Reliable Transport Connection Cache Eviction for Data Center Networking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Reliable transport protocols require large memory for connection information, which when cached, can lead to slow transactions and wasted space due to fetching data and keeping inactive connections in the cache.
Innovation Solution
Implement a cache eviction system that prioritizes connection entries based on their likelihood of receiving transactions, using resource count and recency, and evicts connections with the lowest priority when the cache occupancy exceeds a threshold, utilizing a priority queue system and 1r2w memory for efficient management.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If connection information is stored in cache to improve transaction speed, then transaction latency is reduced, but cache memory is wasted when connections are inactive
Solution Approach 1:
The patent implements dynamic cache eviction by continuously monitoring connection activity and adjusting cache contents accordingly. The system evicts inactive connection entries from cache and reloads them only when needed, transforming the static cache into a dynamic structure that adapts to changing connection patterns. This resolves the contradiction by maintaining fast access for active connections while freeing cache memory for inactive ones.
Solution Approach 2:
The system changes the state parameter of connection entries by tracking their activity status and using this information to determine eviction candidates. By monitoring whether connections are active or inactive and adjusting cache retention based on this parameter change, the system optimizes both transaction speed and memory utilization.
2Reliability
If connection information is fetched from memory to cache when requests are received, then cache hit rate improves, but transaction latency increases due to fetching overhead
Solution Approach 1:
The system performs preliminary actions by pre-loading connection information into cache before it is actually needed. By anticipating which connections will be active and loading their information in advance, the system reduces the frequency of cache misses and the associated fetching overhead, thereby improving both cache hit rate and transaction latency.
Solution Approach 2:
The patent implements mechanisms to skip the cache fetching step entirely for connections that are already in cache. By maintaining a high cache hit rate through intelligent eviction and pre-loading, the system allows most transactions to proceed directly from cache without the time-consuming memory fetch operation, effectively rushing through the transaction process.
3Speed
If all connection information is kept in cache to ensure fast access, then transaction speed is maximized, but cache occupancy exceeds available memory capacity
Solution Approach 1:
The system extracts only the necessary subset of connection information from the total set of connections and keeps it in cache. By identifying and retaining only active or frequently accessed connection entries while evicting inactive ones, the system maintains fast transaction speed for relevant connections without exceeding cache memory capacity.
Solution Approach 2:
The patent applies local quality by providing different cache retention treatments to different connection entries based on their individual activity characteristics. Rather than uniformly caching all connections, the system selectively maintains high-priority active connections in cache while allowing low-priority inactive connections to reside in slower memory, optimizing both speed and memory utilization.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Aspects of the disclosure are directed to high performance connection cache eviction for reliable transport protocols in data center networking. Connection priorities for connection entries are determined to store the connection entries in a cache based on their connection priority. During cache eviction, the connection entries with a lowest connection priority are evicted from the cache. Cache eviction can be achieved with low latency at a high rate.