This document discusses deploying Accelrys Enterprise Platform (AEP) servers in a load balanced configuration for high availability. It recommends using a staging server to test configurations before deploying to production nodes. All nodes should be configured identically and share storage. A load balancer should be configured to distribute traffic evenly across nodes. Applications need to be packaged and deployed identically to each node to ensure consistency across the load balanced farm. Load balancing improves availability, scalability and performance but requires additional infrastructure and configuration.
(ATS6-PLAT09) Deploying Applications on load balanced AEP servers for high availability
1. (ATS6-PLAT09) Deploying
Applications on Load Balanced AEP
Servers for High AvailabilityConrad Agramont
Senior Product Manager, Enterprise Technologies
Accelrys Enterprise Platform
conrad.Agramont@Accelrys.com
2. The information on the roadmap and future software development efforts are
intended to outline general product direction and should not be relied on in making
a purchasing decision.
4. • High Availability
• High Concurrency
• Load Distribution
• Great for “Small Jobs”
• Windows and Linux
• Leverages Third-Party Load-
Balancing systems
Web Farm/Load Balancing
AEP01 AEP02 AEP03 AEP04 AEP05
Corporate Network
LOAD BALANCER
Back-End Network
Accelrys Enterprise Platform 9.0
Web Farm
5. • Connecting to AEP over
the Internet without VPN
• Connecting to PP server
via a different URI than
server name
Internet Facing
INTERNET
FIREWALL
AEP01 AEP02 AEP03 AEP04 AEP05
DMZ
LOAD BALANCER
Back-End Network
Accelrys Enterprise Platform 9.0
Internet Deployed Web Farm
7. Operating Systems
• Windows Server 2008 R2 (64 bit)
• Windows Server 2012
• Red Hat Enterprise Linux 5 (64 bit)
• Red Hat Enterprise Linux 6 (64 bit)
• SuSE Linux 11 (64 bit)
8. Storage Requirements
• Shared Storage used by all nodes
– Jobs
– User
– Public
– Shared
– Temp
• Performance Considerations
9. Windows Server for Shared Storage
• Disable distributed directory caching
– Registry Update
[HKEY_LOCAL_MACHINESYSTEMCurrentControlSetservicesLanmanWorkstationParameters]
"FileInfoCacheLifetime"=dword:00000000
"FileNotFoundCacheLifetime"=dword:00000000
"DirectoryCacheLifetime"=dword:00000000
– Registry Update provided
<Server Install>installLoadBalancingConfig_Enable.reg
10. Load Balancers
• What we tested with (F5)
– Hardware
– Virtual
• AEP generically works with different load balancing
technologies
– Hardware Device
– Software
– Virtualization
12. Planning and Deployment
• What to do with the Installer
• Everyone has their own XMLDB
• What doesn’t work in this deployment
– Pipeline Pilot Authoring
– Model Building Protocols (they write into the XMLDB)
• Applications, Protocols, & Components need to use
Packaging to be distributed across the nodes
13. Step 1: The basics
• Directory Services
• Database Connections
• Storage
– Use local temp folders for better job performance
– Shared Storage required for all nodes (Jobs, Users, Shared Directory)
• Recommend 100 GB Network dedicated
• NFS Supported, but not recommended
• Fast Storage technologies are recommended (e.g. IBM GPFS)
14. Step 2: Build your Staging Server
• Deployed using GUI or Unattended Installation
• Same configuration as production
• Export settings that will be used for Load-Balanced Nodes
Staging01
15. Step 2: Build your Gold node
• Deployed using
Unattended Installation
• Configure Reverse Proxy
• Enable Load Balancing
• Export Configuration
• Set Folder Locations
AEP01
Staging01
(ATS6-PLAT07) Managing AEP in an Enterprise Environment
16. What do we have now?
• Single Node
• XMLDB is no longer “Writeable”
AEP01
Staging01
17. Step 3: Deploy additional nodes
• Deployed using automated purposing technology and
leverage unattended installation
• Import Configuration from “Gold” node
AEP01 AEP02 AEP03 AEP04 AEP05
Staging01
18. What do we have now?
• Bunch of nodes, but no load balancing
AEP01 AEP02 AEP03 AEP04 AEP05
Staging01
19. Step 4: Load Balancer
• Deploy the load balancer (see Vendor directions for
details)
• Enable SSL at the load balancer
• Set Affinity
AEP01 AEP02 AEP03 AEP04 AEP05
Staging01
20. What do we have now?
• A web farm
AEP01 AEP02 AEP03 AEP04 AEP05
Staging01
21. Deploying Applications
• First deploy to the Staging server and test your regressions
• Before deploying to the web farm, use maintenance mode to
pause requests going to the servers
• Xcopy package to each node
• Use PkgUtil to deploy package
• Turn off maintenance mode
• Run tests
• Pat yourself on the back AEP01 AEP02 AEP03 AEP04 AEP05
Staging01
(ATS6-DEV06) Using Packages for Enterprise Application Delivery
22. • Always have a Firewall between AEP and the Internet
• At the Firewall or Proxy Server level
– Only expose the end-points that are required for the service
(e.g. don’t expose Web Port if not required)
– Don’t allow access to Admin Portal via the Internet
A word about security
23. • Load-Balancing provide high availability and concurreny,
but require additional configuration settings
• Load-Balancing is an infrastructure pattern the requires
servers, shared storage, and networking equipment all
working together
• Load-Balancing requires planning and guidance for
application development, deployment, and operations
• See the AEP 9 Deployment Guide for more details.
Summary