Archive for May, 2009

VirtualIron has finally been bought by Oracle with a finalised acquisition which has been an industry suspicion for a few months now, see http://www.informationweek.com/blog/main/archives/2009/03/oracle_to_buy_v.html;jsessionid=2FZ22WY3F2H1WQSNDLPCKHSCJUNN2JVN


This really is quite a tactical move by Oracle, as I have highlighted before on a previous post see > http://vmlover.blogspot.com/2009/05/oracle-still-helping-red-pills.html, Oracle’s main Virtualisation platform currently is the Oracle VM hypervisor which is a hypervisor offering that quite clearly lacks any tangible benefits which are available within market leading products such as Vmware, HyperV and Xen, and with the recent acquisition of Sun Microsystems provides Oracle with the brains in a jar responsible for current xVM architecture and Unix offerings, both Oracle VM and xVM are XEN based hypervisors which ensure that he can relatively just bolt on the best bits of each component without having to do major architectural changes to the underlying hypervisor method.
This may see increase in development of heterogeneous migration/management of x86/Unix hypervisors from Oracle, they may develop xVM manager to be the ultimate management product similar to Microsoft SCVMM which can manage competitors to suck them up and then move across to Oracle offerings.
On a hot topic and providing some possible general views, this purchase may also see some interesting developments within the VM Hypervisor Wars that are currently going on which is like a really fast rollercoaster at the moment and I am sure that many customers are considering what options they have based on what they spend in other areas of IT. For example if you are a large Oracle shop and are already forking out for Licensing or have an ELA with them, you run high end systems with Solaris then Oracle growing in size on the Hypervisor front may just kill any current and future aspiration you have as a techie with your current preferred virtualisation technology, it will simply come down to the fact of cost and the bottom line figure, they won’t care if you can hot add CPU/RAM, scale to excessive amounts of RAM and all that good greatness in Vsphere…fact.
Current competitors to VMware entering the hypervisor market with dominance in other areas of IT infrastructure such as Microsoft don’t have the luxury of being able to charge customers for maintenance costs on every level of the Infrastructure stack (if you have this that is) IBM have done this for eon’s with mainframe except they missed a trick and do not now have an x86 offering (yet), now that Oracle has bought SUN and your datacentre starts to look like an Oracle datacentre.
If your a VMware customer today it is probably not going to be the breeze to gain buy in for it or keep using it within your organisation as it was 12-18 months ago, Put the costs of VMware (with recent rises) in front of your beancounters and that cost and end figure being spent on alternative hypervisors today is at heavy risk of being on the next Oracle license negotiation being thrown in the direction of Oracle HQ (if they have any sense that is)
It is a shame, I feel VMware is likely to start losing “some” ground on the Virtualisation market at a time when they have just announced technology we wouldnt have thought about 3 Years ago. Back in the boardroom this may mean that any VMware evangelist who is passionate and wants to keep VMware within their architecture and roadmaps definitely will be needing solid core business case and technical benefits/drivers to ensure you can keep using the technology in force, so be prepared!!.
Amongst many requirements main ones needed to succeed in gaining VMware buy in get to know your product further than just what it offers within the technical benefits and product feature set, get to know really how licensing works based on the overall cost and value add that is wrapped around this and also ensure you liaise with a dedicated VMware SE and make sure they do large amounts of selling for you if you are that passionate about VMware.
As a summary overall changes with Oracle are not likely to not start to start to hit IT departments for 12-18 Months, so lets hope current changes adrift at Vmware with licensing in vSphere and other nonsense that may arise doesn’t make this more difficult to determine the decision makers purely on cost and ease of licensing renewals.

Some of you guys may have read and seen my views on Cisco UCS and read about how I feel Cisco may or may not gain massive popularity within the datacentre infrastructure space.

