Virtualization, Cloud, Infrastructure and all that stuff in-between

VMware announced today that it was undoing the vRAM based licensing it announced to much boo’ing last year and will revert to the original per-CPU model (without the core limitations), I had hoped that vRAM was an intermediary step to pure per-VM licensing to help people (and the industry/channel) over the hump to move from legacy to something more radical – it would seem this will not be the case for the forseeable future.

Whilst I admire VMware for listening to customers that complained about vRAM changes and judging by the applause in the room when incoming CEO Gelsinger made the announcement you’ll probably not agree with me but I think this is a backward step given the stated vision for building cloud infrastructure. In my humble opinion they should have dispensed entirely with per-host, per-socket or even standard/enterprise/enterprise plus editions and bundles and focused purely on a per-VM (or vApp/Group of VMs) feature license. crazy? moi? maybe – but let me state my case…

Cloud is all about dynamic workloads, rapid provisioning and self-service, not being tied into X capacity or capability which has to be paid for up-front, in advance just incase you might need to use it one day – it’s about pay as you go – pay for what you use; it means you can’t take too many risks, or be too innovative because you need to sink significant cost upfront to make things happen or go to established clouds like Amazon, Azure, google etc. where someone has already taken that risk.

I’ve previously written about the need for the software industry to move away from legacy perpetual licensing models and move to a rental/subscription based model with lower cost of entry – allowing real flexibility to scale up and down, and allow businesses to attribute true cost to service lines or business units before the fact, rather than after.

Why not set a lower unit cost for vSphere per-VM but let consumers buy premium vSphere features and add-on products (like SRM, vCenter Ops, etc.) a la carte? on a PER VM basis then people can buy commodity hardware as and when they need it – without having to absorb large chunks of software license costs per physical asset, this is a good way of removing the cost barrier for small organizations (that may have them looking at Microsoft or other solutions and reduces the risk/uncertainty in planning implementation – pay for what you use (or license what you think you’ll need).

By way of a very simplistic example – with some dummy costs to illustrate the point is below

There is a base license to use (LTU) cost – which is per VM, regardless of how many ESX hosts, clusters, sites, datacenters it runs on, so you merely specify on a monthly/quarterly/annual basis how many VMs you are running – or have vSphere report the number (this is how it works for VMware’s service provider licensing – VSPP).

So the per-VM cost is built up in layers – base and premium functionality, pay for what you need, pay for what you use – don’t be tied into bundles, editions and planning for changes that may never come – easy to step up and down in levels of functionality for individual vApps (or even individual VMs).

Depending on how it’s implemented this may need some changes to the product to facilitate – maybe it’s not a hard block that you cannot enable a feature, but vCenter will report that vApp 1,3 and Test 1 are out of compliance with your licensing (think host-profiles) and give you a click-through option to uplift your VMware licensing rental agreement, or even disable the functionality for offending vApps.

Hardware is cheap and commodity but slow to procure and has a traditional manufacturing/distribution chain behind it, you can’t generally sell it back or scale it down.

Software is relatively expensive but quick to procure and implement (automatable, even in terms of the software defined datacenter)

Neither are particularly flexible today with legacy sales, channel and distribution infrastructures that underpin a massive workforce in VARs, SI companies etc.

It’s hard for hardware vendors to price and sell their products on a flexible basis (they have to cover cost of design/manufacturing for a physical asset (although I’ve previously written that I think this is where VCE should have focused here instead of becoming just a reseller/SI)

Software vendors can do much better to clean up here today as they are less-burdened with a traditional manufacturing and distribution chain – whilst they obviously have shareholders to satisfy and costs to cover they can be much more innovative with how they sell and distribute their products – taking a longer vision and more incremental revenue (maybe less satisfying to shareholders and analysts, but..)

In the cloud (especially so in the private cloud) Workload is important, not infrastructure; and cost should be attributed accordingly.

None of this will change overnight – but I do believe this is a better model for the future – feel free to disagree (constructively) in the comments!

