Inclusive caches are commonly used by processors to simplify cache coherence. However, the trade-off has been lower performance compared to non-inclusive and exclusive caches. Contrary to conventional wisdom, we show that the limited performance of inclusive caches is mostly due to inclusion victims—lines that are evicted from the core caches to satisfy the inclusion property—and not the reduced cache capacity of the hierarchy due to the duplication of data. These inclusion victims are incorrectly chosen for replacement because the last-level cache (LLC) is unaware of the temporal locality of lines in the core caches. We propose Temporal Locality Aware (TLA) cache management policies to allow an inclusive LLC to be aware of the temporal locality of lines in the core caches. We propose three TLA policies: Temporal Locality Hints (TLH), Early Core Invalidation (ECI), and Query Based Selection (QBS). All three policies improve inclusive cache performance without requiring any additional hardware structures. In fact, QBS performs similar to a non-inclusive cache hierarchy.

Networks are growing in size and complexity, resulting in increased alarm volume and number of unfamiliar alarms. Often, there is no proportional increase in monitoring personnel and response time to faults suffers. GTE deployed Telephone Operations Network Integrated Control System (TONICS) in 1993 to support its network management operations. To stay competitive in the face of continued staff reductions, increase in network size, and monitoring complications related to deregulation of the telephone industry, GTE is introducing artificial intelligence techniques into TONICS. Alarm Correlation Engine (ACE), the system described in this paper, is part of the effort. ACE aids network management by correlating alarms on the basis of common cause to provide alarm compression, filtering, and suppression. In conjunction with its ability to carry out prescribed responses, it improves response times and increases productivity. ACE was developed with the following requirements: reliability, speed, versatility (handle alarms from different switches and networks), ease of knowledge engineering (field technicians must be able to construct, test, and modify correlation patterns), handle in real time multiple network problems, and finally, interface smoothly with GTE's TONICS system. ACE's strength lies in its domain specific correlation language which facilitates knowledge engineering and in its asynchronous processing core that enables integration into a real-time monitoring system.