- Agnostic live migration,
- Infinite snapshot backup/restore,
- Natively reads Xen, Virtualbox, KVM, VMWare and HyperV virtual disks and configs,
- Can “clone on migrate” a running image to a “dmz” for analysis or a “remote cluster” for easy DR snapshots,
- Supports I/O and processor limits per VM, and
- Runs most workloads at least 95% bare metal.
Well, here's were we stand on that:
- Not yet,
- Close: new VCB solution will have API and lots of options,
- Do I still want this feature?
- Getting Fault Tolerance instead,
- Got this one, and
- Close: 90% bare metal.
On top of that, we're closer to private-public confederations that would allow vCustomer "A" to export/import to/from vCloud vender "B" if each are running vSphere.Also - and you've all seen this by now - we're getting a plugable vSwitch supervisor with many E-class switching capabilities...
Plus, we're getting application awareness in the networking back-end that will allow tweaking and performance tuning like never before... and much needed security tools to help with compliance and multi-tenant lockdown... And what about hot-add of vCPU, memory, devices, etc.? Wait and see what happens next...
The "100% Virtualization" mantra was introduced in the keynote by VMware COO Todd Nielson including desktop virtualization with additional capabilities through 2010. He said only about 10% of workloads have been conquered (world wide) so there's a lot of opportunity out there.
Following-up from Todd's intro was Boyd Davis, Manager of Server Products for Intel. Intel's continuing it's pitch that a 2P Nehalem-EP running VMware is better than 9 older 2p single-core Xeons and will ROI in 8-months on power savings alone. Is that kudus to EP or a mea culpa on the failings of the old FSB technology? In either case, welcome to the virtualization game, Intel...
Interestingly, Intel's positioning the 5500-series EP behind the 6-core, 7400-series 4P server based on consolidation ratios, cost-per-VM and TCO. This means AMD truly has a window of opportunity with Istanbul to make a dent in Intel's momentum - using Intel's own marketing data against them...
Most interesting from Boyd and Intel was the inclusion of the new testing data suggesting that the new IOv 10Gbps NIC delivers 9-9.6Gbps of measured throughput in new ESX/vSphere (iSCSI vs. iSCSI w/jumbo). Coupled with the Cisco Nexus 1000V technology, this interface really changes the landscape for vSphere servers. Now if I can just remember where I parked my 24-port 10Gbps switches...
What confused me about Boyd's keynote was his comparison of Xeon/ISA systems to "proprietary RISC" systems. VMWare is not a play in that space, but even if consolidation and TCO are considered, they "savings" is dwarfed by the sheer scale of the integration costs to migrate (ahead of software vendors) from RISC to x86_64. Davis has been in this market too long to forget his audience...
No comments:
Post a Comment