Beyond internal consumption, all data points to Exchange evolving into a better workload to run within virtualization.  Much of that can probably be attributed to Microsoft’s own virtualization technology, but Exchange on ESX benefits just the same.  Performance gains out of ESX 4 make for a good combination with the improved I/O for Exchange 2007.   Initial data for Exchange 2010 continues the trend of making Exchange a better workload in general and making it more appropriate to virtualize.

Primarily, the session was a presentation of lab work dones to measure the work loads using load generators and test environments.  Those numbers are good to create best practice documents, but sometimes real world experience trumps these experiments.   So by ‘eating their own dogfood’, I think VMware is endorsing the solution whole heartedly.  No more, yes it should work with no problems.

But in the real world, we all realize there will be problems from time to time, and so support is a big issue.  Microsoft, for its part, does not endorse running much of anything in a virtual machine.  Of course, we all know it works – just look at Microsoft education services – all of their coursework is done in VirtualPC.

The session also covered support options, wisely, as that is a major barrier for adoption.  Avenues for support include a Microsoft Premier support contract, which gets you support in any configuration; Microsoft support via the SVVP (Microsoft Server Virtualization Validation Program) with ESX 3.5 U2, Win 2008 and Exchange 2007; or support through an OEM like HP, etc., who are large enough to be able to offer their own support based on experience.

The session also focused on ways to establish resiliency and failover.  This discussion covered Single Copy or Failover Clusters – the traditional MCSC clustering technique, Cluster Continuous Replication (CCR) for Exchange, VMware Fault Tolerence and protection utilizing VMware HA.  I have a chart below that outlines some of the benefits and drawbacks from each method:

Ultimately, though, of the many solutions offered, there is no one correct way to implement this. It ultimately depends on your needs and which configuration most benefits your scenario. But the session provided lots of practical information for helping make those determinations for your real-world deployment.

Virtual Connect is about consolidation by reducing the number of physical connections required.  From the fiber channel modules, two 4-port VC-Fiber modules connect an entire chassis to the fabric, and then using NPIV, the fiber channel traffic is sent to individual blade servers within the chassis.  The 16 blade slots all share the 4 ports of each VC module.

The new development on the ethernet Virtual Connect side is a technology known as Flex-10.  Flex-10  is a new technology which helps to overcome some of the traditional challenges of blades (such as lack of NICs) when looking at blades for virtualization projects.  Virtual Connect Flex-10 allows up to 4 times as many NIC connections per blade server.  In a simliar way to the fiber channel implementation, a single Flex-10 physical NIC (pNIC) on a blade server can be divided into 4 logical NICs or FlexNICs.  These FlexNICs can be configured to different port speeds and capacities.

A blade server with two integrated Flex-10 NICs will have 8 logicial FlexNICs on the motherboard, and each one of those has it’s own MAC address.  These MAC addresses, like earlier blade NICs, can be overridden in the Virtual Connect server profiles and can be assigned virtualized MAC addresses, so the Flex-10 servers get the same portability that earlier server profiles in Virtual Connect offered.

The Flex-10 technology is a two parter – requiring both Flex-10 Ethernet modules on the interconnect bays of the blade enclosure and Flex-10 NICs in the blade servers.  Fortunately for companies with existing investments, the Flex-10 Ethernet interconnect modules are backwards compatible with earlier integrated NICs on your existing blades.

Each FlexNIC can be sized from 100Mb up to a full 10Gb speeds in 100MB increments.  The incredible thing, is that with the two available mezzanine cards and the integrated pNIC’s, Flex-10 technology allows for up to 24 FlexNICs per blade – an incredible amount in a such a compact space.

We have plans of implementing this technology, so hopefully I can post a follow up from personal experiences once we begin that project.

Data Recovery, like VCB, is an agentless backup technology used to grab either full VM image or file level backups of virtual machines in ESX.  Data Recovery also includes de-duplication technology and backup to disk, where VCB is just a method of obtaining the thin VMDK file to be backed up by a third party solution.  Data Recovery also makes use of innovation in the virtual hardware version 7 which allows for block level change tracking.  Although, Data Recovery can apparently (not 100% sure) backup earlier versions of virtual hardware, it won’t be nearly as fast because they lack the block level changes.

Data Recovery deploys in two parts – a virtual appliance and a plug-in for vSphere client.  The virtual appliance is imported from OVF format and with some basic configuration is ready to begin backups.  An IP must be configured and a VMDK must be added to the virtual appliance as a target for the de-duplicated data.  You may have two destination storage locations of up to 1TB each for a total of 2TB per virtual appliance.

Creating jobs for back are pretty simple.  A nice feature for this is that you may choose folders, hosts or clusters as part of backup jobs – meaning that any new VM’s included in that folder will automatically be backed up in addition to the existing VM’s.  This is a nice function for future-proofing your backup strategy.

Backup jobs are scheduled with a backup window to work and there can be up to 8 jobs at a time running, but the virtual appliance does all the scheduling and deciding of when to run the backups.  After an initial backup is run with all data, incremental backups are run from that point on – grabbing on change blocks.  Retention policies are also set for your backup stores and then enforced to keep a number of versions for your backups so that you can go back to a point in time backup.

