Forum Discussion
How does Microsoft distribute the data between all their data centers?
- At the moment tenants are homed in a region, in your case this would seem to be Emea, and those are the data centres. My tenant is in Dublin and Amsterdam. I doubt it makes much performance difference, it's more about regulatory requirements.
However, I am confident that there would be better performance if, for instance, users from Singapore have to connect to the servers in Singapore instead of the ones in EMEA.
Thanks!
- Microsoft's network connects to the internet in many locations, you won't come all the way round the internet to get to an EU DC, you'll pick up their wan somewhere more local. Once you get on to it then you get a pretty standard level of performance. Remember that there is a CDN that is replicating those core files like sharepoint scripts and applications like Outlook work entirely cached these days, no user response is waiting on servers.
Most latency comes from local network, local ISPs and their peering arrangement, the physics of distance make little effect. Nothing on earth is more than 20,000Km away, light in fibre can do that in less than 0.1seconds.I agree that Microsoft has a very strong peering around the world hence it should take no more than a couple of hops before entering their backbone network. However, even Microsoft cannot overcome the laws of physics and the speed of light. If you take the direct distance between Singapore and Amsterdam, it's about 10,500km which would take about 35ms for light to travel in a vacuum or about 70ms round-trip. In reality though, it's unlikely that the round-trip latency would fall below 150-160ms at best.
I also agree that some contents are pushed out to the edge via CDN. For example, those icons that appear in the browser when using OWA are delivered via the CDN. Yet, almost nothing comes from the CDN when using the Outlook client. On top of this, the protocol used by Outlook to connect to Exchange Online is either MAPI/HTTP or Outlook Anywhere and both protocols are more sensitive to latency than standard HTTP(S) that OWA uses. Finally, while it's true that cache mode helps with the sending/receiving emails, not everything within Outlook uses cache mode. For example, access to Online Archive does not use cache mode hence latency affects performance. It's a known issue that large Shared Mailbox doesn't work well in cache mode and latency also comes into play here.
Just to give you an idea, sending a 10MB attachment on a LAN took 9s while it took 58s with a WAN latency of 60ms.
Going back to Aldin's comment about performance improvement in moving the mailboxes to their regions - yes, this is definitely true but this isn't something that's currently available for the majority of the O365 customers. I've worked with large customers who are O365-Dedicated customers and it's indeed possible to move the mailboxes closer to the users. However, there is also a downside to this in terms of data sovereignty and what happens when the user travels around the world.
Last of all, the company that I work for, Riverbed Technology, has a solution known as SteelHead SaaS that can ensure the user's performance regardless of where they are in the world. Feel free to reach out to me if you're interested in learning more about this solution.
Thanks,
Blanco
- Blanco, There's more exchange specific data with reference Le method here https://blogs.technet.microsoft.com/neiljohn/2011/06/15/outlook-2010-network-latency-test-results/. We can't really judge the latency across Microsofts network, but expect it to be good, that data shows that on the most data intensive tasks it might be a 2x or 3x impact. Mapi/https is perhaps likely to improve this. Outlook since 2013 manages connection well in that the application won't lock, and remote mailboxes are far better than they used to be in 2007. Personally I've worked from 5000 miles away and seen little impact on normal use of any service, and I'm not getting performance related issues raised by staff half the world away.
