Distributed Cloud Cache for VM Image Provisioning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In cloud computing environments, the provisioning of virtual machines is hindered by the large size of virtual appliance (VA) image files, which require significant bandwidth for transfer, leading to degraded server provisioning speeds due to geographically dispersed hosting centers and centralized image repositories.
Innovation Solution
A distributed, cooperative cache is formed across cloud hosts, where template images and patches are stored and shared, with a control routine coordinating caching activities to service requests from local resources whenever possible, and periodically updated to minimize total cost using a cost function.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a centralized image repository is used to store template images, then image management is simplified, but network bandwidth consumption increases and provisioning speed decreases
Solution Approach 1:
The patent divides the centralized image repository into multiple distributed cache nodes deployed across different geographic locations. Each cache node stores copies of template images and patches locally, segmenting the monolithic repository into distributed storage units. This reduces network bandwidth consumption by serving images from local caches rather than always fetching from the centralized repository, while maintaining simplified management through coordinated cache invalidation and updates.
Solution Approach 2:
The patent introduces cache nodes as intermediary components between the centralized image repository and cloud hosts. These intermediaries store and serve template images and patches locally, reducing direct traffic to the centralized repository. The cache nodes act as mediators that can satisfy requests locally when possible, thereby reducing network bandwidth consumption while the centralized repository maintains overall image management control.
2Productivity
If template images are transferred from centralized repository to cloud hosts, then provisioning can be performed, but provisioning time increases due to large file sizes
Solution Approach 1:
The patent pre-loads template images and patches into distributed cache nodes before they are needed for provisioning. The cache nodes proactively fetch and store commonly requested images and patches in advance, so when provisioning requests arrive, the images are already available locally. This preliminary action eliminates the need to transfer large files during the provisioning process itself, significantly reducing provisioning time while maintaining full provisioning capability.
Solution Approach 2:
The patent creates and distributes copies of template images and patches to multiple cache nodes across the cloud infrastructure. Instead of transferring large image files from the centralized repository to each cloud host during provisioning, the system uses local copies stored in distributed caches. This copying approach allows rapid local serving of images, reducing provisioning time from minutes to seconds while maintaining the ability to provision any required image type.
3Ease of operation
If cloud hosts access the centralized image repository over the Internet, then image retrieval is possible, but server provisioning speed degrades significantly
Solution Approach 1:
The patent implements local cache nodes at each cloud host or in close proximity to multiple hosts, creating geographically distributed storage points. Each cache node serves its local neighborhood of cloud hosts, eliminating the need for remote Internet access to the centralized repository. This local quality approach ensures that image retrieval occurs over fast local network connections rather than slower Internet connections, dramatically improving provisioning speed while maintaining easy image retrieval capability.
Solution Approach 2:
The patent transitions from a single-dimensional centralized repository architecture to a multi-dimensional distributed cache network. Instead of all hosts accessing a single central point, the system creates multiple access points distributed across different geographic dimensions. This dimensional change allows hosts to access images from the nearest local cache node, reducing network latency and improving provisioning speed while maintaining Internet-accessible retrieval capability through the distributed network.
Data Source
AI summary
A method of provisioning in a cloud compute environment having a set of cloud hosts associated with one another. The method begins by forming a distributed, cooperative cache across the set of cloud hosts by declaring a portion of a data store associated with a cloud host as a cache, and storing template images and patches in the cache. Caching activity across the distributed, cooperated cache is coordinated by having the caches share information about their respective contents. A control routine at a cache receives requests for template images or patches, responds to the requests if the requested artifacts are available or, upon a cache miss, forwards the request to another one of the caches. Periodically, the composition of the distributed, cooperative cache is computed, and the template images and patches are populated into the caches using the computed cache composition.