Multicast Data Distribution for Low Bandwidth Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems management solutions, such as Microsoft SMS, face challenges in efficiently distributing software and data to branch sites with low or intermittent bandwidth connections, leading to high network loads and performance degradation, especially when multiple machines attempt to download data simultaneously.
Innovation Solution
The implementation of data distribution apparatus that utilizes multicast technology, allowing multiple machines to act as sources for distributing data over a local network, with a selection mechanism to recover missing data from other machines, reducing the load on the central network and spreading the processing load across machines.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If multiple machines independently download data from a central site simultaneously, then each machine can obtain updates independently, but the network load on the larger network becomes excessively high
Solution Approach 1:
The patent segments the download process into two phases: a coordinated phase where one machine downloads data to local storage, and a distribution phase where multiple machines retrieve data locally. This segmentation reduces WAN traffic while maintaining independent update capability through local data sharing
Solution Approach 2:
The patent introduces an intermediary mechanism (coordination protocol and local storage system) that mediates between the central site and multiple machines. The intermediary coordinates download timing and manages local data distribution, reducing direct simultaneous connections to the central site
2Loss of energy
If one machine acts as a server to distribute updates locally, then network load on the larger network is reduced, but the performance of the server machine degrades significantly
Solution Approach 1:
The patent performs preliminary action by downloading and storing complete update data on the server machine before distribution begins. This allows the server to serve multiple clients simultaneously from local storage without real-time processing overhead, maintaining server performance while reducing network load
Solution Approach 2:
The patent uses copying by storing a local copy of the complete update dataset on the server machine. This copy serves as a local distribution source, eliminating the need for the server to process multiple real-time download requests and thus preserving server performance
3Loss of energy
If data is downloaded over a large network to a local machine, then central network load is reduced, but bandwidth limitations of the larger network still constrain the distribution speed
Solution Approach 1:
The patent segments the distribution architecture into a slow WAN link for initial data acquisition and a fast LAN for local distribution. This segmentation allows the bandwidth-constrained WAN to be used only once for data acquisition, while subsequent distribution leverages high-speed local network connections
Solution Approach 2:
The patent transitions from a single-dimension centralized distribution model to a two-dimensional model combining WAN and LAN networks. Data flows through the WAN dimension once for acquisition, then distributes through the LAN dimension for rapid local dissemination, overcoming WAN bandwidth limitations
Data Source
AI summary
Embodiments of the invention are concerned with data distribution apparatus, for use on each of a plurality of machines in distributing a set of data over a network amongst said machines. The apparatus comprises:i) multicast apparatus for use in multicasting said set of data over the network for receipt by others of said machines; andii) a data retrieval arrangement for use in retrieving data from a set of data previously multicast by at least one of said machines, and supplying retrieved data to the multicast apparatus for use in multicasting the retrieved data over the network to others of the plurality of machines,wherein more than one of said plurality of machines is enabled to act as a multicast source in distributing the set of data over the network.In embodiments of the invention, a first machine is elected as download master and multicasts downloaded data, via a group address, to the networked machines. Each machine receiving the multicast data monitors the data blocks received and, if a machine has received an incomplete set, for example because it came online late or as a result of unreliability, it can access a shared cache of another machine having a more complete set. Such a machine then starts a secondary multicast so that other machines can also complete their sets. Only one machine provides a secondary multicast at one time, to prevent high traffic in the network. While the download master is still downloading, the secondary multicast function is cycled around amongst the networked machines to avoid overload of any one machine.