First of all, I want to say that I've been deploying Exchange for years, since the days of Exchange 5.5, from 15 years ago. I've never had any real connectivity issues unless the in-house server crash.... well, that is until now. I've deployed many on-premise Exchange 2010 servers and this is the first "in the cloud" hosted Exchange solution that I've inherited from someone who deployed it. There has been nothing but connectivity issues + slowness problems.
- Using the same Windows user account logged into the same laptop/pc, I can use a different Outlook profile and open a mailbox against an on-premise Exchange 2010 server and it has NO issues whatsoever. Launching the mailbox profile pointed to Office 365 would sometimes takes forever, especially between 9 and 11am... umm, that is business hours and it appears that the CAS or MB servers are swampped on the Microsoft's side because sometimes I get repeated login prompts even though I know my password is correct after I verify it via OWA.
- The same symptom occurs when I simply try to view the Outlook profile for Office 365, and sometimes it times out completely and says "The server is offline, Retry, Work Offline, or Cance." In my experience, the Exchange server is either down or there is no connectivity to the internet. I'm on a 50MB down/10MB up Comcast highspeed cable connection and I have no issues connecting to other clients' on-premise Exchange 2010 servers. I get the same result using the same Outlook profile in different networks, some as fast as a 100MB fiber connection. So Microsoft??
- I initially thought the "trying to connect" message is an Outlook profile + OST cache file issue but it is not. I deleted the profile and removed the several xml files related to autodiscover process and rebuilt both the autodiscover + OAB + OST file from scratch. I get the same initial slowness during login + sporadic drops which results in "trying to connect" error message in Outlook. I should add that the autodiscover process is fine because "test email auto-configuration" results come back fine.
- I do a nslookup against outlook.office365.com and it yields 6 public IP addresses. If I continue to do nslookups, I get a round-robin DNS return that continuously rotates the DNS entries. This is probably not the way to load balance your farm of Exchange CAS servers. How about using the same set of IP addresses but use hardware load balancers behind which check on actual server load before shooting an user over? Arbitrarily shooting a connection to a busy server would yield a time-out!
I'm an admin for 15+ Exchange 2010 servers for several different clients of ours and this is the worst connectivity/network issue I've seen with Exchange. I get a daily complaint from these guys about their Outlook saying it's "trying to connect" right around 10am and 3pm, which is about when people are the busiest working.
My theroy:
1. the Exchange CAS server farm or the backend MB servers are gettting swampped
2. There isn't enough bandwidth on Microsoft's side to handle the millions of connections
I know I can't be the only one seeing this problem. I'm trying to gather info from others to find out if you are seeing the same. We need a group push to fix this or I'm taking them off to an on-premise Exchange server.