Documenting a virtualization project

Experiences in Virtualization

VDI project is on the move

Posted by martijnl on March 23, 2007

Since the last update on February 8th a lot of progress has been made and we now have an almost functional VDI environment for our offshore development work in Mumbai, India and we already have interest from different other parts of the organization to implement this for their outsourced or offshore/nearshore projects (Spain and two major Dutch clients).

The VDI cluster of two HP DL585’s is up & running with ESX3 and 125 VDI seats. The VM’s themselves are running on a 2TB dedicated allocation in the EMC Clariion SAN. After assigning key users both in Holland and in India the templates for the various expertises (Java, Microsoft, Oracle, Software Testing) have been made and approved by the process managers in Holland and India.

The modelling for the templates varies, some expertise areas like Software Testing have created one template for all users, some expertises have created templates on a per project base (MS, Java). For this last group a process was created for template lifecycle management. Basically, we deploy a standard expertise template in a “staging” area, the projectleader takes care of bringing the template up to project standard, the template is stored in the library and the projectmembers are tagged to this template. At the end of the project the template is “degraded” to the template archive.

Leostream logoOn the Leostream side all went well, 75 seats of the Connection Broker were ordered directly via the Leostream site, and the correct license key was mailed the next working day. Configuration was pretty straightforward with only one issue. This related to putting the virtual desktop to suspend state after a users logs off. This was solved in an update of the software, a proces which is easy to do, backing up the configuration in one file is also a nice feature.

In Leostream Active Directory group membership (one group in AD for each expertise) is matched againt a Leostream policy, only one match possible. Within Leostream groups of VD’s receive a “tag” which can be linked to the policy and a policy can hold multiple tags. With this we can assign users to multiple types of VD’s (for instance if someone is working on multiple projects or cross-competence) by simply adding the tag to the policy and without extra fiddling in the Active Directory. After the users

logs in at the VDI site, they can choose a VD from a list which is controlled by their policy/tag settings and after that they receive a .rdp file which they can open with Remote Desktop from their own desktop. We however decided to use the Leostream Connect Tool which has to be installed at the physical desktop of the user. With this the user simply starts the tool, fills in his credentials and Leostream automatically picks a VD from the user’s available group of VD’s, creates the correct .rdp file, starts the VD and performs the user logon via Remote Desktop and suspends the VD after user logoff.

So basically all configuration on the cluster, Leostream and the templates is finished, we are now focussing on completing documentation, test plan and most of all, waiting on the 20Mb NL-Mumbai connection to be delivered. We had some unexpected delays due to a relocation of the development center to a different building on the same compound without our knowledge. As this specific building is off-net (no fiber coming into the building), so an extra fibre extension was required.

So, we still love VMware and VDI 😉


10 Responses to “VDI project is on the move”

  1. JeffG said

    I used to use leostreams P>V when it was too new and was bugged quite to heck. I did like their responsiveness to support issues back then. I have been pushing our migration to VM, but our Oracle DBA is always bringing up issues of how VM does not work or fails to meet spec for Oracle db’s. How did you get past the fact that Oracle does not support VM? They have too many links on their support sites stating this fact to dodge it. I have yet to see one of our DBA’s oracle servers even come close to challenging a gues vm, but I can’t name any large companies moving production to VM.. even our VMWare reps can’t help me resolve the Oracle lack of support issue.

  2. Arjanb said

    JeffG, Oracle does support running their products on VMware. From Metalink Note 249212.1:
    “For Customers running Oracle on VMWare. No limitation on use or distribution.” They even
    have a joint customer support agreement with VMware and an alliance to optimize deployments.
    The fact that your Oracle DBA states that VMware “does not work” or “fails to meet spec”
    sounds like “afraid of the unknown”. Both statements are poorly argumented because VMware
    obviously does work, we for example run 104 databases and 3 application servers over six
    virtual Linux servers, without any problem. And “fails to meet spec” is simply a matter of
    correct server sizing on the ESX host. Just size fysical memory and CPU on your virtual Oracle
    server exactly the same as on the fysical one and you can’t go wrong.

