This paper presents a state-based scheduling algorithm for the selection of the next logical process (LP) to be run on a processor in a Time Warp synchronized parallel discrete event simulation. In our solution, state information related to the LPs in the immediate predecessor set of a given LP is used to compute its scheduling priority. This distances our algorithm from previous solutions wher...

In the last years, many database-as-a-service (DBaaS) systems have started to offer their functionalities through REST APIs. Examples are record stores like DynamoDB and Azure Tables, object stores such as Amazon S3 as well as many NoSQL database systems, for instance Riak, CouchDB and ElasticSearch. Yet today, there has been no systematic effort on deriving a unified REST interface which takes...

This work studies the interaction of Web proxy cache coherency and replacement policies using trace-driven simulations. We speciically examine the relative importance of each type of policy in aaecting the overall costs, the potential of incorporating coherency issues in cache replacement and the inclusion of additional factors such as frequency of resource use in replacement and coherency poli...

The bulk synchronous parallel (BSP) model used by synchronous graph processing systems allows algorithms to be easily implemented and reasoned about. However, BSP can suffer from poor performance due to stale messages and frequent global synchronization barriers. Asynchronous computation models have been proposed to alleviate these overheads but existing asynchronous systems that implement such...

As Internet video goes mainstream, we see increasing user expectations for higher video quality and new global policy requirements for content providers. Inspired by the case for centralizing network-layer control, we present C3, a control system for optimizing Internet video delivery. The design of C3 addresses key challenges in ensuring scalability and tackling data plane heterogeneity. First...

Due to high data volumes and unpredictable arrival rates, continuous query systems processing expensive queries in real-time may fail to keep up with the input data streams resulting in buffer overflow and uncontrolled data loss. We explore join direction adaptation (JDA) to tackle CPU-limited processing of multijoin stream queries. The existing JDA solutions allocate the scarce CPU resources t...

OBJECTIVE To examine the effect of branding, as indicated by brand name, on evaluation of the cigarette smoking experience. DESIGN Between-subjects and within-subjects experimental study. Participants were randomly allocated to smoke a cigarette from a pack featuring a premium brand name and a cigarette from a pack featuring a value brand name. Within each condition, participants unknowingly ...

Traditional decision tree classifiers work with the data whose values are known and precise. We can also extend those classifiers to handle data with uncertain information. Value uncertainty arises in many applications during the data collection process. Example sources of uncertainty measurement/quantization errors, data staleness, and multiple repeated measurements. Rather than abstracting un...

Stochastic gradient MCMC (SG-MCMC) has played an important role in largescale Bayesian learning, with well-developed theoretical convergence properties. In such applications of SG-MCMC, it is becoming increasingly popular to employ distributed systems, where stochastic gradients are computed based on some outdated parameters, yielding what are termed stale gradients. While stale gradients could...

Deep learning with a large number of parameters requires distributed training, where model accuracy and runtime are two important factors to be considered. However, there has been no systematic study of the tradeoff between these two factors during the model training process. This paper presents Rudra, a parameter server based distributed computing framework tuned for training large-scale deep ...