Cisco have provided direct from the horses mouth responsive feedback on comments such as mine that industry bloggers have raised and I am mightily impressed they have done this, it certainly takes a lot for an individual to represent the organisation to provide responses to feedback and possible criticism composed on various web blogs. The interview with Wendy Mars Cisco Director of the UCS Intiative is available in video format to view on http://www.youtube.com/watch?v=fDUbttbatBo&feature=channel_page , responses were very good and cisco like, analytic and positive with no fear of that what has been said by the likes of myself is likely to give UCS a damaged impression at such an early stage and it is a very positive review. Cisco have as i’ve said before not got into the blade server market to dabble, they are here to become number one.

Collectively the blog commentary has probably come across in some respects as quite a hard critism on Cisco and pointed quite aggressively and looked at flaws that the new converged strategy they are using. Run up to the California UCS Launch was marketed from the outset by Cisco themselves in moderate build up and I think a lot of people not privy to internal discussions with Cisco directly were quite amazed at how architecturally basic the technology and strategy is yet on the flipside of this amazed at how it is likely to turn datacentre consolidation up a notch by the sheer simplistic nature of converged backplanes and converged IO between SAN and LAN with FcOE.

In response to my question which was “Do enterprises want to unify networks, storage and servers?”, It does seem that large amounts of views on UCS have come from a bottom up engineering nuts and bolts perspective not from the top down C level view or even a middle view which is an Architect who needs to invest in the right technology yet not sacrifice functionality, scalability etc for their organisation/customer.

I think a lot of engineering folk are most likely afraid or more probably concerned at how Converged networking affects the SAN or LAN admin as we know it and the role that they both play in IT today, I think it will be very hard to find any C level bods that have as a harsh view on how it will integrate, post descriptive recommendation from trusty aide they will most likely love the overall benefits with the reduced opex costs of running a datacentre with fewer cabling requirements and costs of fabric switching and Lan switching (longer term not immediately). Engineering folk will probably find it similar to Server Virtualisation and blade backplane technology such as Virtual connect and will need to find time to adapt with the new way of working within a structured IT department.

Lets hope we see more response to blogger opinions and questions on the internet, the organisations and companies that we comment on are nine times out of ten obtaining free marketing for large amounts of revenue gain in return so videos like this to appease us are the least they can/should do.

Seems a few peeps are playing…I mean evaluating the new hot add CPU/RAM feature within VSphere.


To summarize the following OS’s only support hot add CPU/RAM regardless of whether a physical or virtual machine;
  • Windows 2003 Enterprise (RAM Only) and Datacenter (CPU & RAM)
  • Windows 2008 Enterprise (RAM Only) and Datacenter (CPU & RAM)
  • Linux running kernel 2.6.14 or above
When looking at how the hot add cpu feature works within the Windows OS it does appear that the running application and processes which are not system or SVCHOST based do not benefit on first plug of the CPU basically meaning that a reboot is required to ensure that the OS Can use the CPU Effectively and applications running as processes other than the supported can actually use the CPU’s, I presume this would mean apps like MS Exchange with Store.exe or SQL servers process which are not currently going to benefit immediately.

This is something you will need to take into consideration as to whether your hot add of a CPU actually is going to benefit. Microsoft SQL 2008 Supports hot addition manually after first plug, so I presume most of Microsoft’s future backoffice applications will too, not so sure how current MS applications likely to be virtualised will cope though.

Below are some links on more detailed information and resources i’ve picked this from for you to look in more detail.

http://www.kernel.org/doc/Documentation/cpu-hotplug.txt
http://download.microsoft.com/download/a/f/d/afdfd50d-6eb9-425e-84e1- b4085a80e34e/SVR-T325_WH07.pptx
http://www.sqlskills.com/blogs/paul/post/SQL-Server-2008-Hot-Add-CPU-(and-affinity-masks).aspx

POST UPDATED 23.07.09 with more clarity on features supported on Enterprise OS’s

Make what you want of the title of this post but VSphere includes functionality within the SDK and within HA that performs a screen dump of your Virtual Machine when you get a bluescreen or VMtools not responding (also something you can do in Vmware workstation) Eric Sloof has coded a great app to see each of your VM’s current screen dump, this can be found at http://www.ntpro.nl/blog/archives/1097-Big-Brother-Really-Is-Watching.html…


