6. Telecom & 5G
Edge use cases
|
|
Multi-access Edge Computing (MEC)
5G has the following effects from the tower to the wireless device:
- 100X increase peak data rate
- 50X increased density of devices
- 50X decrease in latency, down to 1ms
However, 5G does nothing for the network backbone from the towers to the data centers. As such, 5G could easily overwhelm the fibre networks, the data centers and the cloud.
The solution is to move compute power from the data center to the base of the tower to handle caching, preprocessing and local processing. This type of work is called multi-access edge computing (MEC).
The actual specific workloads and the economics for MEC are as yet unknown. Thus, investing now in data centers at the base of towers, even single rack solutions, is expensive and risky. One may find out that the hardware selected before engaging the market is completely wrong for the actual workload.
Edgecell is inexpensive, linearly scalable, low power, remotely managed platform as a service (PaaS). Edgecell is the first viable platform for deploying MEC to thousands of locations. As a PaaS, Edgecell is an operational expense (OPEX), not a capital expense (CAPEX). Hardware is not purchased and can be easily replaced to meet the changing needs of workloads. As such, Edgecell is a low risk means for the 5G community to deploy MEC now and rapidly evolve as it learns the specific workloads and economics of MEC.
Scalable edge computing
Edgecell is the first server designed for the true edge. We designed and built Edegcell out of necessity. We tried to use off the shelf servers and single board computers to deploy distributed software such as Kubernetes at the edge but found nothing that would meet the requirements at scale.
The conditions for the true edge are as follows:
- No data closet (or data closet is full, overheating and underpowered)
- No technical staff onsite
- Hundreds or thousands of locations
- Potential limited or intermittent network bandwidth
- Demand for compute power always increasing
- Many novel opportunities to apply machine learning
Edgecell meets these requirements with the following features:
- No special equipment needed (such as airconditioning or top of rack routers)
- Very low energy consumption
- No echnicians required
- Start with smallest footprint
- Easily add more computepower
- Run distributed software
- Run machine learning models
- Remotely monitor, manage and upgrade
In particular, we had three fundamental design objectives for Edgecell: simplicity, scalability and remote management.
Simplicity
Simplicity we defined as follows:
“If you can delivery a pizza, then you should be able to install an Edgecell device.”
If you look at the back of most Edgecell devices, there are only two ports: power and ethernet. Any other server or single board computer has multiple ports. Each port exponentially increases the possibility of someone not technical making a mistake in installing.
No USB or video ports also means that no one can take a Edgecell and use it for their own personal use. A Edgecell is a server, period. All control of the Edgecell is remote and centralized. No one at the edge should do anything to configure a Edgecell. As such, no one can.
Scalability
Linear scalability is the ability to incrementally increase the available compute power without an additional sunk cost. Edgecell is designed for distributed software which by its nature runs on multiple servers. On the cloud, one pushes a button to add more servers to a cluster. We needed that same capability at the edge.
We assessed previous attempts at scalable servers. Blade servers were meant to be a linear scalable solution, but in practice no one ever bought a blade chassis without all the blades already installed. There are two reasons for blades not scaling linearly as intended:
- The sunk cost of a chassis
- The technical skills necessary to access in the data center
Remote management
Edgecell has a built in provisioning system that enables us to remotely install and update software, down to the driver and operating system level, even when the Edgecell is behind a firewall.
Edgecell GitOps enables update to the operating system and platform (eg. Kubernetes) to thousands of devices. It also enables customers to deploy application (container) updates to thousands of locations.
Edgecell has an event-driven architecture designed to scale. It is multi-tenant and can be deployed in a shared-nothing environment. That means large customers can have their own Edgecell installation and provide multi-tenant access to devices to their customers.
Edgecell as MEC
Edgecell serves as multi-access edge computing (MEC) in three key areas:
- The micro tower
- The enterprise campus
- The remote office, branch office, small office and home office (ROBO SOHO)
We will review each area here.
Micro tower
Edgecell works at the edge on the carrier side of the network. Edgecell Edge platform as a service (PaaS) for Kubernetes provides a cluster at any location that is:
- Remotely managed
- Highly available
- Easily installed
- Easily scalable
- Able to run machine learning models
- Cost effective
- Low energy
- Small
Edgecell can be deployed not only at towers but also micro cells.
To build a highly available Kubernetes cluster at the edge with traditional hardware is expensive. Even micro data centers such as AWS Outpost cost more than $500,000 each per year to operate. They require an additional construction of housing, power and cooling infrastructure.