You can register for VMworld EMEA at the early bird discount price now using the link in this article https://community.emc.com/community/connect/everything_vmware/blog/2011/09/14/get-the-early-bird-discount-for-vmworld-europe–even-at-the-last-minute

You can still get this ticket at a discounted price courtesy of EMC… travel isn’t too bad at present either, there still seems to be good availability of both official and non-conference hotels in Copenhagen.

I was able to price out a 2 day, 4 night trip (fly out Monday night after work –>conference Tues/Weds –> fly back on an early AM flight Thursday) with direct flights (SAS) and hotel (CABINN City) from a London airport (LHR) for approx. £380 – it means you miss the last day of the conference, but it gets you back into London around 9am to get to the office (assuming you work near LHR).

Taking this approach you also can make the best parties (Veeam and official VMworld party). My advice = do as may hands-on labs as possible while you are on-site at the conference and take advantage of the interactive group discussions, meet the expert sessions etc. you can catch up on the presentations as you get access to most of content on-line (slide/video/audio) post-event so you can catch any sessions you miss with the compressed timescales.

Great deal if you are tight on time away from the office and you couldn’t make it to VMworld US earlier in the year

Good on EMC for providing this sponsored registration link (c.£683 after discount), which should equate to a trip cost of approx £1k (for 2 days training/networking) – you’d be hard-pressed to find a formal training course for that in the UK with equivalent deep dive content without taking you out of the office for an entire week.

This year VMware are enforcing session registration for VMworld, my understanding is that if you don’t register and a session is full, you wont get in!!

this is a good thing IMHO, in previous years there have been long queues (although I hope they have better badge-scanners than in the past, otherwise there are mad queues!) and this has led to missed sessions, but the downside is that you’ll need to book, pay for and plan out your schedule early on.

So, to avoid missing out logon to the portal at http://www.vmworld.com and choose schedule builder – it’s pretty quick to do.

I have completed my registrations this evening but I’ve also seen a couple of sessions that are “sold-out” already, so get in quick!

vApps were introduced as part of the vSphere 4 release but were largely a forgotten area of functionality until now.

The concept of a vApp is as a bar-code for an IT service, where that service consists of a number of inter-dependent virtual machines containing applications that provide a service – for example a website. the vApp contains a number of virutal machines and is tagged with required levels of service and other pertinent information like start-up order, dependencies and required networks etc. to allow them to run successfully.

For example a corporate Sharepoint service could be grouped and deployed as a vApp containing relevant domain controllers, DNS, SQL and MOSS VMs to allow it to run – from a VMware perspective you manage and deploy the servers as a whole vApp rather than individual VMs.

With the vCloud Director (vCD) announcements it’s clear what VMware’s intention was; vApps are core to the service catalog concept for vCD, you don’t just pick virtual machines you can pick ready-to-use and self-contained application stacks to deploy and un-deploy.

However, if you think about it, it’s not as simple as it might seem once you go beyond the infrastructure level as you’ll still need to do in-guest engineering and automation to make this sort of deployment model successful but it’s a good foundation to work from.

This type of rapid provisioning and the level of in-guest automation required to make it useful can be problematic with Windows guest OS’es – there are still tight dependencies on domain controllers, forests and domain SIDs to get around for many applications. As more and more Microsoft applications move to PowerShell at the core this becomes more feasible but architecturally speaking it’s a problem for anything other than trivial applications.

The guest automation story is much better for Linux VMs deployed as part of vApps as scripting and automation is at the core of Linux deployment and always has been but it’s not done for you, vCD just handles the {virtual} infrastructure provisioning; tailoring and automating the resultant guest OS images is up to you but there is much more precedent on this space.

Strategically, Springsource makes a lot of sense for these sort of container deployments, the use of application frameworks breaks the dependencies on the underlying OS and makes applications much more flexible and portable, but this is an evolution away from current enterprise applications.

I took a different approach to VMworld this year, usually I try to cram in as many sessions as possible and don’t usually spend much time on the hands-on labs. – this year I am planning to do a 60/40 mix of labs and sessions. Because the sessions are audio recorded I can review them at a later date and make the most of the hands-on labs whilst I’m on-site.

