It may
seem hard to believe, but even in this day and age, there are still
some enterprises that are cloud-averse. For whatever reason, they are
hesitant to have their applications or data in the cloud and prefer to
host (almost) everything in their own on-premises data center. While
this strategy works for some, it’s getting harder to adhere to, mostly
because there are some essential applications that have no option for
on-premises hosting. Or, that option is just so inconvenient that it’s
almost not worth the trouble.
For example, suppose your
organization wants to use an industry-leading CRM application like
Salesforce. Well, there is no on-premises version of this SaaS
application. If this is the CRM your company wants to (or must) use,
then a cloud deployment is the only viable option. Then, there’s
Microsoft Exchange. Managing Exchange Server on premises is challenging
at best, with some customers complaining they aren’t getting the support
they need from Microsoft. For its part, Microsoft would prefer that
customers migrate to the cloud-based Office 365 for the best overall
experience with the most up-to-date features and functions. The bottom
line is this: enterprise applications increasingly are hosted in the
cloud, and organizations need to adapt to a WAN architecture that
includes the cloud as a primary location for doing business.
This is a fundamental issue for enterprises that still use a legacy
MPLS-based, hub-and-spoke WAN architecture. That is, the corporate data
center is located on-premises at some central location. Branch offices
are generally connected via MPLS circuits to bring their traffic into
that data center to access both local applications as well as
cloud-based applications—and often for security as well. The only egress
point to the cloud is through that central data center. Thus, branch
traffic is “hair-pinned” because it is brought from the branch, to the
data center, to the cloud, and then reversed to get back to the branch.
Needless to say, this adds a lot of latency and causes performance
issues for the applications, which in turn leads to unhappy users who
complain about slow applications.
A
great way to get around this architectural limitation is to deploy
regional CloudHubs, which many believe, this writer included, are the
future of WAN architecture.
CloudHubs
are virtual data centers that consist of racks of switching and routing
equipment deployed in carrier-neutral, colocation data centers, such as
those operated by Equinix and CyrusOne. Then, these CloudHubs are
interconnected with high-capacity, low-latency circuits that create a
high-performance core network. At the edge of this network, an
enterprise can interconnect using its existing carrier services, whether
they are MPLS networks, dedicated internet access networks, Ethernet,
or private line services. The organization also can directly connect its
branch offices, remote and mobile users, partners, and of course, its
own data centers to the core network. This model essentially brings the
enterprise into the cloud with virtual data centers that help create a
distributed infrastructure.
Suppose an enterprise has its
corporate headquarters in Dallas and branch offices in Los Angeles,
Atlanta, New York and Chicago. The enterprise can either create its own
CloudHubs at colo data centers in those cities, or it can rent space in
an existing CloudHub. Either way, having a network presence in a series
of local CloudHubs allows the enterprise to peer with various
cloud-based applications and platforms. In most instances, CloudHubs are
one hop or one cross-connect away from SaaS and IaaS providers. Peering
means that traffic travels much shorter distances, which in turn vastly
reduces latency and increases performance of cloud apps.
There
are numerous advantages to this approach. Using a CloudHub as a core
component of the WAN allows an enterprise to reach SaaS and IaaS
directly at the regions in which their instances are hosted. Also,
consider the way that applications are constructed today, using microservices
in containers that are widely distributed in different locations. The
components of an application talk to each other through APIs, or
application programming interfaces. To be able to keep up with this new
way of constructing and running applications, an enterprise’s
infrastructure needs to be distributed as well, which is what CloudHubs
enable.
For example, Microsoft Office 365 runs on a variety of
microservices. It doesn’t all run in an application sitting in a
specific place. Companies need to be able to peer with Microsoft in
different regions to be able to get to these microservice containers in
the fastest way possible.
Having access to distributed CloudHubs
can allow enterprises to securely connect to SaaS and IaaS via fiber
connections, which can eliminate the need for advanced security. Also,
CloudHubs offer bandwidth on demand which can be used to facilitate
cloud migration and backups as well as business growth.
CloudHub Building- Pre-built CloudHubs are ready for use
Now,
back to that comment about building or renting a series of CloudHubs.
An enterprise can go to any collocation data center and install a rack
and power. Then, they buy and install networking infrastructure like
routing, switching and security. They deploy and manage all this
themselves. They put this same sort of infrastructure in a series of
collocation data centers and then interconnect them with high-speed
communications. This forms the core of the network. If expansion is
needed, the company can go to more colo centers whenever the service is
needed.
Some companies offer an alternative to the DIY model.
These are CloudHubs located in dozens of locations around the world.
They can be multi-tenant or single-tenant.
Additionally, they are fully functional and ready for move-in. An
enterprise can simply rent capacity and be up and running in no time,
and all maintenance and operational activities are done by the provider.
This model provides organizations with the agility and flexibility they
need so that they can add or move sites frequently, such as through
M&A activity or business growth.
Either way – pre-built or
DIY – CloudHubs are a vital component for enterprises to ensure optimal
performance of cloud-based applications.
Mark Casey is CEO and founder of Apcela. Mark is a progressive leader focused on leveraging emerging technologies and his deep knowledge of the global telecom and IT markets to drive business results for his clients. Over 25 years, Mark has played an instrumental role in complex network optimization and digital transformation programs for Global Tier-1 carriers and Fortune 500 companies. Under his leadership, Apcela pioneered the development of an award-winning high performance, low-latency global cloud platform.