Abstract: Techniques are provided for assigning workloads in a multi-node processing environment using resource allocation feedback from each node. One method comprises obtaining feedback from distributed nodes that process workloads, wherein the feedback for a given node indicates (i) an allocation of resources, and (ii) a number of executing workloads. In response to receiving a given workload to be processed, candidate nodes are identified to execute the given workload; and the given workload is assigned to a given candidate node based on an amount of available resources on each candidate node and/or a stability of resource adjustments made for each candidate node. The stability of the resource adjustments made for each candidate node can be evaluated based on a maximum resource adjustment made for a given candidate node relative to a maximum resource adjustment made for each of the candidate nodes.

Abstract: Techniques are provided for assigning workloads in a multi-node processing environment using resource allocation feedback from each node. One method comprises obtaining feedback from distributed nodes that process workloads, wherein the feedback for a given node indicates (i) an allocation of resources, and (ii) a number of executing workloads. In response to receiving a given workload to be processed, candidate nodes are identified to execute the given workload; and the given workload is assigned to a given candidate node based on an amount of available resources on each candidate node and/or a stability of resource adjustments made for each candidate node. The stability of the resource adjustments made for each candidate node can be evaluated based on a maximum resource adjustment made for a given candidate node relative to a maximum resource adjustment made for each of the candidate nodes.

Abstract: Techniques are provided for dataflow execution time estimation for distributed processing frameworks. An exemplary method comprises: obtaining an input dataset for a dataflow for execution; determining a substantially minimal data unit for a given operation of the dataflow processed by the given operation; estimating a number of rounds required to execute a number of data units in the input dataset using nodes assigned to execute the given operation; determining an execution time spent by the given operation to process one data unit; estimating the execution time for the given operation based on the execution time spent by the given operation to process one data unit and the number of rounds required to execute the number of data units in the input dataset; and executing the given operation with the input dataset. A persistent cost model is optionally employed to record the execution times of known dataflow operations.

Abstract: Methods and apparatus are provided for data operation pre-processing with a probabilistic estimation of operation value. An exemplary method comprises extracting feature values from a data set; identifying a set of operations that previously processed data sets comprising the extracted feature values; determining whether to execute an operation from the set of operations before an explicit request for the execution of the operation based on a probabilistic evaluation of a value of pre-processing the operation; and executing a set of instructions for the operation when it is determined that the operation is to be executed. The set of instructions comprises, e.g., stopping execution of other operations being executed, freeing resources required by the operation, and/or allocating resources required by the operation.

Abstract: Techniques are provided for dataflow execution time estimation for distributed processing frameworks. An exemplary method comprises: obtaining an input dataset for a dataflow for execution; determining a substantially minimal data unit for a given operation of the dataflow processed by the given operation; estimating a number of rounds required to execute a number of data units in the input dataset using nodes assigned to execute the given operation; determining an execution time spent by the given operation to process one data unit; estimating the execution time for the given operation based on the execution time spent by the given operation to process one data unit and the number of rounds required to execute the number of data units in the input dataset; and executing the given operation with the input dataset. A persistent cost model is optionally employed to record the execution times of known dataflow operations.

Abstract: Techniques are provided for automatic placement of cache operations in a dataflow. An exemplary method obtains a graph representation of a dataflow of operations; determines a number of executions and a computational cost of the operations, and a computational cost of a caching operation to cache a dataset generated by an operation; establishes a dataflow state structure recording values for properties of the dataflow operations for a number of variations of caching various dataflow operations; determines a cache gain factor for dataflow operations as an estimated reduction in the accumulated cost of the dataflow by caching an output dataset of a given operation; determines changes in the dataflow state structure by caching an output dataset of a different operation in the dataflow; and searches the dataflow state structures to determine the output datasets to cache based on a total dataflow execution cost.

Abstract: Lossless content-aware compression and decompression techniques are provided for floating point data, such as seismic data. A minimum-length compression technique exploits an association between an exponent and a length of the significand, which corresponds to the position of the least significant bit of the significand. A reduced number of bits from the significand can then be stored. A prediction method is also optionally previously applied, so that residual values with shorter lengths are compressed instead of the original values. An alignment compression technique exploits repetition patterns in the floating point numbers when they are aligned to the same exponent. Floating point numbers are then split into integral and fractional parts. The fractional part is separately encoded using a dictionary-based compression method, while the integral part is compressed using a delta-encoding method.

Abstract: Distributed content-aware compression and decompression techniques are provided for data, such as seismic data. Data is compressed by obtaining a file chunk of a plurality of file chunks of a larger file, a start offset of the file chunk, and a data descriptor indicating how the file chunk should be processed based on characteristics of the larger file. Headers in the file chunk are compressed using a substantially lossless header compression technique to obtain compressed headers. Samples in the file chunk are compressed using a substantially lossless sample compression technique to obtain compressed samples. The compressed headers and compressed samples are packed into a single bit stream comprising a compressed version of the file chunk. The compression can be performed in a physical or logical layer of storage nodes of a file system or in compute nodes of a computing cluster. The compression can be executed on demand by an external agent and/or in a background mode by a storage operating system.

Abstract: Data compression with window-based selection from multiple prediction functions is provided. A predefined default predictor and a plurality of other predictors are applied to a floating point number to generate a plurality of predictions. A compression metric over a collection of floating point numbers is evaluated for the default predictor and the plurality of other predictors. Based on the compression metric, (i) the floating point number is encoded using the predefined default predictor, or (ii) the collection of floating point numbers is encoded using one of the other predictors. Stored indexes indicate which predictor was used for the encoding. A set of predictors out of a larger set of predictors can be determined for a specific data set based on a performance-based ranking. The default predictor and the alternate predictors can be represented as ensembles of predictors.

Abstract: Lossless content-aware compression and decompression techniques are provided for floating point data, such as seismic data. A minimum-length compression technique exploits an association between an exponent and a length of the significand, which corresponds to the position of the least significant bit of the significand. A reduced number of bits from the significand can then be stored. A prediction method is also optionally previously applied, so that residual values with shorter lengths are compressed instead of the original values. An alignment compression technique exploits repetition patterns in the floating point numbers when they are aligned to the same exponent. Floating point numbers are then split into integral and fractional parts. The fractional part is separately encoded using a dictionary-based compression method, while the integral part is compressed using a delta-encoding method.