Lokasi ngalangkungan proxy:   [ UP ]  
[Ngawartoskeun bug]   [Panyetelan cookie]                
Skip to content

Commit 859f2cd

Browse files
ShannonLeavittShannonLeavitt
authored
Merge pull request #15741 from MikeRayMSFT/20200701-bdc-perf-guidelines
Stage perf tuning guide
2 parents acaed2a + 90f45ca commit 859f2cd

2 files changed

Lines changed: 170 additions & 0 deletions

File tree

‎docs/big-data-cluster/performance-guidelines-tuned.md‎

Lines changed: 166 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,166 @@
1+
---
2+
title: Performance best practices for SQL Server Big Data Clusters
3+
description: This article provides performance best practices and guidelines for running SQL Server Big Data Clusters on Kubernetes
4+
author: mihaelablendea
5+
ms.author: mihaelab
6+
ms.reviewer: mikeray
7+
ms.date: 06/30/2020
8+
ms.topic: conceptual
9+
ms.prod: sql
10+
ms.technology: big-data-cluster
11+
---
12+
13+
# Performance best practices and configuration guidelines for SQL Server Big Data Clusters
14+
15+
[!INCLUDE [sqlserver2019](../includes/applies-to-version/sqlserver2019.md)]
16+
17+
This article provides best practices and recommendations to maximize performance for applications that target services running within a big data cluster.
18+
19+
The following guidelines focus on recommendations for configuring the Linux operating system hosting the Kubernetes worker nodes where BDC will be deployed on. As a best practice, configure the tuning profile before deploying the big data cluster. The settings included in the proposed tuning profile were validated during the case study conducted by Microsoft and Intel. The results of the study are published for download in this [whitepaper](https://aka.ms/sql-bdc-spark-perf/).
20+
21+
> [!TIP]
22+
> For tuning configurations specific to SQL Server on Linux, see [Performance best practices and configuration guidelines for SQL Server on Linux](../linux/sql-server-linux-performance-best-practices.md). Also, other best practices like index design for SQL Server databases, still apply.
23+
24+
## Proposed Linux settings using a tuned `mssql-bdc` profile
25+
26+
Create a **tuned.conf** configuration profile with the content below.
27+
28+
```bash
29+
[main]
30+
summary=Optimize for Microsoft SQL Server Big Data Clusters TPC-DS performance
31+
include=throughput-performance
32+
33+
[sysctl]
34+
#network tunings
35+
net.ipv4.conf.default.rp_filter=1
36+
net.ipv4.tcp_timestamps=0
37+
net.ipv4.tcp_sack = 1
38+
net.core.netdev_max_backlog = 25000
39+
net.core.rmem_max = 2147483647
40+
net.core.wmem_max = 2147483647
41+
net.core.rmem_default = 33554431
42+
net.core.wmem_default = 33554432
43+
net.core.optmem_max = 33554432
44+
net.ipv4.tcp_rmem =8192 33554432 2147483647
45+
net.ipv4.tcp_wmem =8192 33554432 2147483647
46+
net.ipv4.tcp_low_latency=1
47+
net.ipv4.tcp_adv_win_scale=1
48+
net.ipv6.conf.all.disable_ipv6 = 1
49+
net.ipv6.conf.default.disable_ipv6 = 1
50+
net.ipv4.conf.all.arp_filter=1
51+
net.ipv4.tcp_retries2=5
52+
net.ipv6.conf.lo.disable_ipv6 = 1
53+
net.core.somaxconn = 65535
54+
55+
#memory cache settings
56+
vm.swappiness=1
57+
vm.overcommit_memory=0
58+
vm.dirty_background_ratio=1
59+
60+
#kernel NUMA
61+
kernel.numa_balancing=0
62+
63+
#filesystem
64+
fs.aio-max-nr=1048576
65+
66+
[vm]
67+
#should be revisited for SQL large pages use in master/data/compute pods
68+
transparent_hugepages=never
69+
```
70+
71+
## Install **tuned** utility on all the Kubernetes worker nodes
72+
73+
To install **tuned**, execute:
74+
75+
```bash
76+
apt-get -y install tuned
77+
```
78+
79+
## Apply tuning settings to all Kubernetes worker nodes
80+
81+
On each of the targeted worker nodes, copy the **tuned.conf** file created above:
82+
83+
```bash
84+
cd /usr/lib/tuned
