Deep learning methods are transforming research, enabling new techniques, and ultimately leading to new discoveries. As the demand for more capable AI models continues to grow, we are now entering an era of Trillion Parameter Models (TPM), or models with more than a trillion parameters---such as Huawei's PanGu-Σ. We describe a vision for the ecosystem of TPM users and providers that caters to the specific needs of the scientific community. We then outline the significant technical challenges and open problems in system design for serving TPMs to enable scientific research and discovery. Specifically, we describe the requirements of a comprehensive software stack and interfaces to support the diverse and flexible requirements of researchers.

Pooling and sharing data increases and distributes its value. But since data cannot be revoked once shared, scenarios that require controlled release of data for regulatory, privacy, and legal reasons default to not sharing. Because selectively controlling what data to release is difficult, the few data-sharing consortia that exist are often built around data-sharing agreements resulting from long and tedious one-off negotiations. We introduce Data Station, a data escrow designed to enable the formation of data-sharing consortia. Data owners share data with the escrow knowing it will not be released without their consent. Data users delegate their computation to the escrow. The data escrow relies on delegated computation to execute queries without releasing the data first. Data Station leverages hardware enclaves to generate trust among participants, and exploits the centralization of data and computation to generate an audit log. We evaluate Data Station on machine learning and data-sharing applications while running on an untrusted intermediary. In addition to important qualitative advantages, we show that Data Station: i) outperforms federated learning baselines in accuracy and runtime for the machine learning application; ii) is orders of magnitude faster than alternative secure data-sharing frameworks; and iii) introduces small overhead on the critical path.

Prior work suggests that users conceptualize the organization of personal collections of digital files through the lens of similarity. However, it is unclear to what degree similar files are actually located near one another (e.g., in the same directory) in actual file collections, or whether leveraging file similarity can improve information retrieval and organization for disorganized collections of files. To this end, we conducted an online study combining automated analysis of 50 Google Drive and Dropbox users' cloud accounts with a survey asking about pairs of files from those accounts. We found that many files located in different parts of file hierarchies were similar in how they were perceived by participants, as well as in their algorithmically extractable features. Participants often wished to co-manage similar files (e.g., deleting one file implied deleting the other file) even if they were far apart in the file hierarchy. To further understand this relationship, we built regression models, finding several algorithmically extractable file features to be predictive of human perceptions of file similarity and desired file co-management. Our findings pave the way for leveraging file similarity to automatically recommend access, move, or delete operations based on users' prior interactions with similar files.

Exponentially increasing data volumes, coupled with new modes of analysis have created significant new opportunities for data scientists. However, the stochastic nature of many data science techniques results in tradeoffs between costs and accuracy. For example, machine learning algorithms can be trained iteratively and indefinitely with diminishing returns in terms of accuracy. In this paper we explore the cost-accuracy tradeoff through three representative examples: we vary the number of models in an ensemble, the number of epochs used to train a machine learning model, and the amount of data used to train a machine learning model. We highlight the feasibility and benefits of being able to measure, quantify, and predict cost accuracy tradeoffs by demonstrating the presence and usability of these tradeoffs in two different case studies.

Scientists' capacity to make use of existing data is predicated on their ability to find and understand those data. While significant progress has been made with respect to data publication, and indeed one can point to a number of well organized and highly utilized data repositories, there remain many such repositories in which archived data are poorly described and thus impossible to use. We present Skluma---an automated system designed to process vast amounts of data and extract deeply embedded metadata, latent topics, relationships between data, and contextual metadata derived from related documents. We show that Skluma can be used to organize and index a large climate data collection that totals more than 500GB of data in over a half-million files.

Distributed communities of researchers rely increasingly on valuable, proprietary, or sensitive datasets. Given the growth of such data, especially in fields new to data-driven research like the social sciences and humanities, coupled with what are often strict and complex data-use agreements, many research communities now require methods that allow secure, scalable and cost-effective storage and analysis. Here we present Cloud Kotta: a cloud-based data management and analytics framework. Cloud Kotta delivers an end-to-end solution for coordinating secure access to large datasets, and an execution model that provides both automated infrastructure scaling and support for executing analytics near to the data. Cloud Kotta implements a fine-grained security model ensuring that only authorized users may access, analyze, and download protected data. It also implements automated methods for acquiring and configuring low-cost storage and compute resources as they are needed. We present the architecture and implementation of Cloud Kotta and demonstrate the advantages it provides in terms of increased performance and flexibility. We show that Cloud Kotta's elastic provisioning model can reduce costs by up to 16x when compared with statically provisioned models.

Big data workflows often require the assembly and exchange of complex, multi-element datasets. For example, in biomedical applications, the input to an analytic pipeline can be a dataset consisting thousands of images and genome sequences assembled from diverse repositories, requiring a description of the contents of the dataset in a concise and unambiguous form. Typical approaches to creating datasets for big data workflows assume that all data reside in a single location, requiring costly data marshaling and permitting errors of omission and commission because dataset members are not explicitly specified. We address these issues by proposing simple methods and tools for assembling, sharing, and analyzing large and complex datasets that scientists can easily integrate into their daily workflows. These tools combine a simple and robust method for describing data collections (BDBags), data descriptions (Research Objects), and simple persistent identifiers (Minids) to create a powerful ecosystem of tools and services for big data analysis and sharing. We present these tools and use biomedical case studies to illustrate their use for the rapid assembly, sharing, and analysis of large datasets.

In the effort to reach beyond 3G, researchers have been actively looking at utilizing new models for network based services. Small mobile, pervasive and ubiquitous devices will benefit from networked services and computation provided by utility computing providers and the virtual organizations that lease resources from them. As an additional factor, we believe that it is critical that the mobile, pervasive or ubiquitous devices be able to dynamically manipulate their resource specifications when obtaining services and resources from the utility computing and communication network. This requires a simple, manipulatable, and preferably modular resource specification structure. This paper presents the Resource Description Graph (RDG). The RDG is used to represent available and required resources for hosts and applications in a directed acyclic graph. The RDG has many desirable properties including inherent security, expressiveness, modularity, and composition. We show that the computational time to match RDG resource specifications, thirty resource types and constraints, is less than 1ms --- demonstrating that the RDG is a practical approach to resource specification with a low computational overhead.