In this excerpt, presenter Tony Payne IBM, demonstrates that with an enterprise that demands "five nines", IBM Sametime is equal to the test. Join one of IBM's most experienced deployment engineers to learn the specifics of HADR deployment for every member of the Sametime family. You'll learn how to achieve high availability for each particular component, and you'll learn the requirements for various disaster recovery scenarios. This excerpt coverings clustering and planning.
2. Agenda
What is HADR?
How do we achieve it?
• Proper Planning
• A whole lot of clustering
•The components
– LDAP and DB2
– Sametime System Console
– Sametime Community
– Sametime Proxy
– Sametime Meetings
– Sametime Media Manager
– Sametime Gateway
3. What is HADR ?
How do we achieve HADR ?
– Capacity Planning
n+1
Vertical Clustering for scalability
Horizontal Clustering for redundancy
For Disaster Recovery
Mirror required components in an alternate datacenter
Set up DB2 HADR for end user datasources
Sametime Meetings and Advanced
If you want true HADR
Best to plan for it from the beginning
4. What is HADR ?
Capacity Planning
– Proper Capacity planning includes using techline resource
– Planning for growth
– Keep in mind what happens when
• A server goes down?
• A node goes down?
• Several nodes go down?
5. Planning for HADR from the beginning
Determine what components are required in a DR scenario
Do NOT scrimp on hardware – use virtualization to your advantage
– Just don't over tax the VM host
Realize ahead of time that every component will atleast be doubled and no components
get deployed as an 'all in one'.
– A 'simple' deployment can get complex very quickly
– Complexity increases once in production and then you decide you need to increase
capacity, but haven't planned for it in advance
• For example – Media Services has to be completely rebuilt to 'cluster later', but you
can go ahead and build a cluster of one at the beginning and expand on it with
little/no downtime.
6. Planning for HADR from the beginning
Get your friendly names in DNS ahead of time and test thru them
– Changing DNS to point to LB vs local machine is easier later than re-deploying
Use loadbalanced hostnames for DB2 and LDAP resources
– Even if this means local hostfile entries during the initial stages that point to single
instance
Plan ahead for needed SSL certificates
Understand your GTM and LTM configurations and options
– Involve your network team at the beginning to plan for what site best fits what
components
7. Planning for HADR from the beginning
Determine what components are required in a DR scenario
Do NOT scrimp on hardware – use virtualization to your advantage
– Just don't over tax the VM host
Realize ahead of time that every component will atleast be doubled and no components
get deployed as an 'all in one'.
– A 'simple' deployment can get complex very quickly
– Complexity increases once in production and then you decide you need to increase
capacity, but haven't planned for it in advance
• For example – Media Services has to be completely rebuilt to 'cluster later', but you
can go ahead and build a cluster of one at the beginning and expand on it with
little/no downtime.
8. Clustering Overview
Sametime uses 3 Clustering Models
Horizontal
Multiple Servers across Multiple nodes
Vertical
Multiple Servers on a single node
'Farm' Model
• Mutliple un-managed servers behind a loadbalancer
Different Components have different requirements when clustered
SIP Clustering Model
HTTP Clustering Model
Sametime does not support clustering across a WAN