This got me thinking wouldn’t it be cool if VMware got a screendump of every virtual machine in use globally and made a big master collage out of this, a bit like what this kid did http://www.timesonline.co.uk/tol/life_and_style/article578081.ece

This is also a good reason to disable screen savers in your VM’s, they consume CPU cycles and also means you wont see what’s going on ;)

Jeff Browning, an EMC employee has raised some exceptionally great points on the current state of affairs that surrounds Oracle support within Virtualised environments. Within his post is a simplified view on why he feels the current support model or more should I say the lack of support model at Oracle is technically floored and really not about the actual technology and more about the corporation building an agenda to basically monopolise and gain more revenue, read it as it sets some ideas running in your head. http://oraclestorageguy.typepad.com/

Being an early adopter of Vmware Virtualisation has meant that in the past I have faced this type of support hurdle many a time and felt of the same opinion, one example being SAP, from the way I see it, it took SAP large amounts of use case and almost eating the dog food themselves first internally to gain trust with ESX 2 being used to host its main B2C training and demo suites.

The scalable benefits of ESX VI3 arrived with larger than 3.6GB of addressable RAM per VM, multiple CPU’s and it was then the start of them being able to comfortably provide customers with full blown production support for running SAP within a Virtualised environment. Microsoft is another one, they have in the last year introduced the SVVP program http://www.windowsservercatalog.com/svvp.aspx to ensure that Microsoft software customers play straight dice and align MS software versions to the latest releases and service pack revisions to gain break fix support from them. Although larger enterprises were able to gain acknowledgement for support entitlement under their Premier support agreement it was always a grey area of what would actually happen when the inevitable phone call had to be made and you had to migrate or replicate a problem with the software which was running say 10000 Exchange accounts! (it just wasnt practical lets face it).


In regards to why we still have a lack of clear positive Oracle alliances is that they more than likely still see Vmware (and other Hypervisors or soft parition mechanisms) as a threat to licensing and associated revenue, this is one of or if not the most important part of Oracles business model. Run considerably more Oracle workloads on less physical processors or run Oracle within a DRS cluster and that’s less revenue for them. On the licensing front they do seem to have tailored licensing for VMware (bit funny as its not supported) to mean that you have to license ALL possible hosts that a VM running Oracle could run upon, as you can gather this is just ridiculous and almost a non starter.

When it comes to aligning Oracle to any goal or hope of a complete virtualised datacentre you can see its just not going to fly and add up on your business case (or at least be easy at the moment), the licensing model is no doubt designed for large mainframe Iron which has no dynamic virtualisation capability such as Live migrate or resource scheduling. Interestingly however they now support LDOM’s which is a Hypervisor based Virtualisation tool (this was predate the SUN acquisition), http://sun.systemnews.com/articles/133/4/opt-sysadmin/21419 I can’t find if they have tailored licensing plan for this yet (they do for CMT Sparc processors) but this is a Hypervisor so no technical issues or arguements exists here then with the hypervisor indirection being used so this just increases the suspicion that Licensing rules the roost over virtualisation support.


Hopefully in future we will see some real movement in Oracle licensing schemes and acknowledged support, it is obvious that they are dependant on each other. Other issues arise within larger enterprises as it is hard to determine and show your data/dba teams that running Oracle on VMware is a reality, I could probably add loads of detail on how and what you can do to achieve this but will save for another rainy day.

The whole support, licensing and 100% virtualised datacentre vision will start to become more interesting when Vsphere arrives with all of its great scalable benefits along with the high availability benefits that are brought to organisations that can enable you to consolidate Tier 1 workloads, This was what happened in my example of SAP and others around release of ESX 3 so hopefully Oracle will acknowledge this and can forklift customers into building enough revolt to get Oracle on the straight and narrow and get focused on who is the most important person at the end of the day…..the paying customer.