Added kernel tuning docs (#3921)

docs/deployment/kernel-tuning/README.md

@@ -0,0 +1,71 @@
+# Kernel Tuning for Minio Production Deployment on Linux Servers [![Slack](https://slack.minio.io/slack?type=svg)](https://slack.minio.io) [![Go Report Card](https://goreportcard.com/badge/minio/minio)](https://goreportcard.com/report/minio/minio) [![Docker Pulls](https://img.shields.io/docker/pulls/minio/minio.svg?maxAge=604800)](https://hub.docker.com/r/minio/minio/) [![codecov](https://codecov.io/gh/minio/minio/branch/master/graph/badge.svg)](https://codecov.io/gh/minio/minio)
+
+## Tuning Network Parameters
+
+Following network parameter settings can help ensure optimal Minio server performance on production workloads.
+
+- *`tcp_fin_timeout`* : A socket left in memory takes approximately 1.5Kb of memory. It makes sense to close the unused sockets preemptively to ensure no memory leakage. This way, even if a peer doesn't close the socket due to some reason, the system itself closes it after a timeout. `tcp_fin_timeout` variable defines this timeout and tells kernel how long to keep sockets in the state FIN-WAIT-2. We recommend setting it to 30. You can set it as shown below
+
+```sh
+`sysctl -w net.ipv4.tcp_fin_timeout=30`
+```
+
+- *`tcp_keepalive_probes`* : This variable defines the number of unacknowledged probes to be sent before considering a connection dead. You can set it as shown below
+
+```sh
+sysctl -w net.ipv4.tcp_keepalive_probes=5
+```
+
+- *`wmem_max`*: This parameter sets the max OS send buffer size for all types of connections.
+
+```sh
+sysctl -w net.core.wmem_max=540000
+```
+
+- *`rmem_max`*: This parameter sets the max OS receive buffer size for all types of connections.
+
+```sh
+sysctl -w net.core.rmem_max=540000
+```
+
+## Tuning Virtual Memory
+
+Recommended virtual memory settings are as follows.
+
+- *`swappiness`* : This parameter controls the relative weight given to swapping out runtime memory, as opposed to dropping pages from the system page cache. It takes values from 0 to 100, both inclusive. We recommend setting it to 10.
+
+```sh
+sysctl -w vm.swappiness=10
+```
+
+- *`dirty_background_ratio`*: This is the percentage of system memory that can be filled with `dirty` pages, i.e. memory pages that still need to be written to disk. We recommend writing the data to the disk as soon as possible. To do this, set the `dirty_background_ratio` to 1.
+
+```sh
+sysctl -w vm.dirty_background_ratio=1
+```
+
+- *`dirty_ratio`*: This defines is the absolute maximum amount of system memory that can be filled with dirty pages before everything must get committed to disk.
+
+```sh
+sysctl -w vm.dirty_ratio=5
+```
+
+## Tuning Scheduler
+
+Proper scheduler configuration makes sure Minio process gets adequate CPU time. Here are the recommended scheduler settings
+
+- *`sched_min_granularity_ns`*: This parameter decides the minimum time a task will be be allowed to run on CPU before being pre-empted out. We recommend setting it to 10ms.
+
+```sh
+sysctl -w kernel.sched_min_granularity_ns=10
+```
+
+- *`sched_wakeup_granularity_ns`*: Lowering this parameter improves wake-up latency and throughput for latency critical tasks, particularly when a short duty cycle load component must compete with CPU bound components.
+
+```sh
+sysctl -w kernel.sched_wakeup_granularity_ns=15
+```
+
+## Tuning Disks
+
+The recommendations for disk tuning are conveniently packaged in a well commented [shell script](https://github.com/minio/minio/blob/master/docs/deployment/kernel-tuning/disk-tuning.sh). Please review the shell script for our recommendations.

docs/deployment/kernel-tuning/disk-tuning.sh

@@ -0,0 +1,48 @@
+#!/bin/bash
+
+## Minio Cloud Storage, (C) 2017 Minio, Inc.
+##
+## Licensed under the Apache License, Version 2.0 (the "License");
+## you may not use this file except in compliance with the License.
+## You may obtain a copy of the License at
+##
+##     http://www.apache.org/licenses/LICENSE-2.0
+##
+## Unless required by applicable law or agreed to in writing, software
+## distributed under the License is distributed on an "AS IS" BASIS,
+## WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+## See the License for the specific language governing permissions and
+## limitations under the License.
+
+for i in $(ls -d /sys/block/*/queue/iosched 2>/dev/null); do
+    iosched_dir=$(echo $i | awk '/iosched/ {print $1}')
+    [ -z $iosched_dir ] && {
+        continue
+    }
+    ## Change each disk ioscheduler to be "deadline"
+    ## Deadline dispatches I/Os in batches. A batch is a
+    ## sequence of either read or write I/Os which are in
+    ## increasing LBA order (the one-way elevator). After
+    ## processing each batch, the I/O scheduler checks to
+    ## see whether write requests have been starved for too
+    ## long, and then decides whether to start a new batch
+    ## of reads or writes
+    path=$(dirname $iosched_dir)
+    [ -f $path/scheduler ] && {
+        echo "deadline" > $path/scheduler
+    }
+    ## This controls how many requests may be allocated
+    ## in the block layer for read or write requests.
+    ## Note that the total allocated number may be twice
+    ## this amount, since it applies only to reads or
+    ## writes (not the accumulate sum).
+    [ -f $path/nr_requests ] && {
+        echo "256" > $path/nr_requests
+    }
+    ## This is the maximum number of kilobytes
+    ## supported in a single data transfer at
+    ## block layer.
+    [ -f $path/max_hw_sectors_kb ] && {
+        echo "1024" > $path/max_hw_sectors_kb
+    }
+done