#1 – TA01 – Managing VMware With PowerShell
#2 – DV02 – VDI versus Terminal Services
#3 – AP02 – Best Practices for Deploying Sharepoint/MOSS 2007 on VMware Infrastructure
#4 – AP11 – Performance Best Practices
#5 – DC21 – VMware – Standardised Platform Provisioning
#6 – AP01 – Best Practice for Successfully Virtualising Active Directory
#7 – DV07 – Server & Storage Sizing for VMware View
#8 – AP05 – SQL Server Performance on VMware: est Practices, Recommendations, Tuning & Troubleshooting
#9 – TA11 – Best Practices to Increase Availability & Throughput for the Future of VMware
#10 – TP12 – Consolidation of Performance Sensitive Applications

I managed to see most of these, but missed a few, so I will be getting in there if I can.  Check this photo here to see what is repeated when and where.

I can highly recommend Brian Madden’s presentation (#2) as I saw it first time round and it was great.

Some excellent stuff, particularly the demos from Bruce & jerry Chen.

vCompute

Stephen outlined that in the upcoming version of VMware – vSphere – a single VM has been able to achieve 23,000 total DB transactions per seconds – 250mbmb/sec of disks I/O, which equates to 510 disks spindles to saturate the I/O.  Pretty impressive and as Paul Maritz said yesterday and Stephen reiterated today – “No excuses not to run databases in a VM”

Also an area that has been seen as a weak point for VMware is web workload.  This can now scale so well that VMs have been tested and can scale up to the equivalent of serving 3bn page hits a day  – eBay ‘only’ gets 1bn!

vStorage

Thin provisioning will utilise the vStorage API and can fit straight into the alarms system in vCenter to alert the admins/storage admins as to when thin provisioning thresholds are about to be breached.

vNetwork

Distributed Switch – Allows single confiugration for all hosts and maintains state
Nexus 1000v – brings visitbility at the per VM level, lots more Cisco related functionality and management can be completed by networking team and takes the onus away from the VMware admin team.

Green Credentials

On the green side vSphere will take advantage of CPU related power saving functionality as well as IPMI using DPM.  When using DPM 50% power savings were made when comparing a cluster to it without DPM enabled.  There’s a video of this section of the keynote here.

Fault Tolerance

Fault Tolerance gets particular mention here and the difference between HA & FT were discussed:

HA – restarts a VM on a different hosts, but causes outages whilst the VM is powered up and the VM is in a crash consistent state

FT – lock-steps a VM with a shadow VM of itself on a different ESX Server – it also reproduces another copy onto another ESX Server to ensure it is still protected even while the original ESX Host is down

VMsafe gets briefly mentioned in the context of vSpere & the VMware ecosystem becoming almost completely policy driven in the future.

vCenter

Linked mode was also discussed and a demo also executed by a guy called Bruce.  You can view the video here.  This now means that you can managed 10 VirutalCenters from a single VI Client GUI.

An enhanced search capability has been added that runs across the whole VMware estate.

Host profiles also got discussed, again with a demo by Bruce. You can view the video here.

vCenter Linux Virtual Appliance was announced and met with whoops of delight from the crowd, who seem to ooze their Open Source credentials whenever possible.

Then Jerry Chen came out alongside Stephen and talked about a couple of cool developments.  For which I have videos:

MVP – Mobile Virtualisation Platform
PCoIP – software implementation of PC over IP to be incorporated into VMware View

Jerry Chen’s section was excellent, so check out the official VMware video of the keynote here, when it becomes available.

The project I am currently working on has a number of Brocade fabrics (4) after adding 3 new ones and managing them would have been a pain if not for Data Center Fabric Manger.  However, there is one massive limitation with DCFM and that’s the free version only manages a single fabric.  ‘What do you expect for free?’ I hear you cry, well I expect the jump up from the free version to be a bit more incremental than it currently is.  If you want to upgrade you can, to the Enterprise, which will cost you a substantial £30k list.  Now this is not bad if you push Enterprise to its limits (1000 ports in a single fabric), but what if you only have 200 ports?  well Brocade have obviously noticed this discrepency and are releasing a middle level version and are hoping to release in the next few months.

