active active setup across 5 sites

Where exactly is the delay in the session process?  Database access?  Latency to target endpoint?  Latency between user & PS node? 

When loading and logging into the BI console , users face delay , and also when the RDP is initiated , they are facing delay and lags are noticed within the session. So even when in the site where management node is deployed, the lag is seen. But ll try to isolate further like you have mentioned and update here..

Hi ​@SFA 

This is a tough one to give you an answer to, there are so many factors that can impact the performance.

Network traffic, Load Balancing, SQL AG performance. end-user location, Workgroup configuration to avoid traffic crossing from DC to DC, UVMs dedicated resources etc.

The list is big, and without any knowledge of your architecture, resources etc. it is difficult to tell.

KR Jens.

Hi All,

Thank you for responding. We are also currently mainly going through the setup as we didnt do the deployment for the customer.

They have 10 appliances. 2 appliances with SQL in one site , one of them is the management node. Other is worker node.

In other 4 sites , we have 2 SQL free worker nodes each .

The SQL always on cluster , primary is one site (closer to the management node) - in one country. and all other sites across 4 countries are async nodes. There is loadbalancer in each site. With the 2 worker nodes behind it for the other site.

When user in Site 2 access the portal , it points to the LB VIP , then it directs traffic to one of worker nodes, the AD authentication goes to each country domain but the loading of the page after authentication takes time. So ideally , request goes to the site and site domain controller but i believe the DB write would happen for the audits etc into the always on cluster via listener to the primary node in the site 1?

The session launched , even when the specific worker node is chosen, the RDP session loading is lagged . So here the DB write would be causing latency as the main node is in Site 1? is it possible to help understand how the flow would be in this scenario?

There is no DB access once the RDP/SSH session is connected to the target endpoint. If there is real session / screen-refresh latency, that is pure network lag / routing. 

Logon page latency could likely be the BIND account connecting across countries to a random domain controller. 

For SQL always-on clusters, yes, the primary node is the read/write node and all writes must go to it.  The other nodes are just read-only.  Database access over WAN will always be slow, but this should just be during initial connection/authentication. 

Do you have load balancer VIPs for each country’s users?  Or is your LB smart enough to direct them appropriately? 

there is LB for each country site. Also when other sites access passwordsafe , it goes only to the worker node right? There are multiple domains added and user will choose their respective site domain, so the likeliness of issue because of bind account is less.

“Database access over WAN will always be slow, but this should just be during initial connection/authentication. “ - this is actually where the delay is..in the initial login and initial RDP login. So in all this , the management node is never coming to the picture for the other sites right? Also i noticed that multi failover subnet was turned off as well. could that also add to the latentcy?

...the management node is never coming to the picture for the other sites right?  Correct, only the worker node the user is connected to.

If your SQL Always-On is sync-mode, then I would have DBA’s check query performance and network latency to the active DB node.  You’ll want as much of your database instance memory and on fast SSD storage, including TempDB.