Packet, the bare-metal cloud computing startup, is jumping on the edge networking train by announcing a new edge compute service with 11 locations. The service is intended for developers that need infrastructure without virtualization or multi-tenancy.
Packet, which received a $9.4 million investment from SoftBank in 2016, makes a bare metal server technology that automates physical servers and networks and provides on-demand compute and connectivity.
The company’s new edge compute locations include Los Angeles; Seattle; Dallas; Chicago; Ashburn, Virginia; Atlanta; Toronto; Frankfurt; Singapore; Hong Kong; and Sydney. These locations join existing cities New York; Sunnyvale, California; Amsterdam; and Tokyo. Additional cities such as Paris; London; Sao Paulo; and Mumbai will be available in October.
The edge compute locations will feature a single server configuration known as a Type 1E instance based on an Intel SkyLake processor but will also have Packet’s full networking and infrastructure automation suite.
According to Packet CEO Zackary Smith, the goal is to move the compute power closer to the developers. “We’re going where the networks already exist,” Smith said, adding that by offering no networking overlay or multi-tenancy, developers can quickly deploy an instance using standard DevOps tools or the Packet portal. Customers can even build a custom layer 2 or software-defined wide area network (SD-WAN) across Packet’s transport network.
Smith said Packet is already seeing major operators experimenting with deploying their evolved packet core (EPC) on bare metal because this can lower their costs and reduce latency.
He also said that Packet envisions that the number of edge compute locations will increase from 10 to 15 to 50 to 100 locations. “If we can get it to 100 locations, that is compelling,” he said.
Smith added that he believes edge computing is in the early stages and he thinks Packet’s approach is more practical for developers because the service lets them get access to a dedicated bare metal server configuration in less than eight minutes.
He also argues that some other edge compute initiatives, such as that of data center startup Vapor IO, which intends to open micro-data centers at the base of a cell tower, are not as practical for developers because developers don’t just want access to a particular carrier on a certain tower but to all operators. “Edge compute will be finite and it will have limits. Multi-tenants make it hard to predict.”