    One thing you can do is create a virtual server for him to play around with, show him the
    advantages of quickly adding RAM, vCPU’s and taking snapshots before important releases or
    updates. He will demand a virtual Oracle server very soon, trust me 😉 And you can always
    create a fallback scenario by keeping the virtualized fysical Oracle server on hold for a
    few months for easy switchback in case of an emergency. We run production and development
    Oracle fully virtual, Oracle 8i/9i/10g, Oracle 9iAS/10gAS and Oracle 10g Grid Control, all
    without any (support) problems. Just go for it 🙂

  3. JeffG said

    I wish it weren’t true but oracle has a thousand such statements on their own sites
    see MetaLink Note #249212.1 regarding no support for production vm’s.

    You seem to have missed a part of their information from that metalink.
    Oracle will not support production deployments of Oracle software on VMware (see MetaLink Note #249212.1).
    I just realized you used the same metalink which is specifically about NOT supporting vmware for PRODUCTION. Darn, I had hoped you had something new. That’s too bad.
    The features you and I see our DBA does not. Some people fight for no real reason.

  4. AndyP said

    Im interested in how you are connecting the sites. Where is the detacenter ?
    are you seeing any latency?

  5. martijnl said

    The sites are connected with a 20Mbit MPLS site to site connection. It is an extention to our existing grid of five local offices. Mumbai effectively being the sixth office on that grid.

    The expected latency will be between 150ms and 200ms with the projection of the supplier being on the low side of that band.

    Planned delivery date of the Mumbai side of this connection is somewhere in May (roughly 5 weeks from now) so we will have actual data by then.

  6. martijnl said


    Sorry to hear the DBA has problems with the VMware solution. As mentioned earlier we keep some physical servers separate for testing production problems on should the supplier make a problem of it being on a virtual server. The savings we have by making it virtual servers make up for the extra effort needed in such a situation.

    The difference is probably a lesser degree of influence from our DBA’s regarding the hardware on which their servers run. We have been using VMWare GSX Server since 2002 and have been running Oracle on VM (on GSX / test and dev) since 2004 if I recall correctly.

    In preparation for this project we tested Oracle on VI3 as early as June / July last year so we knew it would work and that it would run production.

  7. AndyP said

    Martijnl when you have more data I would be interested on latency and actual usefullness of the remote connections. what type of application will they be using? are they keyboard intensive apps with alot of screeen refreshes. I am working on a similiar project with connections from the US to mumbai, but from your description all the offices are local in India or am I incorrect?

  8. martijnl said


    To clarify my earlier post: all existing offices are in The Netherlands. Because of the way our Telco provider is delivering the connection we can view the Mumbai office (from a network standpoint) as “the sixth office”. For Mumbai extra firewalling and access control is applied because of the remote status of the office.

    The work done in Mumbai is done in teams based on competence:
    – Microsoft (mostly .Net work with Visual Studio Team System)
    – Java (using, among others, Eclipse, MyEclipseIDE, Altova XMLSpy and scripting work with UltraEdit)
    – Oracle (this could be Oracle Forms Builder, Webforms or Portal and plain PL/SQL programming with PL/SQL Developer for example)
    – Software testing (making testscripts and performing tests on various systems using plain Office/wordprocessing and (Mercury) QualityCenter)

  9. Emerson C. Marchni said

    One asks. In this project of VDI you quotation the use of protocol RDP. I have heard of some manufacturers who different things, but for what vi in its project the use of an environment of SBC on VDI if make necessary.
    He is correct to affirm this?
    Today I have more than 2500 users inside of the environment of Server Terminal and a total of 55000 stations inside of the company. We are starting to make deploy to move for Seen Windows.

  10. description here

    VDI project is on the move

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: