neomantra

High Performance Redis

In response to this brief blog entry, @antirez tweeted for some documentation on high-performance techniques for Redis. What I present here are general high-performance computing (HPC) techniques. The examples are oriented to Redis. but they work well for any program designed to be single- or worker-threaded and asynchronous (e.g. uses epoll).

The motivation for using these techniques is to maximize performance of our system and services. By isolating work, controlling memory, and other tuning, you can achieve significant reduction in latency and increase in throughput.

My perspective comes from the microcosm of my own bare-metal (vs VM), on-premises deployment. It might not be suitable for all scenarios, especially cloud deployments, as I have little experience with HPC there. After some discussion, maybe this can be adapted as [redis.io documentation](https://redis.io/do

neomantra

#!/bin/bash
# Copyright 2013-2019 Neomantra BV. All rights reserved.
# Released under the MIT License.

usage()
{
    cat >&2 <<EOF
usage:  $0 [-h] [-v] [-c cpulist] ppid1 [ppid2 [...]]

    Given a list of parent process IDs, this script finds their child

neomantra

# Builds lua_ipython_kernel on centos

ARG IMAGE_BASE="centos"
ARG IMAGE_TAG="7"

FROM ${IMAGE_BASE}:${IMAGE_TAG}

RUN yum install -y \
        epel-release \
    && yum install -y \

neomantra

# Dockerfile for Aeron C++ SDK
#
# Copyright (c) 2017 Neomantra BV
# Released under the MIT License, see LICENSE.txt

FROM debian:stretch-slim

MAINTAINER Evan Wies <[email protected]>

ARG AERON_VERSION="master"

neomantra

/// cc -o bson_float_test bson_float_test.c  -lbson-1.0.0                                                                                    

#include <stdio.h>
#include <libbson-1.0/bson.h>

int main(int argc, const char* argv[])
{
    bson_t b;
    bson_init(&b);

neomantra

#include <iostream>
#include <boost/fusion/adapted/std_tuple.hpp>
#include <boost/fusion/algorithm/iteration/for_each.hpp>

struct One {
    int value() const { return 1+rand(); }
};

struct Two {
    int value() const { return 2+rand(); }

neomantra

#include <iostream>
#include <boost/fusion/adapted/std_tuple.hpp>
#include <boost/fusion/algorithm/iteration/for_each.hpp>

struct One {
    int value() const { return 1+rand(); }
};

struct Two {
    int value() const { return 2+rand(); }

neomantra

local ffi = require("ffi")
local  new = ffi.new

--[[
All numeric calculations are performed with doubles. Using
floats for storage saves memory (for big arrays). But arithmetic
is usually slower due to the required float<->double conversions.
]]
local Matrix
local MatrixStructure = ffi.typeof("double[16]")

neomantra

-- Converts a LuaJIT bytecode number to name
--
-- usage:   luajit  bcname.lua bytecode_number [bytecode_number2 [...]]
--
-- example:
--    $ luajit-2.1 bcname.lua 71 72
--      VARG
--      ISNEXT
--
-- From:  http://www.freelists.org/post/luajit/frames-and-tail-calls,1

neomantra

$ luajit ~/Desktop/serialize_metamethod.lua 
2 =%= 3
cdata<struct 98>: 0x0004d738