What advantage is there to having a broker?

[TDJB3]

I have briefly looked over several products and don't see a real advantage to a broker, yet I don't want to overlook something.

We are building a case to use virtual machines instead of PCs. The plan considers having individual machines for each user and not pools of machines to share.

Thanks for any thoughts.

BTW - I really did appreciate the post in regards to the number of virtual machines per blade server.

Another BTW - If your interested in trying to calculate your eletrical savings by switching to virtual machines and using thin client devices, check out this site http://michaelbluejay.com/electricity/computers.html

[mreferre]

Have a look at this:

http://it20.info/files/3/documentation/entry20.aspx

There are a couple of pages that outlines what a broker is supposed to do.

Massimo.

[admin]

We typically see the tipping point for picking up a broker at around 100 - 150 users. Managing that many users statically can get cumbersome. There are a number of paths to take at that point in that you could roll something really simple in house with the right design of devices etc. so you do not need a heavy weight broker or even a simple off th shelf one.

You will probably want something though even if its simple.

[hharold]

Hi there,

our main reason on why to use a broker is to give users 'Power Management'.

The ability to reset their own virtual desktop (in a hang situation for example), or to power on a virtual machine in the event it got shutdown.

If you have 10 virtual desktops you could have a support group to do this,

but when you get to a lot of virtual desktops (200 in our situation) you really want the users to be able to do this themselves.

cheers,

Harold

[mreferre]

Out of curiosity .... what is the device you are letting them access the vm's from ? Is it a standard Windows PC with a browser portal ?

Thanks. Massimo.

[hharold]

Right now we are using HP Thin Clients in the t55xx range in kiosk mode.

These thin clients run Windows CE.

Till now they are accessing the VDI desktop through a custom made RDP connection in the HP Connection Manager.

Each user has its own VDI desktop named after their Windows username.

We are waiting for LeoStream to get there Connection Broker to start a RDP connection through there browser version on Windows CE.

cheers,

Harold

[mreferre]

Thanks.

Massimo.

[jjohn]

I think you're referring to the Leostream Connect product, which was recently ported to CE (no small task). I believe HP was made aware of this work, but talk to your rep. Customers exert more pressure than vendors.

The CB just sends text files to clients and hopes that someone there understands what to do with them

[dgrobler]

So in your model, where each user has in fact it's own PC, but just virtual, I'm not sure there is much of an advantage in having a broker as long as the provisioning of additional vms/thin clients is simple enough.

But do you really want to stick to that model? It might be more efficient for you to roll out standardized desktops, that share all the same characteristics being able to be used by a group of users interchangeable. Also the way you upgrade virtual desktops might be more efficient just using templates. So the whole procedure around software distribution can be looked at from a completely different angle.

Also you might want to leverage those features as suspending a Vm when not used, creating snapshots automatically or manually and other nice things that had been so hard in the traditional PC world.

So the more you leave the traditional deployment path the higher the need to have someone in the middle.

-Dirk

[mreferre]

>So the more you leave the traditional deployment path the higher the

>need to have someone in the middle

You are right ..... but on the other hand the deeper you try to convert your users from personal desktops to NON personal desktops the more they will complain about what you are doing.....

I agree with your analysis (in fact there might be tools that would help with cheating the user such as roaming profiles etc etc) but this requires the IT to re-design a bit their Windows deployment standards and the end-user will have to be limited in what they could do (no more add-remove programs etc etc).

As I said I agree this would be the best .... but we can't forget that we need to deal with these people in the end ......

Massimo.

[leo2]

Removed. Logged in as the wrong user.

[dgrobler]

Complexity depends on the relationship between client/user and vm you have to maintain. If you have 1-1 relationship like userid matches VM name, well then there is not much of a brokering need. This resolution can be done in a straightforward manner. Even with a couple of hundreds of VMs I don't see the big advantage of a broker.

But as Massimo indicated, once this is implemented the desire to use other features will probably increase.

[leo2]

Disclosure: We sell Connection Brokers.

Answer to you question:

Connection Brokering is turning out to be a glue function, and like most glues you only need them when you need to stick several things together.

At the most basic level associating users with associated Hosted Desktops is a trivial task which is often best performed by a Post-It stuck on each user's monitor showing the IP address of their Hosted Desktop. This solution has significant advantages such as lower power consumption, portability, and it may even be cheaper than a CB.

So why do people buy our product even when they only have a small number of Hosted Desktops?

In comes down to

Authentication and single sign on - particularly when the users are coming in via a Cisco, F5, or Juniper SSL VPN.

License and Virtualization resource management - suspending and powering down unused machines.

Provisioning - especially as we move towards full integration with application virtualization.

Logging for regulatory reasons - HIPAA, Sabine-Oxley, etc.

and Thin client setup and control.

But, having said all that our best customers are people who have had a manual system, or even have written their own CB, and now want to automate the process.

