Atif Qadeer @atifqdr
Senior Systems Engineer, NSBU, UK, VMware
Jason Meers, Systems Engineer, VMware
In a nutshell, Abstraction
Traditional Approach of migrating data centres is challenging
The problem – Complexity, Inefficiency, Inflexibilty. (Move while keeping the lights on)
ABSTRACTION = Agility
Perfect DMDB enables DC migration because the environment and dependancies are understood – If only…
Talk about MANY problems and issues that can cause migration process to fail.
Migration Process – Private DC(s) to Private DC
Audit –> P2V –> SDDC –> Migration
*Discovered audit stage was very improtant, to understand the requirement, rather than what is actually deployed. ie Does the server need that many CPU or NICs?
*Unknown legacy equipment such as network switches created many problems.
Created a pair of 3 node clusters in both data centres and migrated VMs into the local cluster 1st. Showed the systems as working or resolved any issues and then long distance vMotioned the VMs to the new home.
Put effort into understanding VM to VM communication as the inter site link was only 1GBps.
Application portfolio was an important deliverable. Most of the apps where bespoke and very important. 2072 physical servers where migrated in 11 months. Main pain point was in the audit stage to understand the requirements. To keep the 'learnt' environment consistant, business groups where introducted. Admins then only have control of a sub set of the whole.
Power came from creating a Service Catalog which implmented Blueprints and cost profiles. Link this with change systems and use pre approved changes to deploy new VMs.
VMware Validated Designs
Combine this best practic information to deliver SDDC
Q&A – Tips for decommisioning systems
Audit stage included 'end of life' planning for VMs and who owned them
Traditional was quoted at £18 mil and 2 years. Actual via SDDC £2.6 mi and 11 months
Was it hard to convince people the technology works?
YES!!! – Had to run many PoCs and demonstrate actual working application to almost very stake holder.
Location of data centres
Everything was inside the M25 but 'all over' London. Large data centres, computer rooms & 2 servers under a desk, plus everything in between. Problems with physical machines that became virtual but still had links to physical devices such as tape libraries.
Question about physical to virtual networking. ie a Physical Oracle server that is communicating with 'new' VMs.
Used layer 2 bridging in NSX to resolve many of the challenges.
Look at NSX Universal objects
NSX Requirements question
"Needs jumbo frames on physical wires"