Traffic Manager vs Load Balancer for RDP sessions

%3CLINGO-SUB%20id%3D%22lingo-sub-140714%22%20slang%3D%22en-US%22%3ETraffic%20Manager%20vs%20Load%20Balancer%20for%20RDP%20sessions%3C%2FLINGO-SUB%3E%3CLINGO-BODY%20id%3D%22lingo-body-140714%22%20slang%3D%22en-US%22%3E%3CP%3EHello%20Azure%20Team%2C%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3EMy%20challenge%3A%3C%2FP%3E%0A%3CP%3E-%20i%20want%20to%20provide%20access%20to%20Virtual%20Desktops%20(VDI)%20deployed%20in%20Azure%26nbsp%3B%3C%2FP%3E%0A%3CP%3E-%20i%20have%203%20regions%3A%20us%2C%20emea%2C%20apac%3C%2FP%3E%0A%3CP%3E-%20in%20each%20region%20i%20have%203%20shifts%20(10h%20each%2C%20with%202h%20overlap)%2C%20so%20each%20region%20provides%2024%2F7%20support%3C%2FP%3E%0A%3CP%3E-%20i%20need%20to%20reinitialize%20each%20VDI%20every%2024h%20(redeploy%20new%20Windows%20Terminal%20Server%20to%20make%20sure%20no%20customer%20data%20is%20there%20for%20more%20then%2024h)%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3EWhat%20would%20be%20the%20right%20design%20%3F%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3EI%20was%20thinking%20to%20use%20Traffic%20Manager%20(DNS%20loadbalancing)%20nested%20profiles%3A%3C%2FP%3E%0A%3CP%3E-%20global%20profile%3C%2FP%3E%0A%3CP%3E-%20nested%20emea%20profile%3C%2FP%3E%0A%3CP%3E-%20nested%20us%20profile%3C%2FP%3E%0A%3CP%3E-%20nested%20apac%20profile%3C%2FP%3E%0A%3CP%3EThen%20another%20layer%20of%20nested%20profile%20inside%20each%20region%20(3%20shifts%20per%20region).%20But%20the%20challenge%20is%20in%20overlaps.%20I%20need%20to%20make%20sure%20my%20shift2%20from%20emea%20starting%202%20hours%20before%20shift1%20in%20emea%20is%20finished%20can%20work%20correctly.%20If%20i%20will%20switchover%20on%20the%20beginning%20of%20overlap%20my%20shift1%20will%20get%20a%20new%20DNS%20A%20responses%20and%20their%20RDP%20session%20could%20be%20redirected%20to%20a%20new%20VDI%20(they%20would%20lost%20all%20data).%20I%20was%20thinking%20to%20increase%20DNS%20TTL%20timer%20to%202%20hours%20-%20but%20that%20looks%20like%20a%20can%20of%20worms%20(i%20would%20not%20have%20failover%20if%20specific%20VDI%20is%20going%20down).%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3EAnother%20option%20is%20not%20to%20do%20DNS%20loadbalancing%20but%20application%20Load%20Balancing%20(using%20LoadBalancer).%20But%20i%20do%20face%20similar%20challenges%20-%20how%20to%20plan%20overlapping%20shifts%20(each%20region%20is%20having%20all%203%20shifts).%3C%2FP%3E%0A%3CP%3EI%20would%20like%20to%20use%20Load%20Balancing%20with%20sticky%20option%20-%20this%20way%20existing%20RDP%20sessions%20would%20be%20redirected%20to%20the%20same%20VDI%2C%20at%20the%20beginning%20of%20the%20overlap%20i%20would%20change%20the%20weight%20of%20the%20VDI%20for%20all%20new%20sessions%20to%20be%20redirected%20to%20a%20new%20VDI%2C%20while%20old%20sessions%20would%20stick%20with%20old%20VDI.%20Would%20that%20work%20%3F%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3EAny%20recommendations%20%3F%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3EThanks%2C%3C%2FP%3E%0A%3CP%3EMichal%3C%2FP%3E%3C%2FLINGO-BODY%3E%3CLINGO-LABS%20id%3D%22lingo-labs-140714%22%20slang%3D%22en-US%22%3E%3CLINGO-LABEL%3ELoad%20Balancer%3C%2FLINGO-LABEL%3E%3CLINGO-LABEL%3ETraffic%20Manager%3C%2FLINGO-LABEL%3E%3C%2FLINGO-LABS%3E%3CLINGO-SUB%20id%3D%22lingo-sub-141122%22%20slang%3D%22en-US%22%3ERe%3A%20Traffic%20Manager%20vs%20Load%20Balancer%20for%20RDP%20sessions%3C%2FLINGO-SUB%3E%3CLINGO-BODY%20id%3D%22lingo-body-141122%22%20slang%3D%22en-US%22%3E%3CP%3EHi%20Cliff%2C%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3EThank%20you%26nbsp%3B-%20that%20is%20interesting%20option.%20It%20would%20not%20work%20for%20me%20(i%20need%20to%20destroy%20VM%2C%20not%20to%20manage%20it%20centrally%2Fclear%20home%20folders).%20To%20solve%20my%20problem%20i%20am%20also%20considering%20other%20set%20of%20solutions%2C%20for%20virtual%20desktops%3A%20not%20to%20use%20Windows%20Terminal%20Server%20but%20Linux%20with%20LXC%20containers%20and%20x11%20apps.%20This%20way%20i%20will%20have%20separation%20on%20container%20level%20and%20would%20not%20have%20to%20destroy%20the%20machine%20(but%20container%20instead).%3C%2FP%3E%0A%3CP%3EWould%20love%20to%20see%20similar%20solution%26nbsp%3Bfrom%20Microsoft.%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%0A%3CP%3EThanks%2C%3C%2FP%3E%0A%3CP%3EMichal%3C%2FP%3E%3C%2FLINGO-BODY%3E%3CLINGO-SUB%20id%3D%22lingo-sub-140981%22%20slang%3D%22en-US%22%3ERe%3A%20Traffic%20Manager%20vs%20Load%20Balancer%20for%20RDP%20sessions%3C%2FLINGO-SUB%3E%3CLINGO-BODY%20id%3D%22lingo-body-140981%22%20slang%3D%22en-US%22%3E%3CP%3EHave%20you%20investigated%20WSMAN%20%3CA%20href%3D%22https%3A%2F%2Fdocs.microsoft.com%2Fen-us%2Fpowershell%2Fmodule%2Fmicrosoft.wsman.management%2Fproviders%2Fwsman-provider%3Fview%3Dpowershell-5.1%22%20target%3D%22_blank%22%20rel%3D%22noopener%20noreferrer%22%3Ehttps%3A%2F%2Fdocs.microsoft.com%2Fen-us%2Fpowershell%2Fmodule%2Fmicrosoft.wsman.management%2Fproviders%2Fwsman-provider%3Fview%3Dpowershell-5.1%3C%2FA%3E%3C%2FP%3E%0A%3CP%3E%26nbsp%3B%3C%2FP%3E%3C%2FLINGO-BODY%3E
Occasional Contributor

Hello Azure Team,

 

My challenge:

- i want to provide access to Virtual Desktops (VDI) deployed in Azure 

- i have 3 regions: us, emea, apac

- in each region i have 3 shifts (10h each, with 2h overlap), so each region provides 24/7 support

- i need to reinitialize each VDI every 24h (redeploy new Windows Terminal Server to make sure no customer data is there for more then 24h)

 

What would be the right design ?

 

I was thinking to use Traffic Manager (DNS loadbalancing) nested profiles:

- global profile

- nested emea profile

- nested us profile

- nested apac profile

Then another layer of nested profile inside each region (3 shifts per region). But the challenge is in overlaps. I need to make sure my shift2 from emea starting 2 hours before shift1 in emea is finished can work correctly. If i will switchover on the beginning of overlap my shift1 will get a new DNS A responses and their RDP session could be redirected to a new VDI (they would lost all data). I was thinking to increase DNS TTL timer to 2 hours - but that looks like a can of worms (i would not have failover if specific VDI is going down).

 

Another option is not to do DNS loadbalancing but application Load Balancing (using LoadBalancer). But i do face similar challenges - how to plan overlapping shifts (each region is having all 3 shifts).

I would like to use Load Balancing with sticky option - this way existing RDP sessions would be redirected to the same VDI, at the beginning of the overlap i would change the weight of the VDI for all new sessions to be redirected to a new VDI, while old sessions would stick with old VDI. Would that work ?

 

Any recommendations ?

 

Thanks,

Michal

2 Replies

Hi Cliff,

 

Thank you - that is interesting option. It would not work for me (i need to destroy VM, not to manage it centrally/clear home folders). To solve my problem i am also considering other set of solutions, for virtual desktops: not to use Windows Terminal Server but Linux with LXC containers and x11 apps. This way i will have separation on container level and would not have to destroy the machine (but container instead).

Would love to see similar solution from Microsoft.

 

Thanks,

Michal