Something else of interest was the vCenter plugin that connects directly with Data Centere Fabric Manager and can trigger actions within vCenter based on policies set at the fabric level.  If an HBA’s bandwidth becomes saturated, a policy can send an alert to vCenter to move it to a host with greater HBA bandwidth availability.

All very neat stuff and the demo worked, so check it out if you deal with Brocade kit.

The best things often are the simplest and this tool is not complicated, but it just did exactly what most SMB businesses want from a backup solution and that is to work, without having to go to another vendor.

A few of the highlights as I saw:

• All managed through an extension to the vCenter Server GUI and completely integrated in a seamless way
• Destination volumes are encryptable
• SAN based backups
• ESX host integrated VCB
• Retention of backups is completely based on policies
  • Few – 7 most recent backups
  • Medium – 7 most recent backups, but kept for longer
  • Many – 15 most recent backups for a similar period to medium
  • Custom – Whatever you want them to be
• Schedule driven
• Multiple jobs can be run at once
• Restore jobs can run alongside backup jobs simultaneously
• Exclusions can be applied
• Restore rehearsal can restore a VM backup without overwriting the original
• Data-deduplication of backups is achieved by only backing up changes after the first full backup
• VSS leveraged using VMware Tools in Windows Virtual Machines

File level backups will be in the final release according to one of the VMware guys manning the lab and the documentation also alluded to this functionality.

I might be wrong, but it looked like it was all being driven by another CentOS JeOS Virtual Appliance, so VMware are staying true to their assertion that the future of software is through dedicated appliances!

Start as you mean to go on.

Check out the picture I took today of all the equipment being used to power the hands-on labs.

I must say I got a little tingle.

Check it out here.

VMworld Europe 2009 Gallery

VMworld Europe 2009 Videos

I know that I’m no Quentin Tarantino or David Bailey for that matter, but some of it miught be of interest.

Now that 92% or customers feel confident that mission-critical apps run well on VMware, these have included some of the following in production:

• SQL Server
• Oracle
• IBM DB2
• SAP

So there is no reason not to virtualise Citrix XenApp servers.  It is very easy to virtualise everything aside from the application servers in a typical Citrix ecosystem, such as:

• License Server
• Data Collectors
• Web Interface
• Datastore Servers

What makes XenApp Servers possible virtualisation candidatres:

Most XenApp deployments still on 32bit due to compatability and vendor support, which can cause application silos to be created.  This in turn can result in under-utilised servers.  On top of this the sweet spot for number of processors is dual core. There are currently far too many issues around 64 bit Operating Systems and the apps running on them for most people to use x64 in a terminal Services environment.  So all these factors can create under utilised XenApp servers.

Advantages:

• x64 & x86 XenApp Virtual Machines can co-exist in the same VMware cluster which makes management a lot easier
• Template provisioning makes creating more XenApp servers a breeze
• Hardware refresh means that you don’t end up with even more under utilised servers when new processor families are introduced etc
• Next generation of Hardware Assist will give a 20% increase in XenApp server performance

Considerations:

• Manage user expectations when discussing consolidation ratios – use scale out not scale up, increase the number of XenApp servers
• Test real-life workloads, not just taking benchmarks on the internet as gospel as they only usually discuss Microsoft Office
• Do not use P2V as it can cause more problems than it solves – Registry bloats and can have HAL issues so avoid at all costs
• Use AMD’s RVI or Intel’s EPT
• Using 2 vCPU template can be the sweet spot, but make sure you test
• 1:5 Consolidation ratio is usually achievable, don’t over-reach

All in all a good session if not really technical enough for my liking.

First up, Maurizio Carli, Head of EMEA, he kicks off with pointing out that last years VMworld attendance was 4500 and this year, despite the difficlt times, attendance is expected to be 4700.  Going to show the popularity of the event despite the current climate.

Maurizio went on to discuss the numbers that VMware is touting more and more often:

100% Fortune 100
91% FTSE 100
100% DAX 100

There is now an increase in the percentage of datacenters default build being virtual, up from 27% to 42% in a single year.

Paul Maritz, CEO & President, is up next and he kicks off with an introduction into how the VMware ecosystem is helping to make IT more service driven.