From what I saw today queues for sessions can be big, although if you get there early it’s not too bad, but this isn’t a new problem for VMworld I don’t think they’ll solve it unless they start to move to Tech-Ed scale venues. with 16k attendees at this VMworld in the US maybe the tipping point is coming, although they have added Moscone West to the facility this year which has helped a lot.

Whilst session queues may have been long the hands-on labs have been pretty quiet in Moscone West with no major queues and it’s open 8am until 10pm Monday and Tuesday so I think I’ll focus on that.

There wasn’t a main keynote on day 1, I quite like this as in VMworld’s of old there was a general keynote on day 1 which was more marketing/product announcements with the more interesting technical keynote and demos on Tuesday.

I did all of the labs for an upcoming cloud related product that cannot be named until tomorrow – which is funny as you can take the cloud director (oops :)) labs today, which is going to be useful as I’ll be working with it when I start at VMware next week 🙂

I also did my joint session with Eric Siebert and Simon Seagrave, we ran out of time for most of the demos I had lined up so I’m going to upload them to YouTube in the next couple of days and post them on my blog if you are interested to see how the vTARDIS performs and is configured.

I look forward to the keynote tomorrow and will try and blog as much as possible – although there are certainly a lot of people doing twitter this year, so maybe just click this link and watch the #vmworld hashtag 🙂

Along with a number of other bloggers I was lucky enough to get a sneak preview of the VMworld 2010 labs setup today.

Wow, the setup is impressive, there is a massive self-paced labs room in Moscone West, offering 480 multi-lab seats, unlike previous years there are no specific areas for each lab, each workstation is self-contained and connects you your chosen lab from the "Lab Cloud" – which will be much better in managing the load and waiting times for popular labs.

You will have to register at the entrance and your session will be allocated to your badge number, there is a comfortable waiting area whilst you are called forward to do your labs; combined with the fact that each seat can be for any lab this is a great idea for managing foot-fall and waiting times.

There are a number of labs sessions pre-provisioned and ready to go and some will be provided on-demand when you are logged on, the ops team will be keeping a close eye on demand and can dynamically adjust the number of pre-provisioned labs to reduce start-up times for popular labs.

There are also labs upstairs where a subject matter expert (or “lab captain”) will run an audience through a presentation of the lab session and will be able to take Q&A and provide more information on the background.

The lab cloud is a heavily customised Lab Manager/vSphere environment offering up 30 different lab setups – each lab session runs from a dedicated vPod – a group of virtualized ESX, AD, vCenter hosts built from a totally automated template and accessed by a thin-client; making heavy use of virtualized ESX hosts (ala vTARDIS, but on a massive scale :))

The back-end infrastructure providing the lab cloud is split across 3 sites, 2 external DC’s and an on-site facility – the lab is closely monitored and automation deals with distributing load across the 3 facilities with resilience – the same infrastructure will be scaled down and will support VMworld Europe, although VMworld Europe 2010 will only have approx 1/2 the number of self-paced lab seats.

As you’ll see from the picture below the self-paced labs room is large, the podium in the middle is the operations centre where VMware staff co-ordinate and manage the labs environment, statistics will be relayed on realtime on the large projection screens.

Each lab workstation has a help button where you can request help from the on-site subject matter experts, I like this model better as it means the SMEs can be dispatched anywhere in the room to help out whilst allowing the maximum number of seats to be balanced across the available labs "on-demand"

I’d strongly encourage you to check out the labs, remember the normal presentation sessions are audio recorded (keynotes are usually video’d) and slides are available post-VMworld but labs are not, so this is your only chance to go hands-on – although the team know this is high on the list of "wants", the .PDF lab manuals will also be made available for download post-VMworld.

Interesting stat of the day, the environment will be creating/destroying about 5,000 virtual machines per HOUR, and over the course of the week they expect to handle 75-100,000 virtual machine create/destroy operations.