VMworld 2012 Live Blog of INF-VSP1168 – Architecting a Cloud Infrastructure
This is the Live Blog of INF-VSP1168 – Architecting a Cloud Infrastructure. You’ll find my recap of the session below.
Speakers / Moderator:
- Chris Colotti- VCDX # 37 VMware, Inc. – Moderator
- Aidan Dalgleish – VCDX # 10 VMware, Inc.
- Duncan Epping – VCDX #007 VMware, Inc.
- Rawlinson Rivera VCDX #86, VMware, Inc.
This will be a panel discussion. The session begins with “Gathering Requirements”
Talk to Customers
- Gather information and document
- Categorize (Requirements, Nice to Have, Constraints, Assumptions
Great advice for consultants and anyone going for the VCDX certification.
Some Example Requirements
- Increase agility / flexibility while reducing costs
- 99.9% Availability
Some Example Constraints
- Hardware already acquired
- Number of NICs dictated per server
- Reuse existing equipment
Some Example Assumption
- Sufficient switch ports
- Storage can handle expected workload
- Properly trained staff
Historical best practices
- Understand why it is a best practice
- It is better to put requirements first and apply the best practices to that requirement
- Constantly evaluate best practices and question whether it still applies to newer software
Some Use Case Examples (The application comes first!)
- Server consolidation
- OPEX savings
- Resource optimization
Conceptualize Your Design
- Building Blocks – Operations, Time to Market, Compliance
Sizing / Scaling
- What does the environment look like today?
- How will size / scaling impact your Design / Project?
- What is the use Case?
Again, all of these details should revolve around the application use case.
- Capacity Planner
With capacity planning, we are trying to identify the anomalies within the physical infrastructure. These anomalies may require some extra planning.
- 2 vs 4 sockets
- Optimal Memory Configurations
- TPS vs no TPS
- What is the sweetspot? – Still seems to be dual socket – 96GB of memory
- Networking is rarely a bottleneck (especially with 10GbE)
- 10GbE will lift all (or most) constrains for a long time
- Again, use the capacity planning report to identify anomalies.
- It’s not just the capacity, but the performance that matters
- Take into account RAID penalties
- Size includes the VM and its associated files
- The RAID level used impacts IOPS
- RTO impacts the number of VMs per datastore. Can you restore within the RTO window?
- What is the vendor?
- AMD vs Intel
- Blade vs Rack
- Embedded ESXi, Boot from SAN, Auto-Deploy, etc?
- Management integration?
- Local SD / USB
- SAN Boot
- PXE Boot with Auto-Deploy
- USB is cheap
- Local disk has higher availability
- SAN Boot, Stateless computing but more costly
- PXE Boot is the best of all worlds
- How many VMs?
- vCenter Appliance?
- Web Client? – Use vCenter Appliance
- SRM / View / vCloud Director? – Consider vCenter Heartbeat
- Statistics level will impact performance / scaling
- Read, read, read the documentation!
- Scale up or scale out?
- Should we put VUM on the same server?
- How many vCenter users?
- How many physical Datacenters?
- Will each physical DC need a vCenter Server?
- How many Datacenters will each vCenter manage?
- Separate clusters for DMZ?
- Separate clusters for test?
- Remember the Datacenter is the boundary for vMotion, not the cluster
- EVC is required for FT, consider enabling it by default
- What type of switch will be used?
- What are the pSwitch capabilities?
- Will VLANs be used?
- Will PVLANs be required? – Consider vShield App
- Network I/O Control is awesome! Use it!
- VNIC level firewall
- DVFilter used for in- and out-bound traffic
- VShield App Firewall per host
- Set rules on vCenter Objects like Resource Pools and Portgroups
- vShield Manager cannot be locked out, Exclude vCenter as well
- 5 tuple firewall
- Load Balancing
- Network Isolation
- Data Security Options
- High Availability option is now available
- Consider a management cluster as more appliances are deployed
Back up all management databases!
- Protocol choice? – Fibre Channel higher performing, NFS is much easier to implement
- Does the array support VAAI?
- Does the array support VASA?
- Can we use Storage DRS? – Auto-Tiering can impact the decision
This wraps up the session. There was a lot of great information for practitioners and those who are pursuing the VCDX certification. This has been one of my favorite sessions at VMworld.