85+
scp -r <sourcePath> ./mssql-bdc
86+
```
87+
88+
To enable this **mssql-bdc** tuned profile, save these definitions in a **tuned.conf** file under a `/usr/lib/tuned/mssql-bdc` folder on all the Kubernetes worker nodes and enable the profile using:
89+
90+
```bash
91+
chmod +x /usr/lib/tuned/mssql-bdc/tuned.conf
92+
tuned-adm profile mssql-bdc
93+
```
94+
95+
Verify it's enabled using this command:
96+
97+
```bash
98+
tuned-adm active
99+
```
100+
101+
or
102+
103+
```bash
104+
tuned-adm list
105+
```
106+
107+
If the profile is changed dynamically, for the new changes to take effect, you must restart **tuned** on all the impacted worker nodes:
108+
109+
```bash
110+
systemctl restart tuned
111+
```
112+
113+
Logs for **tuned** service can be found at */var/log/tuned/tuned.log*.
114+
115+
Optionally, you can configure the tuning profile on one node in the Kubernetes cluster and use the script below to copy it over and configure on the remaining nodes.
116+
117+
```bash
118+
#!/bin/bash -e
119+
# This script takes a list of servers (IPs like `cat ~administrator/workerhosts)) as input
120+
# and update these servers with the local version of mssql-bdc tuned.conf.
121+
122+
is_root() {
123+
local is_root_set=0
124+
if [ "$EUID" -ne 0 ]; then
125+
echo "Please run as root"
126+
else
127+
is_root_set=1
128+
fi
129+
return "${is_root_set}"
130+
}
131+
132+
# function main
133+
if is_root -eq 0; then
134+
exit 0
135+
fi
136+
137+
while [ $# -gt 0 ]
138+
do
139+
# sometimes, people add non-breaking space characters to their *host* files.
140+
WORKER_IP=$(echo "$1" | sed -e 's/\xC2\xA0//g')
141+
echo -e "updating mssql-bdc tuned on \e[42m${WORKER_IP}\e[0m"
142+
# the following commands assume tuned was not set up and active...
143+
# TODO: add the tuned install and status checks
144+
ssh "${WORKER_IP}" mkdir -p /usr/lib/tuned/mssql-bdc
145+
scp ~administrator/tuned.conf "${WORKER_IP}":/usr/lib/tuned/mssql-bdc/tuned.conf
146+
ssh "${WORKER_IP}" tuned-adm profile mssql-bdc
147+
ssh "${WORKER_IP}" systemctl restart tuned
148+
ssh "${WORKER_IP}" tuned-adm active
149+
shift
150+
done
151+
152+
```
153+
154+
## Next steps
155+
156+
For more resources including reference architectures for SQL Server Big Data Clusters, see:
157+
158+
* [Case Study: SQL Workloads running on Apache Spark in MS SQL Server 2019 Big Data Cluster](https://aka.ms/sql-bdc-spark-perf/)
159+
160+
* [HPE Reference Architecture for delivering insight across all your data with Microsoft SQL Server 2019 Big Data Clusters](https://h20195.www2.hpe.com/V2/GetDocument.aspx?docname=a50001963enw)
161+
162+
* [Dell EMC PowerStore: Microsoft SQL Server 2019 Big Data Clusters](https://www.dellemc.com/resources/en-us/asset/white-papers/products/storage/h18231-dell-emc-powerstore-sql-server-big-data-clusters.pdf)
163+
164+
* [Microsoft SQL Server 2019 Big Data Clusters: A Big Data Solution Using Dell EMC Infrastructure](https://infohub.delltechnologies.com/t/microsoft-sql-server-2019-big-data-clusters-a-big-data-solution-using-dell-emc-infrastructure/)
165+
166+
* [Microsoft SQL Server 2019 Big Data Clusters on Cisco UCS Reference Architecture](https://www.cisco.com/c/en/us/solutions/collateral/data-center-virtualization/unified-computing/sql-server-on-big-data-cluster-on-ucs.html)

‎docs/toc.yml‎

Lines changed: 4 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -294,6 +294,10 @@
294294
href: big-data-cluster/notebooks-manage-bdc.md
295295
- name: Run a notebook with Spark
296296
href: big-data-cluster/notebooks-tutorial-spark.md
297+
- name: Performance
298+
items:
299+
- name: Tuning guide
300+
href: big-data-cluster/performance-guidelines-tuned.md
297301
- name: Secure
298302
items:
299303
- name: Kubernetes RBAC for BDC

0 commit comments

Comments
 (0)