The key behind this being the VCD-OS, the foundation for the cloud and on top of this both the vCloud and vClient.  All of these trying to drive towards a Software As A Service model.

A breif history of VMware followed:

1999 – The Client Hypervisor
2000 – The Server Hypervisor
2003 – Cooperating Hypervisors & Management
2009 – The Virtual Datacenter OS

Back to the cloud, the question – How to move IT to a service driven model?

Use standardised building blocks for hardware
Build a scalable software solution
Policy based security & compliance
SLA driven management
Existing & multiple future apps model

One of the key attributes of virtualisation is encapsulation – take an existing application and package it all up and treat it in a flexible way.  This allows the injection of new intelligence in a non-disruptive way:

Security
Compliance

Allowing uniform application of these policies across the estate as and where you see fit.  Giving unprecedented management and flexibility.

Framework model application development can abstract the Operating System away from the developers.  Virtualisation allows the encapsulation of existing applications whilst providing the framework for the new way of writing applications.

Virtualisation is the key to making all of the above happen in an evolutionary way!

VMware vSphere again has become officially official as I mentioned yesterday The key points here are:

vCompute – Hardware Assist & Extended Live Migration
vStorage – Storage Management, Storage Replication, Storage Virtual Appliances
vNetwork – Network Management
Availability – Data Protection, Clustering
Security – Firewall, Anti-Virus, Intrusion Prevention, Compliance
Scalability – Dynamic Resource Utilisation

vCenter Suite, discussed next and elaborates on the SLA management mentioned earlier.  Paul Maritz, outlined the fact that vCenter Suite is to be very open and extensible.  Also talk of a portal cropped up, that sounded a lot like Life-Cycle manager with a new name, but I might be wrong, I’ll try and find out more when I can.

Paul makes what will be seen by some a bold statement “With the introduction of the vSpehere there is no reason why you cannot virtualise 100% of applications”.  I have been saying this for a long time and only really disuade the virtualisation of certain machines based on funky hardware rather than resource utilisation.  Nonetheless I think that customers are going to be very keen to test this claim out.

Kurt Glazemakers, from Terremark, came on stage to discuss their Enterprise Cloud.  Enterprise Cloud is a hosted cloud of compute resources that essentially provides resource pools to end users.  A web front-end to the back-end VMware datacenter allows the end-user to create any number of Virtual Machines within their own resource pool without any manual intervention.

Next up Joe Arnold of EngineYard, a Ruby On Rails specialist company.  Bringing a straightfoward interface to creating Ruby On Rails servers via a VMware vCloud without having to bother with what is happening at the back-end.  The demo here was seriously underwhelming I must say, a little wizard ran for a couple of minutes, that looked like it was building to a serious finale and then essentially a blank web page appeared with a couple of logos.  At this point somebody let out a sarcastic “Woo Hoo”, which I think summed it up fairly well.

Next, Zvi Guterman, IT Structures, a company set up to allow proof of concept environments from a vCloud.  The focus here seems to be the rapid deployment of Virtual Appliances for purposes of demonstration and testing.  Not a lot to say here, as not a lot happened during the presentation.

vClient next and Paul Maritz is back.

Mottos – Provision Users Not Devices & Address All Users

CVP – Client Hypervisor gets a mention here and is apparently very well developed, but it was very much glossed over, with little detail.  I’m planning on interviewing someone from VMware to get more information.

VMware are looking to te get to the point where the end user machine is completely irrelavent and as long as you can get to your corporate VCD-OS you can use your desktop.

Paul announced a new partnership with Intel to bring optimisations for the forthcoming VMware Client Hypervisor.  Specifically relating to vPro and writing intelligence:

Use the hardware to verify the correct VM is running
Ensure that security is wrapped around the VM
Out-of-and management of the device

All very wishy washy at the moment, but I’m sure the Intel stand will bring us more information and to that end so will I.

VMware View roll-out will be complete in 2009, which I think will include a client hypervisor, but I’m not entirely sure.

Maurizio Carli is back to introduce Dr. Wolgang Krips of SAP, who starts off with the phrase “for the next 25 minutes I will talk about…”  to the sound of collective groans around the room and a near mass exodus.

That’s it for now, but this is only the tip of the iceburg for now.

Stay tuned.