In many cases, the data center a customer wants to reach an AWS Local Zone from is a self-built data center in its own building or perhaps in a colocation facility that does not happen to be an AWS Direct Connect Location. In cases such as this, an AWS Direct Connect Partner can offer what is known as a hosted connection:
Figure 6.12 – AWS Direct Connect via partner Centurylink in Berlin
Hosted connections offer several benefits over dedicated connections:
Simplicity: Customers can rely on the partner to handle the administrative and operational aspects of the physical connection, simplifying the setup process.
Incremental scaling: Hosted connections offer a range of bandwidth options, starting from 50 Mbps and going up to 10 Gbps. This flexibility allows customers to choose the right capacity for their workloads and scale it up in line with throughput demand.
Speed of provisioning: As they are managed by partners who have pre-established infrastructures in the appropriate locations, hosted connections tend to be set up faster.
Cost-effective for lower throughput: Organizations that do not need the full capacity of a dedicated connection can opt for smaller bandwidth increments, which is typically more cost-effective even though a hosted connection costs slightly more.
Cost difference for a hosted connection
On a per-Gbps basis, hosted connections do cost slightly more than dedicated connections. However, this difference is small enough that it usually does not impact a customer’s decision except at the highest of scales.
Continuing with the two examples mentioned previously of a customer connecting to the Berlin zone, let’s assume the connection speed is 1 Gbps and the customer transfers 50 TB in and 50 TB out every month. The costs per month in USD are as follows:
| Port cost (1 gigabit per second) | Transfer cost (50TB) | Total | |
| Dedicated | $219.00 | $1,024.00 | $1,243.00 |
| Hosted | $240.90 | $1,024.00 | $1,264.90 |
Figure 6.13 – Simple comparison of hosted and Direct Connection costs
Accounting for spikes and overhead, this much transfer is essentially utilizing the entire 1 Gbps connection for the entirety of the month. What happens when the customer’s throughput needs scaling to 1.5 Gbps? What if the customer’s utilization pattern is sporadic and they need to account for 95th percentile spikes in the 2.7 Gbps range? In those cases, the small savings from a dedicated connection are more than cancelled out by the lack of granularity.