Destination storage may be unmounted and exported, backed up or otherwise saved, though this is a manual process.

As another note, this is only available on vSphere hosts and cannot backup ESX 3.5 or vCenter 2.5 infrastructures.

Why would you want to do a long distance VMotion?  In my case, we have two data-centers – geographically close to one another.  We currently stretch our cluster between the two locations and it allows us to float VM’s using VMotion between the two.  The problem is that if we lose our primary datacenter, all storage is presented from here.  Long Distance VMotion is the notion of having two separate clusters, one in each datacenter, and being able to VMotion between them.

What was really news to me from this session (I’ll get to what was presented) was that we can present the same data stores to two different clusters and have them recognized on both clusters.  I am pretty sure I tried this way back in the 3.0 days and it failed to work.  This must have been added in 3.5 or 4.0 – I have not tried in recent years.

So, what was presented?  The three companies worked together to identify and trial a solution to allow for long distance VMotion.  At this point, there is a very narrow set of criteria must be satisified to be support and for this to occur.  Much of the restriction comes on the storage side, but network also presents some problems.  Apparently, everything you need in vSphere is there, if you separate each datacenter into its own set of hosts.

Requirements

• Distance between datacenters must be less than 200 km
• A single instance of vCenter to control both clusters
• Each site must be configured in it’s own cluster – the cluster cannot be stretched
• Dedicated gigabit ethernet network
• A single VMware Distributed Switch stretched across both clusters (ok, didn’t know we could that either)
• Same IP subnet configured on both clusters for the VM to run
• Cisco DCI (Datacenter Interconnect) type technology – if you have something similiar by another vendor, you’ll be supported – this means that you should have a core network that can handle routing traffic to either location for the IP – the VM networks must be stretched between datacenters
• Datacenter storage should be R/W on both sides
• VMFS Storage is presented to both clusters – VM’s in each datacenter are run from the local storage LUNs.
• Must have less than 5 ms latency and at least 622 mbps bandwidth (OC12)
• No FT across sites!

Surprisingly, we have most of this configured in our environment and its been status-quo for us for several years.  The biggest difference between our environment and this spec configuration is that we run a stretched cluster to achieve this.  Our datacenters are very close to one another and we only present storage from our primary datacenter so that we don’t have a split-brain scenario.  But, it does give me new things to think about and talk about with co-workers.  We currently don’t run two clusters or SRM because we like the flexibility to VMotions between datacenters – with that now a possiblity, we may have something new to investigate…

The session went through some of the canned reports that VMware has produced that leverage both the vCenter Database Schema and vCenter Database Views that VMware has produced.  The database views are a supported and document method to access this information and the schema will provide you access to all information, but is unsupported and may change from release to release as changes to the underlying schema occur.  In other words, if you code your report against the schema – your reports may break when you update vCenter and its database.  Database view documentation can be found at http://www.vmware.com/resources/techresources/759.

Primarily, VMware has released reports which meet some of the basic needs customers were asking for – such as Inventory and Configuration reports, Auditing reports (for accounts and VMotions for instance) and Utilization reports for hosts, vm’s, datastores and clusters.

I realize now that I didn’t write down the URL of where to get these canned reports so that I can deploy them when I return home, so I will have to search for that.  I will post an update once I find it.

HP is busy working on new technology to integrate vCenter and the HP hardware giving administrators greater visiblity into the hardware state and ease of management from a central pane of glass. This software is part of the Insight Control suite of products which HP already produces.  The new plug-in will sit on a Virtual Center client and will talk directly to the ILO, OA, and other agents on HP hardware to report back things like temperature, processor state, memory errors, etc. into the vCenter console.  HP reps also told me that the software goes bidirectional and can send information from vCenter over into HP SIM.

Power control also gets integrated as we continue to see HP work on power management in its rack mount and blade enclosures.  Power and heat were major issues addressed with the C7000 chassis and the functionality is being extended into the view from vCenter.  Today, if you enabled VMware Distributed Power Management (DPM) and it shuts down a host because the virtual machines are idle, HP SIM sees this as an error state and begins to alert administrators due to a failure of a host.  In the future, the two-way communication will allow for HP SIM to be smart enough to recognize that the host is offline due to DPM and not to a critical error.  We’ll gain insight (sorry, had to)  into vCenter from the HP tools.

HP also touted its new VDI hardware package which includes Proliant server along with a Lefthand SAN solution to provide for a very flexible, single-rack solution for deploying VDI.  The press release touts it as the first sub-$1000 per seat solution for deploying VDI.

For those who haven’t heard of HP’s Lefthand technology – Lefthand Networks was a start-up SAN provider who used commodity hardware to create iSCSI SANs with advanced snapshotting and replication features.  HP acquired this company last year and is steadily working to bring their iSCSI solution into their Storageworks lineup.  This solution seemed to use a purpose built Lefthand solution with what appeared to be a 3 to 5 U enclosure with two or three sliding shelves housing multiple mini SAS drives to accommodate terrabytes of data.  The disk enclosure shown during the presentation looked more like HP’s SFS enclosures than the Lefthand SAN offerings – the P4000, etc. – but we were told it was Lefthand powered.

