Scale-Out Data Center Architecture

hastywittedmarriedInternet και Εφαρμογές Web

8 Δεκ 2013 (πριν από 3 χρόνια και 10 μήνες)

91 εμφανίσεις

Scale
-Out Data Center
Architecture

UKUUG Spring 2011 Conference
Leeds – 24.03.2011
Bernd Erk
Agenda
INTRODUCTION TO NETWAYS
SCALABILITY DEFINITION
TYPICAL SCENARIO
- DATABASE
- WEB AND APPLICATION SERVERS
GROWTH
DIFFICULTIES
ON TIME
ON TIME
ON TIME
ON TIME
ON TIME
ON TIME
4 SLIDES
3 SLIDES
6 SLIDES
6 SLIDES
2 SLIDES
6 SLIDES
DESTINATION
TIME
REMARK
- FILE AND BLOCK-I/O
SUMMARY
ON TIME
ON TIME
2 SLIDES
6 SLIDES
Q & A
ON TIME
1 SLIDE
INTRODUCTION TO NETWAYS
BOARDING
Introduction

to
NETWAYS
§

Founded in 1995
§

GmbH since 2001
§

Open Source since 1997
§

30 full time employees
§

Specialized in Open Source Systems Management
and Open Source Data Center Infrastructure
Skills


Monitoring


Performance Management


Configuration Management


Service Management


Knowledge Management


Asset Management


Identity Management


Backup & Recovery


High Availability Solutions


Cluster Solutions


Load balancing


Virtualization


Storage Solutions


Firewalls


Databases


Voice over IP
Open Source
Systems Management
Open Source
Data Center Solutions
Managed Services
Consultancy
Conferences

§

Open Source Data Center Conference 2011
§

6 - 7 April 2011 in Nuremberg
§

http://www.netways.de/osdc


§

Open Source Monitoring Conference
§

5 tracks with speeches & workshops
§

29 - 30 November 2011 in Nuremberg
§

http://www.netways.de/osmc



Conferences
§

33 years old
§

Head of Operations at NETWAYS

§

Responsible for Consulting, Development
and Managed Services

§

Background in Database Infrastructure and
Service Oriented Architecture
§

Open Source Projects


Icinga
team member

Introduction
to Bernd Erk
Root
of „all“
evil

SOAP
REST
SCALABILITY

DEFINITION

BOARDING
Scaling

Methods

§

Scale-Up


add as many resources as possible to a single node
system
§

Scale-Out


add as many resources as required, to your application
using additional nodes
§

“Mixed Mode”


add nodes based on virtualized platform to optimize
hardware usage
Scaling

Methods
-
Pictured

§

Scale-Up

§

Scale-Out

§


Mixed Mode


Scale
-Out
Challenges

§

Management
§

Bottlenecks
§

Workload
Distribution
TYPICAL SCENARIO
BOARDING
Typical

Scenario

Firewall

Firewall

SRV 1
Framework
Database
WebServer

Shared

Firewall

Typical

Scenario

§

WebServer /
Application
Server


Apache


Lighttpd



?? Tomcat /
JBoss

§


Framework




PHP


Ruby


Python


?? Java
§

Database


MySQL



PostgreSQL



NoSQL

SERVER
Framework
Database
WebServer

GROWTH
DIFFICULTIES
BOARDING
DATABASE
BOARDING
Database
Bottlenecks

SRV 1
Framework
Database
Webserver
Shared

Firewall



SQL/Model


QPS


Disk-IO

Database Solutions
§

Database Tuning
§

Query- and
Index-Optimization

§

Slow
Query Analysis
§

Reduce

number
of QPS


Application
Tuning


Caching
(
Memcache
)
§

Scale-Out

Database-Facility

Database
Scale-Out

SRV 1
Framework
Webserver
Shared

Firewall

DB1
DB2
Database
Database
Replication


Database Scale-Out Problems
§

Separate
Read
/
Write

Queries

§

Replication
Lag
§

Failover

Detection

§

High
Availability

§

Load

balancing

§

Replication

Bottlenecks

Database
Scale-Out
Solutions
SRV 1
Framework
DB-Master
DB-Slave
Database
Database
Master-Slave

SRV 1
Framework
DB-Master
DB-Slave
Database
Database
Master-Slave


MySQL“-Proxy

SRV 1
Framework
DB-Master
DB-Master
Database
Database
Master-Master

Database
Scaleout

Extended

Database-Facility

Master-Slave
Master-Slave
Master - Master
DB M1
DB M2
DB M1S1
DB M1S2
DB M2S1
Database-Facility

Master - Slave
Database
Sharding

Master - Slave
DB
Products

DB
Customers

Products-Slave

Products-Slave

Customers-Slave

WEB &
APPLICATION

SERVERS
BOARDING
Database-Facility

Web Server
Bottlenecks

SRV 1
Framework
Webserver
Shared

Firewall

DB1
DB2
Database
Database


RPS


Session-Handling



Memory

Web Server Solutions
§

Web Server Tuning
§

Try
a different Webserver


Apache


Lighttpd



Nginx

§

Scale-Out

Database
Facility
Web Server
Scale-Out

SRV 1
Framework
Webserver
Shared

Firewall

DB1
DB2
Database
Database
SRV 2
Framework
Webserver
Web Server
Scale-Out
Problems
§

Deployment
§

Synchronous Files
§

High Availability
§

Session-Handling
Web Server
Scale-Out
Solutions
Loadbalancer

Loadbalancer

SAN Storage
App
Server
Facility

Database
Facility

SRV 1
Framework
Webserver
SRV 2
Framework
Webserver
Shared

Load

balancer
and
Firewall
Web Server
Scale-Out
„Mixed“
App
Server
Facility

XEN A1
XEN A2
XEN A3
XEN A4
Application

App
2
App
1
App
3
App
4
App
5
App
6
Search

Search
1
Search
2
Memcached

Memcached
1
Memcached
2
FILE- AND BLOCK-I/O
BOARDING
I/O
Bottlenecks

SAN Storage
App
Server
Facility

Database
Facility

SRV 1
Framework
Web Server
SRV 2
Framework
Web Server
Shared

Load

balancer
and
Firewall


File-I/O



Disk-I/O

Image-Delivery

Bottleneck

App
Server
Facility

SRV 1
Framework
Web Server
SRV 2
Framework
Web Server
Shared

Load

balancer
and
Firewall
SAN Storage
Image-Delivery
Solution
Image
Facility

Image 1
Image 2
Ningx

Varnish

Ningx

Varnish

SAN Storage
SAN Storage
SAN Storage
Storage

Scale-Out

SAN Storage
SAN Storage
SAN Storage
DAS
DAS
SSD DISKS
SSD DISKS
SUMMARY
BOARDING
BOARDING
Tool
Facility

Image
Facility

Database
Facility

App
Server
Facility

Architecture

Overview

XEN A1
DB1
DB2
SAN Storage
Image 1
DB3
Image 1
DB4
XEN A2
XEN A3
XEN A4
XEN T1
Shared
Firewall
and

Load

balancer

Bernd‘s
5
Scale
-Out Rules
§

Rule
1:
Prevent
I/O
§

Rule
2:
Isolate

application

facilites

§

Rule
3:
Prepare

for

sharding

§

Rule
4:
Don‘t
handle
sessions
in
file

system

§

Rule
5: BE FAST - DON‘T BE SLOW
Q & A
BOARDING
BOARDING
Questions
and
Answers


NETWAYS GmbH
Deutschherrnstrasse
15-19
90429
Nuremberg


bernd.erk@netways.de

http://www.twitter.com/netways


www.netways.de
blog.netways.de