Future Networks

One line is not enough

Connect the router and that’s it? Unfortunately not. Anyone wishing to connect hybrid cloud services with the company network should take four points into account.

Connecting hybrid cloud services to the company network is more complicated than many IT experts think. There are four key factors for the technical connection.

Replies in a telephone conference arrive with a delay, or the picture in a video conference is choppy. Many companies that obtain voice and video services from a cloud encounter such problems. The cause of that often does not lie in the cloud data center, but in “cloud connectivity,” i.e. the network connection between the cloud and user. Connecting a cloud to the company network (WAN) is anything but a trivial matter – and even more so if a company pursues a hybrid cloud strategy, i.e. uses private and public clouds. The technical requirements for networking depend on four factors.

1. Video needs speed.

Anyone only wanting to obtain applications from the cloud that are non-critical as regards transmission, such as e-mail services or backups, does not need to worry. They tend to be insensitive to loss and delays of data packets and so work with every type of cloud connectivity. The very opposite is the case for connecting cloud ERP (enterprise resource planning) systems or real-time applications, such as video and IP telephony from the cloud: they place very high demands on data transmission. For voice, for example, the standardization organization ITU-T recommends a packet loss rate of less than five percent and delays below 150 milliseconds. To keep both of them low, not only the technology used for cloud networking plays a role, but also the geographical conditions.

2. The Earth is large.

You might think that if an international company uses an external cloud, it only needs one connection to the cloud provider’s network. Yet it’s not that simple. Big cloud operators usually distribute their data centers over different regions. A user in Sydney accesses data storage in Australia, and a user in London accesses it in Western Europe. The reason: the shorter the route the data takes, the shorter the transmission (latency), which means that real-time applications in particular work better.
This geographical distribution would not make sense if a company just created one connection to the cloud, for example in Frankfurt, Germany. The employee in Sydney would then access the data center in Australia via Frankfurt – and send data around the globe. Latencies of more than 360 milliseconds would be a certainty. That is why the company network needs several regional connections to the cloud provider’s network: at least one per continent, with additional branches.

3. A cloud seldom comes alone.

Yet that is not all: According to the Cicero Group, almost three-quarters of large companies worldwide cooperate with multiple cloud providers. That means they need to provide several regional connections for each provider. Accomplishing that requires a bit of prior know-how, since each provider implements the connection in a technically different way, has its own security and configuration specifications, and has its own charging model. One example: whereas some offer room for the customer’s routers in their own data center, others only connect their network to a colocation provider who installs the customer’s routers at its premises. Things might be easier in future for companies whose network service providers offer preconfigured edge routers for networking with the big cloud operators.

4. Standard is not enough.

Different transmission technologies are used depending on the type of cloud (private or public). To date, the public cloud has mostly used virtual private networks (VPNs) over the Internet. However, the Internet network quality is not sufficient for time-critical applications. This is why more and more companies are using powerful MPLS connections for access, such as in the case of the ExpressRoute offerings from Microsoft or other private connections. Alternatively, they choose special Internet routing that optimizes packet loss and latencies.
In contrast, private clouds are networked almost exclusively via MPLS VPN anyway. Ethernet or Layer 1 services are used only if the quality or bandwidth requirements are even higher, for example in order to connect a cloud location to headquarters or to interconnect multiple data centers.

Crucial for success

Up to now, many companies have attached too little importance to network aspects in implementing hybrid cloud sourcing. This has to change. After all, a hybrid cloud strategy can only succeed if all cloud services are connected to the company network quickly, securely and with adequate quality and cost-effectiveness.

Further articles