Thursday 18 June 2009

Cloud Computing and the Virtualised Datacentre

VMware event at Old Trafford Theater of Dreams (TM)

Piled over to Manchester from Sheffield on the train to attend a day of sessions on Cloud Computing and the Virtualised Datacentre. I expected we would get the tram to the ground but major maintenance work was being carried out on the tram infrastrucutre in the city centre - they have a replacement bus service on to St Peter's Square but it is as quick to walk. This added a little delay, as well as mysterious stops every minute or so along route, taking lessons from the tube I expect!

We missed the first few minutes of the keynote speech but it was general background stuff, there were lots of references to computing in the cloud, internal and external clouds.

Three sessions were to run simultaneously during each session block, we split up and took a session each. Not all sessions were repeated through the day.


The first session I attended was on storage, this focussed on EMC and how they were ahead of other storage vendors in tying in to the new vStorage API framework available for vendoirs to hook in to vSphere. No surprise really that as EMC own VMware, but Netapp and other vendors either already have or will be working on those areas too. The EMC capabilities will be typical of the sort of stuff coming - Netapp already have have rapid cloning and other functionality that is triggered from the VI Client using the new API. Thin provisioning from the VMware side (not required if you are already doing it at the storage side), sVmotion that can be done within the storage array rather than via an ESX host, recovering VMs from native storage snapshots (rather than VM snaps) from
within the VI CLient, storage tiering, where storage can be moved from expensive disk to less expensive disk, nice stuff like that.


The second session attended was a very good one from Cisco. Lots of interesting stuff albeit with a cheesy video about the infrastructure 'tribes', the Networking, Storage, and Server tribes and how as virtualisation encompasses all these areas there is now a fourth tribe, and the potential for disharmony - I don't think we've had any of that but as I occupy three of those roles I'm not about to own up to disagreeing with myself;-)

Much was said about putting the network team back in control of the networking, right down to the virtual switch in ESX - something that has been the department of the Virtual Environment admin until now. This is a good thing, there should be a consistency in how switches are managed even when virtual. New technology (vNetwork in VMware which is built on by Cisco soft/hardware) allows a distributed vSwitch to be created, this means port/machine policies, statistics, rules and the like follow the virtual machine when the machine is migrated. There are other ways of picking up some of these statistics by monitoring each VM directly, something we do now, but the fine level of control normally available to Network Admins has been missing. An example of one of these levels of control is security. Currently all VMs attached to the same vSwitches can communicate with each other unhindered as when talking to each other over the same vSwitch traffic is never sent out to the managed LAN, in many environments this is not desirable and there may be requirements that machines are walled off in various ways.

The VMware environment needs to be running at least Enterprise Plus to be able to use this new technology. There is a Cisco Virtual Supervisor Module (VSM) that exists as a VM that controls the DistributedvSwitch. The Nexus 1000v is the Cusco hardware that sits out in a rack provides ports that create an extended ethernet in to the virtual environment, as physical hosts are added to the virtual environment they automatically pick up the distributed vSwitch configuration.


No comments:

Post a Comment