For more information, see the original press release: http://www.hp.com/hpinfo/newsroom/press/2009/090901a.html

For information about the Insight Control for VMware, see: http://h18013.www1.hp.com/products/servers/management/integration.html

The second lab was pretty good – still plauged with a few technical difficulties.  This lab largely worked, albeit slowly, because all of the lab environment was running from a single LUN (not a great idea). The lecture portion of this lab was quite good, so here a few notes:

• VMware View contains the ability to prevent certain classes or hardware ID’s of USB devices from being able to attach to virtual desktops
• We were exposed to a tool called USBlyzer which allows you to get the hardware, vendor ID and other needed information for masking away these devices from the View desktops
• The USB limiting functionality is coordinated by registry entries and by the View tools loaded in the virtual desktop.
• There are command line tools which allow for mapping default VM mappings for users and the ability to point user data disks back to the correct VMDK if you have to purge or redeploy a virtual desktop
• We discussed the problems with User Data Disks and the mistaken notion that these were somehow roaming profiles for users
• We talked about basic troubleshooting which often relies on DNS resolution – when in doubt, check DNS, if all else is working, check DNS again.
• We talked about some of the things during a View deployment that could come into play – such as vCenter pointing to one Domain Controller and the View Desktops pointing to another (thanks to configuration in AD Sites and Services).  Because replication takes 15 minutes to occur, by default, in this configuration, View Desktops will try and join the domain, but computer accounts won’t exist.
• Dynamic DNS is almost an necessity and even in environments where BIND or other DNS is used, most people setup a Windows DNS instance for Dynamic registration if only for the  virtual desktops – then including a forwarder to the primary BIND DNS or other.
• And in case I didn’t say it already (I know I did, but here goes) – Problems with View deployments are almost always DNS.
• We were exposed to two command lines:
• vdmadmin – used to map users to default virtual desktops – vdmadmin -D -d desktop-pool -m machine -u domain\userid - can also be used to list user information, show entitlement or policies
• sviconfig – used to backup or restore databases for View Composer – also used to create a View Composer database
• See http://www.vmware.com/pdf/viewmanager_cl_tool.pdf for more information.

The afternoon was a partner session, which although I could have attended (nice little “Partner” tag on my credentials), but I chose not to.  Last year, the partner session was mostly how to sell and sales related marketing, but was interesting only because of the early access to the vSphere release information.   From other coverage online, the main announcements centered around cloud initiatives, which we heard about the following morning during the first general keynote.

• 12,448 attendees at this year’s VMworld – down from last year in Las Vegas
• Tod Nielson, VMware COO, challenged partners last year to try and recruit any of the 40 Fortune 1000 companies not using VMware to purchase VMware – 10 companies were recurited and those partners were rewarded with free passes to this year’s VMworld – not a bad challenge.  The same offer is extended to partners who can recruit any of the remaining 30 Fortune 1000 companies not using VMware.
• Paul Maritz, President and CEO, takes the stage – talking about cloud initiatives
• Martiz talks about how VMware has been driving capital expenditure savings and now also operational expenditure savings.
• Talks about the growing family of vCenter products, including Chargeback, Capacity IQ, Orchestrator, SRM, ConfigControl, etc.
• Commitment to small business – showed off the IT in a Box solution targeted at small business.
• Also targeted it small to medium business – VMware Go announced – service offering for users of the free ESXi product to quickly configure and deploy ESXi for their uses – a way to engage the community which is tapping into free ESX offering.
• vCloud Express – a service offering from cloud providers (like Terramark, who demo’d) to quickly and easily provision VM’s in the cloud all based on VMware’s vSphere platform.  At rates starting at about 3 cents per hour – or about $40 per month, its a good cost judging from my web hosting days…  and its all protected like any other vSphere cluster.  I’d certainly pay those prices!  I have a friend who runs a hosting company in mind that I may mention this offering to…
• We also saw a demo of the PC over IP technology that has been developed.  TELUS demo’d and discussed their company all over a PCoIP connection – and it looked flawless.
• Last, but not least, Martiz addressed the recent acquisition of SpringSource.  As soon as the demo began for this and source code appeared on screen, the flood gates open and people began leaving in mass.
• SpringSource’s Rod Johnson took the stage for the demo and showed off SpringSource’s customized Eclipse framework which now offers a Deploy to internal cloud, Deploy to external cloud or Deploy to the local – aka the “very private cloud”.
• The integration to deploy to cloud looked, to me, like Macromedia’s half-baked solutions for deploying code in Dreamweaver – it never really does the best job you could on your own – and probably not what a developer really wants – but good for a novice without a sys admin.
• Glossed over was SpringSource’s Hyperic offering, though Martiz did remark that he thought this is what monitoring will look like in the future – and I agree.  If only they could do it agentless – somehow leveraging ESX instead of agents – fingers crossed.