Pruning can be used in Convolutional Neural Networks(CNNs) to reduce the consumption of computations. In the iteration of pruning, a fixed ratio of filters or weights are pruned after evaluating their importance according to a certain criterion, which has the risk of leading to mis-pruning when the distribution of the importance is imbalanced. We propose a strategy to assist pruning for CNNs at filter level. Firstly, in order to evaluate importance and non-importance about filters in the network, we design two fuzzy membership functions respectively, then we evaluate all filters by their membership function value, finally, α-cut set is employed to determine the ones to be pruned . Experimental results demonstrate the effectiveness of our strategy, since it can obtain more compression than the popular pruning algorithms at similar accuracy level.

We propose DotSlash, a self-configuring and scalable rescue system, for handling web hotspots at dynamic content Web sites. To support load migration for dynamic content, an origin Web server sets up needed rescue servers drafted from other Web sites on the fly, and those rescue servers retrieve the scripts dynamically from the origin Web server, cache the scripts locally, and access the corresponding database server directly. We have implemented a prototype of DotSlash for the LAMP configuration, and tested our implementation using the RUBBoS bulletin board benchmark. Experiments show that by using DotSlash a dynamic content web site can completely remove its web server bottleneck, and can support a request rate constrained only by the capacity of its database server.

No abstract available.

The paper presents two new mechanisms, global attributes and preference filters, that improve the efficiency and extendability of the service location protocol (SLP). Global attributes describe common service properties, which can be used to perform queries across different service types, and can be standardized for service management. Preference filters facilitate processing of search results, such as finding the best match, in SLP servers. They can reduce the amount of data transferred to the client, and better support thin clients with limited resources and capabilities.

No abstract available.

The service location protocol (SLP), a proposed IETF standard, provides a flexible and scalable service discovery framework for IP networks. It can be deployed with or without a directory service. This paper presents mSLP, the mesh-enhanced service location protocol. mSLP enhances SLP with a fully-meshed peer directory agent (DA) architecture. Peer DAs exchange service registration information, and maintain the same consistent data for shared scopes. mSLP improves the reliability and consistency of SLP directory services. It also greatly simplifies service agent (SA) registrations in systems with multiple DAs. mSLP is backward compatible with SLPv2 and can be deployed incrementally.