VMworld 2009 Ask the Experts Follow-Up

Amongst the most valuable sessions at VMworld 2009 for me this year was the Ask the Experts sessions. Not for any specific question exactly, but for the brain share that was gathered up on stage, and in the quality of the questions being asked. I will do my best to summarize them here based on my ‘live tweets’, because of this however, they will be largely unformatted.

TA2259 – Ask the Experts, Virtualization Design

This session was storage heavy, but was excellent.

• Ask the experts – “When do you need seperate VMDK Files?”  <- Every logical drive, maxims, backups, pvscsi, snaps, dedupe, etc
• Ask the experts – from @vseanclark – “Go big or scale out in vSphere?” <- You can scale up in vSphere, but It’s a question of risk.     Followed up with a discussion of scheduling, not at 1:1, go for the reference architectures as a building block model.     General consensus is that scale out is better, due to scheduling and multi-threadedness
• Ask the experts – “Hardware or software iscsi HBAs, or mixed case?” <- generally software
• Ask the experts – “Direct map luns to in guest iscsi v rdm/vmfs/cluster?” <- in guest=historical; rdm=not perf, clustering, use VMFS     Across the vendor community they’re working to eliminate any case that breaks the VMDK model.     TODAY – Do what is required to support your apps.
• Ask the experts – “NFS v iSCSI?” <- don’t look @ cpu cost, tcp conns. MBps=fastest single link, scale out nfs for better thru-put     Mix of both is best. There may be operational benefits by vendor.
• Ask the experts – “Memory Overcommit? Is it better to swap or vSwap?” <- OS swap is better than vmKernel swap, start low w/ mem     Some experimenting has found PVscsi to use guest caching, and gets crazy perf with guest memory     set reservations for databases, as they bypass guest OS memory mgmt and swap anyways. Slotsizes/TPS etc.
• Ask the experts – “There are more benefits to virtualization than consolidation” <- AGREED
• Ask the experts – “Monitoring IO in vSphere w/o calling san” <- vscsi stats, vendor plugins, perf stats, IO DRS, datastore autotune
• Ask the experts – “PSA, Benefit or tune PSA?” <- Up to vendors to prove the benefit.     Every array is different, lowest possible costs, best outcome, but vendors need to prove benefit.
• Ask the experts – “Hardware raid config to use” <- It depends. YES IT DOES!
• Ask the experts – “Do vApps fill a niche, or will they be main stream?” <- Currently a niche, we’re in the early days of vApps
• Ask the experts – “%rdy When is 5% to 10% an issue?” <- Are the users whining? Its all about scheduling & smp
• Ask the experts – “Learning resources for vSphere?” <- @scott_lowe’s book! http://bit.ly/28fbk7

TA2259 – Ask the Experts, Virtualization Design – #2

This session was notable because there were not as many storage questions as the first 🙂

• Ask the experts – “Find a solutions provider that you can trust & you can work with.”
• Ask the experts – “Growing your environment, did you seek a solutions prvider?” <- Long experience & assistance from VMware
• Ask the experts – “What are the drawbacks to using a dedicated snapshot datastore?” <- Management and operation
• Ask the experts – “SQL Dataload & testing, snapshots?” <- If you must, to it on the array.
• Ask the experts – “Are storage vendors working on array intergration?” <- Yes, exciting things in this space.
• Ask the experts “Excessive CPU slows things down, how do I review that?” <-scheduling has changed, vsmp adds overhead, start small “%rdy, context switching”
• Ask the experts – “How do you change organizational mind sets (groups)?” <- Change is scary, but make it valuable to them.
• Ask the experts – “What hurdles w/ Desktop Virtualization?” <- its complicated, graphics, tco, etc
• Ask the experts – “What hurdles w/ Desktop Virtualization?” <- its about user management and experience
• Ask the experts – “What hurdles w/ Desktop Virtualization?” <- it changes the way things operate, manage expectations
• Ask the experts – “What hurdles w/ Desktop Virtualization?” <- it changes buss ops, w/o that it all falls apart
• Ask the experts – Use your vMUG!
• Ask the experts – “Anti-affinity & FT” <- It depends. It needs to be tested. < Good Q, he’s got them stuttering
• Ask the experts – “Virtualizing Low latency apps, QOS and IO garantees?” <- Not always a good use case for Virt. IO DRS is coming.
• Ask the experts – “Assign a NIC to a VM?” <- VMDirectPath – NIC to the VM and slightly lower latency.
• Ask the experts – “Local storage for VDI?” <- Not quite… Use a VSA or so.
• Ask the experts – “My boss administers his own snapshots…” <- Oh Shi…
• Ask the experts – “Snapshots take a long time to commit…” <- Yes, they do, but try to use the storage level snapshot.
• Ask the experts – “Snapshots != Backups” <- Agreed!
• Ask the experts – “Where did the nursery go?” <- LOL What’s a Nursery? A resource pool where the VM is born…
• Ask the experts – “FCOE?” <- Converge the networks, but the standard isnt 100% ratified yet

And finally… some video of the session from Mr Chad. Thanks!