How to Optimize Your Office 365 Performance with Network Peering

This article was originally published on CloudTweaks on June 19 2019.

Microsoft Office 365 usage has grown significantly in recent years. More than 56 percent of organizations all around the world now use Office 365 cloud-based services. In terms of user count, Office 365 is the world’s most widely used cloud service, with one in five corporate employees now using at least one of the applications.

It’s great that so many companies are getting the benefits from this set of productivity tools—but are they getting the best possible performance from their applications? Many business users of the applications complain of having poor user experiences, either with slow response time or audio or video interactions that “break up.” When this happens, it’s often the company’s own network causing the performance issues.

Gartner recently surveyed its customers that use Office 365 and 20 percent reported experiencing a networking problem. Another 22 percent reported performance issues where the network was a possible cause. Thus, it’s important that companies properly prepare their network connectivity to reduce or eliminate such issues and to optimize their Office 365 performance.

Traditional hub-and-spoke networks weren’t designed to support a fast user experience for cloud applications. To resolve performance and security issues, enterprises often must re-think their network design.

How to Optimize Your Office 365 Performance with Network Peering

The Office 365 Network Connectivity Principles

How Office 365 is delivered to users has evolved significantly since the product’s inception, and it’s still undergoing a lot of transformation. What started off as single-user access and ecosystem of a company’s Office connectivity needs is now comprised of dozens of global, micro-services delivered over a distributed content delivery network (CDN). When a user launches an Outlook client to access their email, or uses Skype audio/video services, or uses SharePoint or OneDrive, their connection from the end user device is going to multiple different cloud-based content delivery networks (CDNs) at any given point in time. These CDNs that serve up various components of Office 365 run in some 120 Microsoft datacenters around the world.

According to Microsoft, the most significant factors that determine the quality of the Office 365 end user experience is network reliability and low latency between the local Office 365 clients and the highly distributed Office 365 service front doors, which are entry points to the Global Microsoft Network. This is Microsoft’s public network backbone that interconnects all of Microsoft’s datacenters with low latency and cloud application entry points (i.e., the service front doors) throughout the world. There are over a hundred locations of these service front doors worldwide to ensure that Microsoft customers are always close to one of these ingress points.

Interestingly enough, many end users have their best performance experience with Office 365 applications when they are on their home networks because of the low latency between the home Internet connection and a nearby Office 365 service front door. For most companies, however, this isn’t a viable option due to security concerns and regulatory compliance requirements. Instead, organizations have to rethink their corporate network topology to optimize application performance between the business office locations and the Microsoft CDNs.

Microsoft outlines its network connectivity principles to Office 365 and says the primary goal in the network design should be to minimize latency by reducing the round-trip time (RTT) from an enterprise’s network into the Microsoft Global Network.

The main challenge associated with companies’ deployment of Office 365 is that they backhaul the traffic from their end users to their core data centers where they have their security stacks and their Internet gateway connectivity. The traffic runs through the security stack proxies, next-gen firewalls, IDS, IPS, data loss prevention, etc., and then out to the content delivery network.

For many enterprises, their premises-based data centers are not optimized for peering with Microsoft’s peering points, so all this backhauling and hair-pinning of traffic creates sub-optimal performance in terms of latency and resulting application performance. Congestion on the segments that the traffic traverses also plays a critical role in performance, and hub and spoke architectures of traditional customer deployments are one of the biggest bottlenecks. Without optimization techniques that deliver the traffic going from a user’s workstation to the closest service front door, the traffic will be impacted by latency as it gets sent around the public Internet.

Distributed cloud hubs provide a better topology approach

A better approach is somewhere between the two extremes, the first being for users to use their own Internet connectivity, which lacks the proper security posture, and the second being the backhauling of traffic to the data center, where there’s good security but long latency. A third and much more efficient approach is where the enterprise opts for some sort of distributed cloud-based gateways – call them cloud hubs – where the security stack as well as the Internet connectivity are both closer to where the end users are. Users connect to the closest cloud hub and their traffic goes through the same security stack that the core data center would offer, hence maintaining the organization’s security needs and compliance requirements, and then accesses the Microsoft peering points.

This cloud hub approach not only provides the required security and cuts down on traffic hair pinning but it also distributes Internet capacity across multiple regions, thus alleviating the bottleneck of the throughput at a data center. In the cloud hub topology, the enterprise has multiple cloud-based data centers to choose from, each with sufficient capacity to serve the users’ needs and giving the traffic a relatively short path to the Microsoft content delivery network.

If these cloud hubs are in vendor-neutral data centers (as opposed to within a typical Tier 1 MPLS provider), they are likely to share residency with the Microsoft CDNs. This further reduces the length of fiber between the cloud hub and the CDN where the Office 365 components are served, reducing latency to milliseconds or even microseconds.

All in all, an enterprise’s instance of Office 365 will benefit tremendously from these performance optimization strategies by 1) bringing the data center closer to the end users, and 2) having those data centers be in the same proximity as a Microsoft CDN. This approach maintains a good user experience, and at the same time, maintains the security and compliance requirements for the organization.

How can we help? 

We love talking about software-defined networks and the cloud! Let us know if we can help by filling out the form. Cheers!