remote content servers. Some background on IP Anycast: Advantages: - IP routing will deliver packets over the shortest path to the nearest host - Client needs to be configured with only a single Anycast host address - Upon failure, routing will forward to the next nearest server with the same Anycast address - Distributed Anycast servers will effectively "load balance" traffic Addressing - Works with IPv4 and IPv6 Routing - Works with standard IP routing protocols - Changes in network topology (configuration) will automatically change to the nearest Anycast server - Route Summarization may increase Route Table size - IPv6 use can add QoS-Quality of Service (policy-based) routing
IPv4-to-IPv6 or Internet Protocol Version 4 to Version 6 also known as 6to4 Relay Router Anycast routers advertise anycast address prefix into IPv4 network where IPv4 packets are routed to nearest 6to4 relay router and then to IPv6 network. Mesh-First Tree Networks are created on the basis that a mesh (interconnected) network is formed BEFORE a routing protocol is determined. Then Source-Rooted Trees are selected based on the routing protocol such as DVMRP-Distance Vector Multicast Routing Protocol. DV-Distance Vector routing protocols maintain a routing table which consists of destination nodes, corresponding costs and next hops (routers). Routing Tables analyze the routes and determine the SPF-Shortest Path First. Then for multicast routing, RPF-Reverse Path Forwarding is used to determine SRT-Source Rooted Trees for content distribution. SPT-Shortest Path Tree is used by only ONE source (the root of the tree) to reach all receivers. Also known as a Directed Tree, packets only flow from the root (source) to the leaves (receivers). SFT content can also be distributed across different trees. ST-Shared Tree is where there are MANY sources to reach all receivers. That is, each tree (source) is also a root (receiver). This is also referred to as SCT-Shared Center Tree as the content is at the center of the distribution tree. There is research that suggests that Shared Trees have 25% less delay than SPT as content is closer to the receiver. Two other factors are important in a multicast network - Dense and Spare modes. DM-Dense Mode uses a "push" model to flood multicast traffic (broadcast) to every device. That is, every device gets the data whether they asked for it or not. SP-Sparse Mode uses "pull" or "pull down" model where the data is retrieved when asked for (on demand). SP allows RP-Rendezvous Points or Shared Root to share information about active sources or shared trees. Dense and Spare mode selection is also based on the amount and urgency (real-time events versus archival) of the content.

This tutorial is written and produced by TECHtionary.com (http://www.techtionary.com) and
OCSForum.com (http://www.ocsforum.com) sponsor of OCS Forum 2010 Expo - June 15-16 - Boulder, Colorado. OCS Forum is a vendor-independent laboratory environment designed for learning, technical guides, knowledge resources and online "live" services. OCS Forum provides planning, project management, consulting, training, case studies, white papers, speaking engagements, market/customer research, network planning and other services. For more go to http://www.ocsforum.com. Call Tom Cross 303-594-1694 or cross@gocross.com for more information and scheduling.