The notion that compute nodes can be deployed anywhere, ranging from a data center to a donut shop to the top of a light pole or the bottom of the ocean. An often overlooked consideration for use cases is the significant difficulties of deploying, managing and servicing the hardware itself in remote locations. While we can be reasonably sure that the hardware deployed in a data center will be well cared for, many remote locations have substantial hardware and environmental constraints. In this session we will share Verizon’s experience in packaging and deploying the uCPE device hardware in support of its Virtual Network Service product. Some of the factors that went into the hardware decision included:
- Global supportability – Just how fast can hardware be replaced in remote locations?
- Cost versa functionality tradeoffs
- Total Cost of Ownership (TCO) – It isn’t always about the lowest initial cost.
Deploying applications in remote locations is not just software and network, in many use cases the form factor of the hardware component itself will determine the success or failure of the project. Some of the hard lessons we will share include:
- Hardware/Software Disaggregation - How generic is generic x86 really in an edge context? Can you assume that the software will work? What is the best way to avoid integration nightmares when there is a mix of in-house and multiple vendor applications?
- Vendor Lock-in - When you are not a hardware manufacturer, is this even possible? Is it better to buy an existing COTS product or build a specialized box from parts?
- Support Contracts – What they mean and why they are critical to the success of the project.
- Continuous integration in a hardware context -- Making sure it all works together has a different meaning when you DO need to worry about the hardware.
- High Availability considerations in remote locations – Which hardware is critical and what needs to be redundant.
