Big Model analytics tackles the training of massive models that go beyond the available memory of a single computing device, e.g., CPU or GPU. It generalizes Big Data analytics which is targeted at how to train memory-resident models over out-of-memory training data. In this paper, we propose an in-database solution for Big Model analytics. We identify dot-product as the primary operation for training generalized linear models and introduce the first array-relation dot-product join database operator between a set of sparse arrays and a dense relation. This is a constrained formulation of the extensively studied sparse matrix vector multiplication (SpMV) kernel. The paramount challenge in designing the dot-product join operator is how to optimally schedule access to the dense relation based on the non-contiguous entries in the sparse arrays. We propose a practical solution characterized by two technical contributions---dynamic batch processing and array reordering. We devise three heuristics -- LSH, Radix, and K-center -- for array reordering and analyze them thoroughly. We execute extensive experiments over synthetic and real data that confirm the minimal overhead the operator incurs when sufficient memory is available and the graceful degradation it suffers as memory becomes scarce. Moreover, dot-product join achieves an order of magnitude reduction in execution time over alternative solutions.

Existing data analytics systems have approached predictive model training exclusively from a data-parallel perspective. Data examples are partitioned to multiple workers and training is executed concurrently over different partitions, under various synchronization policies that emphasize speedup or convergence. Since models with millions and even billions of features become increasingly common nowadays, model management becomes an equally important task for effective training. In this paper, we present a general framework for parallelizing stochastic optimization algorithms over massive models that cannot fit in memory. We extend the lock-free HOGWILD!-family of algorithms to disk-resident models by vertically partitioning the model offline and asynchronously updating the resulting partitions online. Unlike HOGWILD!, concurrent requests to the common model are minimized by a preemptive push-based sharing mechanism that reduces the number of disk accesses. Experimental results on real and synthetic datasets show that the proposed framework achieves improved convergence over HOGWILD! and is the only solution scalable to massive models.

Online aggregation provides continuous estimates to the final result of a computation during the actual processing. The user can stop the computation as soon as the estimate is accurate enough, typically early in the execution, or can let the processing terminate and obtain the exact result. In this demonstration, we introduce a general framework for parallel online aggregation in which estimation does not incur overhead on top of the actual processing. We define a generic interface to express any estimation model that abstracts completely the execution details. We design multiple sampling-based estimators suited for parallel online aggregation and implement them inside the framework. Demonstration participants are shown how estimates to general SQL aggregation queries over terabytes of TPC-H data are generated during the entire processing. Due to parallel execution, the estimate converges to the correct result in a matter of seconds even for the most difficult queries. The behavior of the estimators is evaluated under different operating regimes of the distributed cluster used in the demonstration.

We present GLADE, a scalable distributed system for large scale data analytics. GLADE takes analytical functions expressed through the User-Defined Aggregate (UDA) interface and executes them efficiently on the input data. The entire computation is encapsulated in a single class which requires the definition of four methods. The runtime takes the user code and executes it right near the data by taking full advantage of the parallelism available inside a single machine as well as across a cluster of computing nodes.