Problem solve Get help with specific problems with your technologies, process and projects.

TSX-EMEA: Architectual Directions for Virtual Infastructure

TSX-EMEA: Architectual Directions for Virtual Infastructure.

Presenter: Yu-Shen Ng, Group Product Manager – ESX server, VMware

This presentation was very much a replay of Steve Herold’s keynote address at the start of this weeks TSX. The subject matter was therefore very familiar content. It was about how VMware leverages CPU enhancements such as:

  • AMD-V and Intel-VT
    For most work loads VMware’s “Binary Translation” system is more efficient. But the hardware assist does help in 64-bit guest operating systems
  • Memory enhancements like “Nest Page Tables” and “Extended Page Tables” (NTP/ETP)
    Currently, the VMkernel must map the VMs memory pages to the physical pages of the ESX host. These new memory features, due next year, will reduce this work load.
  • Device I/O Virtualization
    VMware will be supporting Intel VT-D and AMD’s IOMMU projects, which allow for a more virtualized presentation of devices, such as fibre-channel

It’s clear that VMware see the immediate future as a hybrid model of a couple different types of virtual machines that are customized on the guest operating systems requirements. This would include various combinations of VMware Binary Translation, CPU Vendors Hardware Assist and Paravirtualization. The term they are using is “transparent paravirtualization”. Where the definition of the VM seamlessly is changed based on the requirements. The message is customer choice. The last couple years of innovation have been about making x86 virtualization work; the future is about making it perform better. So far, very familiar territory.

So you might wonder why I attended this session. Well, as Group Product Manager, if there is anyone who knows what VMware will be doing in the future, it was this guy. This was a chance to take the speaker “off topic” and ask about VMware innovations rather than VMware merely reacting to hardware improvements.

VMware have been typically very reluctant to disclose any information about new features for Vi-3. The argument given is they are prevented from doing so because of legal restrictions in the US; you can’t talk up your product in an effort to raise your stock value. Additionally, they would only be helping the competition by giving their big ideas away. However, myself and the group were able to “put the squeeze” on and were able to get a few hints from him. Here they are:

Live data migration
Currently available to sys-admins to move data of a running VM from ESX 2.x/VMFS2 to a ESX3.x/VMFS3 volume. The natural extension of this is a tool for data management for ESX 3.x as well
Third Party Virtual Switches
Allowing other companies developping virtual switches in the ESX host to have properties and features you would expect to see in conventional physical switches. So rather than unmanaged switches of ESX currently (except for vLAN support, security settings and load-balancing settings) the switch would be more like a conventional Cisco Catalyst switch.
Better Management of Virtual Switches
The ability to create switches across many hosts, as well as being able to store the vSwitch information in an XML file for portability
HA non-stop
VMware HA without the crashing of the VMs. VMs continue to run elsewhere on another ESX host. I would expect VMware to redevelop the “replayer” feature currently in VMware Workstation 6
Improved Hardware Support
Currently we are dependent on VMware writing VMware drivers for devices; this means it is necessary to validate any hardware purchases against a VMware HCL. It looks like VMware wish to open this out further to allow other vendors to speed up the process by having a program to validate third party drivers

Improved VM Virtual Hardware
Asked if we will get improvements inside the VM the answer was yes – pressed further we were promised 8 virtual CPUs in a VM

Dig Deeper on VMware Resources

Start the conversation

Send me notifications when other members comment.

Please create a username to comment.