Apache Spark (Español)

Apache Spark es un framework de computación en clúster de código abierto desarrollado originalmente en el AMPLab de la UC Berkeley. En contraste con el paradigma MapReduce basado en el disco de dos etapas de Hadoop, las primitivas en memoria de Spark ofrecen un rendimiento hasta 100 veces mayor para ciertas aplicaciones. Al permitir que los programas del usuario carguen datos en la memoria de un clúster y lo consulten repetidamente, Spark está bien adaptado a los algoritmos de aprendizaje automático.

Esta traducción de Apache Spark fue revisada el 2019-02-09. Si existen cambios puede actualizarla o avisar al equipo de traducción.

Instalación

Instale el paquete apache-sparkAUR.

Configuración

Algunas variables de entorno se establecen en /etc/profile.d/apache-spark.sh.

ENV Valor Descripción
PATH $PATH:/opt/apache-spark/bin Spark binaries

Es posible que deba ajustar la variable de entorno PATH si su shell inhibe /etc/profile.d:

export PATH=$PATH:/opt/apache-spark/bin

Habilitar el soporte de R

El paquete R de sparkR se distribuye con el paquete pero no se compila durante la instalación. Para conectarse a Spark desde R, primero debe compilar el paquete ejecutando

# $SPARK_HOME/R/install-dev.sh

como se describe en . También puede desear compilar la documentación del paquete siguiendo las instrucciones en . Una vez que se haya compilado el paquete sparkR R, puede conectarse utilizando .

gollark: Specifically, 22 bytes for the private key and 21 for the public key on ccecc.py and 25 and 32 on the actual ingame one.
gollark: <@!206233133228490752> Sorry to bother you, but keypairs generated by `ccecc.py` and the ECC library in use in potatOS appear to have different-length private and public keys, which is a problem.EDIT: okay, apparently it's because I've been accidentally using a *different* ECC thing from SMT or something, and it has these parameters instead:```---- Elliptic Curve Arithmetic---- About the Curve Itself-- Field Size: 192 bits-- Field Modulus (p): 65533 * 2^176 + 3-- Equation: x^2 + y^2 = 1 + 108 * x^2 * y^2-- Parameters: Edwards Curve with c = 1, and d = 108-- Curve Order (n): 4 * 1569203598118192102418711808268118358122924911136798015831-- Cofactor (h): 4-- Generator Order (q): 1569203598118192102418711808268118358122924911136798015831---- About the Curve's Security-- Current best attack security: 94.822 bits (Pollard's Rho)-- Rho Security: log2(0.884 * sqrt(q)) = 94.822-- Transfer Security? Yes: p ~= q; k > 20-- Field Discriminant Security? Yes: t = 67602300638727286331433024168; s = 2^2; |D| = 5134296629560551493299993292204775496868940529592107064435 > 2^100-- Rigidity? A little, the parameters are somewhat small.-- XZ/YZ Ladder Security? No: Single coordinate ladders are insecure, so they can't be used.-- Small Subgroup Security? Yes: Secret keys are calculated modulo 4q.-- Invalid Curve Security? Yes: Any point to be multiplied is checked beforehand.-- Invalid Curve Twist Security? No: The curve is not protected against single coordinate ladder attacks, so don't use them.-- Completeness? Yes: The curve is an Edwards Curve with non-square d and square a, so the curve is complete.-- Indistinguishability? No: The curve does not support indistinguishability maps.```so I might just have to ship *two* versions to keep compatibility with old signatures.
gollark: > 2. precompilation to lua bytecode and compressionThis was considered, but the furthest I went was having some programs compressed on disk.
gollark: > 1. multiple layers of sandboxing (a "system" layer that implements a few things, a "features" layer that implements most of potatOS's inter-sandboxing API and some features, a "process manager" layer which has inter-process separation and ways for processes to communicate, and a "BIOS" layer that implements features like PotatoBIOS)Seems impractical, although it probably *could* fix a lot of problems
gollark: There's a list.
This article is issued from Archlinux. The text is licensed under Creative Commons - Attribution - Sharealike. Additional terms may apply for the media files.