Friday, May 16, 2008

Comparing MSCS/VMware/DFS File & Print

The following table shows information I was using to compare various Windows HA file and print solutions, including MSCS, VMware, VMware+MSCS, DFS, VMware+DFS and stand-alone servers. There are no recommendations, and most need to be adjusted or at least considered for your environment, but it might help crystalise your thoughts as it did mine.


Comparison Microsoft server ClusteringVMware HAMicrosoft Clustering on VMware HADFSStand-alone server(s)VMware HA with DFS for file shares
Highly AvailableYYYNNY
Satisfies SLAs??????
Maximum nodes8Limited by host hardware2N/AN/ALimited by host
Failover time<2 minutesVMotion or server startup time<2 minutesSPFSPFVMotion or server startup time
Single server Disaster Recovery – Software Failure<2 minutesSnapshot, server rebuild or manual recovery procedure<2 minutesSPFSPFpshot, server rebuild or manual recovery procedure
Single server Disaster Recovery – Hardware Failure<2 minutes< 30 seconds< 30 secondsSPFSPF< 30 seconds
Licensing2003 Enterprise per nodeDataCentre + CALs (depending on VM design)DataCentre + CALs (depending on VM design)2003 Standard2003 Standard2003 Standard
Hardware Failure – Data CommunicationsSingle/teamed NIC for prod interfaceNIC redundancy depending on virtual solution NIC redundancy depending on virtual solution + cluster-specific requirementsTeamed NICTeamed NICNIC redundancy depending on virtual solution
Hardware Failure – HBASingle HBA per nodeHBA redundancy depending on virtual solutionHBA redundancy depending on virtual solutionSingle HBASingle HBAHBA redundancy depending on virtual solution
OS Disk ConfigurationBasic DynamicBasicDynamicDynamicDynamic
Hardware utilisationPhysical serversVirtual serversVirtual serversPhysical serversPhysical serversVirtual servers
Cost allocationCost model requiredCost model requiredCost model requiredPer server/LUNPer server/LUNCost model required
Scalability/Flexibility – adding new nodes/LUNsYYYNNY, DFS for file
ManageabilityMSCS skills requiredVMware skills requiredComplex combination of both technologiesDFS skills requiredExisting skills, but increased per serverDFS and VMware skills required
User access to shares via UNCSingle nameMultiple namesSingle nameSingle nameMultiple namesDFS namespace
Future proofing – migration to new hardware/OSModerately complicated migrationRelatively simple upgrade path, reattaching LUNs or adding another VMModerately complicated migrationRelatively simpleRelatively simpleRelatively simple
Hardware on Vendor HCL??????
Backup/restore?Standard file backupVCB or ?Standard file backupStandard file backupVCB or ?
Printer administrationSimplified with Cluster 2003Duplicated effort on each print serverSimplified with Cluster 2003N/ADuplicated effort on each print serverDuplicated effort on each print server
Service and Event MonitoringCluster monitoring requiredStandard monitoring for virtual servers, host monitoring requiredCluster monitoring for virtual servers, VMware host monitoring requiredStandard monitoringStandard monitoringStandard monitoring for virtual servers, host monitoring required


1. Basic disks on a Microsoft server cluster can be extended if new space is visible on the LUN. The disks cannot be dynamic in MSCS.
See http://technet2.microsoft.com/windowsserver/en/library/cd4d0a84-6712-4fbc-b099-2e8fefeb694c1033.mspx?mfr=true


Wayne's World of IT (WWoIT), Copyright 2008 Wayne Martin.

3 comments:

Anonymous said...

Hey Wayne, nice table. One consideration. The DFS column could change quite a bit if you consider it from a domain based root rather than a stand-alone root. With domain based roots, you have built in high availablity and because the DFS structure is stored as a blob in AD and replicated to all DCs. This will change the viewpoint on things like scalablity and backup. Again good work on the comparison.

Thanks

Anonymous said...
This comment has been removed by the author.
Anonymous said...

Wonderful post. welcome to visit our links,and for more information's Business Continuity UK :)

Post a Comment