Hope this answers your question.

Cheers

David

CTO Leostream

[mreferre]

>At the most basic level associating users with associated Hosted

>Desktops is a trivial task which is often best performed by a Post-It stuck

>on each user's monitor showing the IP address of their Hosted Desktop

Dave, I would add to your list that the Post-It solution would NOT give you user mobility..... if they sit somewhere else they either have to have their post-it with them or they wouldn't know where to connect to ......

Massimo.

[sgrinker]

Staple the Post-It to the user then...? :smileygrin:

[mreferre]

mh...... I know some users that would also want to have a mirror to read it....

Massimo.

[Anders]

Problem is you'd have to hire the entire j4cka$$ crew...

Your company would go bancrupt within the week... :smileygrin:

As MR. Crosbie said, for small environments, no CB may be cheaper.

But do remember total cost of ownership,

a CB can automates a lot of repetetive tasks.

I believe the inflection point is less than 100-150 users as mentioned,

but I got no calculation to back that up...

It is dependent on a lot of factors though.

\- Anders

[davlloyd]

A broker is a connection cop, it is there to control the assignment between user and machine and the connection experience (rules). There are many options available (these days) you just need to determine your needs and see what suits. In its most simplest form writing your own is not a huge chore as all you need is a front end to a database (or xml file) that pushes rdp connections down in the required format.

Massimo has the Matrix up for a feature comparison which is a good start point to see what is out there. There are a couple of vendors that support CE based clients and worth having a dig around. Personally I would look at why CE? MS have stated that they will not port RDP 6.0 support to CE so chances are there may be a hardware refresh (reflash, RAm upgrade, etc) required to support next generation or things such as multiple screens.

By aligning with other OS's you will also open the field of intergrated Broker support up further.

Interesting note on this as well is that Microsoft are to include Connection Brokering in there Veridian Virtual Management Suite (in beta now). Shame Veridian itself will not be a viable alternative for a couple of years yet (they have dropped the DRS/Vmotion equivalent out of the spec list).

[leo2]

The answer to user mobility and post-its is, of course, to write it on the palm of the user's hand when they come in the door in the morning.

This has much less risk that stapling the post-it to their body - but does have a unassignment problem if they do not wash their hands correctly.

On the issue of thin clients I think that there will be significant changes in the next 18 months in this space.

Please remember that we sell a Connection Broker - but I am trying to focus on industry trends rather than individual players - with one exception - because their technology highlights the trend.

The key problem with any thin client is that it has an operating system of some kind, anything from firmware to Windows XPe. If the OS is like the one in your network switch then your support costs are alot lower than if it is a variation on the OS in your Windows desktop.

In addition the thinner the OS the lower the cost of the thin client because there is less cpu and memory required.

So why are we about to add support for Windows CE? Simply because of USB. There are enough users who want to plug in USB devices and need a driver to run in the thin client that WinCE becomes the compromise between cost/support and functionality.

Personally, I think this is a transient phase - you can see the next generation of client interface devices in the devices based on the Teradici chipset - which we have started to support. They offer dual or quad monitor support with full motion video and USB.

The problem - it will be at least another 18 months or so before this technology will work with virtualization, so for now it is all about physical machine support.

So what do we do for the next two years? Well the clear need is to solve the USB problem without requiring drivers on the client device - which is technically very doable. Once that is solved by the thin client vendors then I think that we will see a strong move away from the WinCE / XPe powered devices because of the associate capital and operational costs - not because I have anything against MS - indeed they are driving our business.

I would be interested in other people's perspectives.

Cheers

David

ps We will be giving away Post-Its with "The Other Connection Broker" printed in the top at VMworld. (We will not be giving away staplers because of the liability issues).

[dgrobler]

Dave,

Disclosure: We sell stateless thin clients and add also some level of brokering to it.

I agree with most of your assessment. This is basically the big thing about VDI that resonates so well with customers. 'Move the PC into the datacenter', with all the attached advantages. Having than still a thin client that needs to be provisioned, managed, security fixed ... just eats a lot of intended benefits. Instead, you have just doubled the amount of PCs in your environment.

So, if you go the route of a firmware driven thin client, we call it ultra thin client, there is no OS and are no local apps to manage. The deployment is more plug and play like. However, there is of course the trade off with limitations in USB and multi media.

Does this prevent then customers from deploying VDI. I don't think so, if the targeted users are well chosen:

• Call center users, with loads of apps, but no (or limited) need for USB or multimedia

• Secure Work from home environments.

• Offshore development

• Health care ...

For these user groups you can deploy today. Without any feature regression, but adding mobility, security, manageability improvements ...

\- Dirk

PS. The post-it broker is a good start and can keep it somewhat simple. We use instead smartcards and hand them out to users. Equally simple, but the users have nothing to remember. They get simply their virtual desktop by putting in their smartcard into the slot.