Edgecell enables companies to deploy a highly available Kubernetes cluster at a fraction of the cost. Plus, it requires no special infrastructure. It can be placed in a box the size of a picnic cooler. That means systems can be deployed in a fraction of the time and without the need for expensive technicians.
Edgecell is an immediate, low risk and inexpensive means for wireless network providers and their partners to experiment and learn the specific workloads and economics of MEC. With remote management built in, deploying and upgrading software to the edge is as easy as deploying to the cloud, so customers can rapidly evolve the software as they learn what is most effective at the edge. Since Edgecell is a PaaS, hardware is not purchased and can be easily replaced to meet the changing needs of workloads as well.
Enterprise campus
5G holds the promise to change how corporate campuses and factories deploy networking. Laptops, tablets, smartphones and IOT devices can use the same wireless networking technology whether they are in the office, on the factory floor or moving across town in a vehicle.
Companies are accustomed to controlling the wireless and Ethernet networks on their campuses. The advantages of 5G come with the challenges of retaining security, efficiency and control. MEC offers an opportunity for companies to overcome these challenges.
One use case in particular is the ability to route corporate bound internet traffic on the 5G network so that it never leaves the corporate network. In this scenario, the company deploys its own micro cells. Each cell has Edgecell MEC. The MEC optimizes the internet traffic. Any packets for the company’s network, such as internal business applications, route directly to the company network and never leave the campus. All other traffic goes to the carrier network as usual.
The Edgecells also cache requests from the carrier network to reduce the bandwidth. For instance, if there is an update for a third party smartphone application that employees use, that update is only requested once from the carrier network. All subsequent downloads go directly from the MEC cache at the tower to the devices on campus.
The result is a win-win for the corporate customers and the wireless carriers. The corporate users have the bandwidth, low latency and ease of use of 5G. The corporate IT department maintains a portion of the security, efficiency and control they are accustomed to with wifi networks. The wireless carriers reduce a portion of the traffic off the network backhaul.
This use case is an immediate and obvious one. As enterprise customers expand their use of 5G networking and explore “the power of the possible,” Edgecell enables them to easily expand the compute power to support those new use cases.
ROBO SOHO
On the customer side, Edgecell has an enterprise role for the remote office, branch office, small office and home office (ROBO SOHO). Virtual desktop infrastructure (VDI) is a critical solution for companies to provide their employees the access they need to internal systems while maintaining strict security monitoring and control. VDI access from ROBO SOHO to a central data center has its frustrations, especially with latency. Now, with the dramatic increase in work from home, these frustrations are causing a significant impact to corporate performance.
Edgecell enables companies to deploy and manage VDI in the ROBO SOHO. The company still has complete control of the VDI, just as they would in the data center, but there is zero latency for the user because the VDI is on site. When combined with 5G, Edgecell as a ROBO SOHO edge computer dramatically changes how enterprises deploy and manage applications for their employees.
Edgecell versus traditional solutions
Verizon Wireless recently announced a MEC solution with Amazon Web Services (AWS). AWS Wavelength uses AWS Outpost, which is a full 19” data center rack of equipment. There are other vendors that offer data center rack equipment as a proposed edge solution.
The following compares Edgecell deployment to an AWS Outpost deployment.
| feature | Edgecell | AWS Outpost |
|---|---|---|
| starting cost per year | $7,500 (3 servers) | $240,000 (testing unit) |
| power | 0.09 kVA | 5 to 15 kVA |
| size | 9" x 7" x 8" | 80" x 24" x "48" |
| 0.3 cubic ft | 53.3 cubic ft | |
| additional equipment | small environmental case | building, generator, air conditioning, security |
| deploy to 10,000 towers | $75 million | $2,400 million |
| annual electric ($0.12/kwh) | $0.9 million | $52 million |
The AWS Outpose price does not include:
- Enterprise supportt
- Kubernetes
Most vendors that offer rack mounted hardware as edge solutions are not PaaS. Most do not include in their pricing the license and support for the OS and Kubernetes.
Start today
Edgecell is a low risk and inexpensive means for wireless network providers, partners and customers to experiment and learn the specific workloads and economics of MEC.
Edgecell is ready to deploy for MEC today. We strongly encourage members of the 5G community to begin their pilots now with Edgecell. We have the inventory, infrastructure and professional services available to engage with your team for immediate deployments for 20 to 100 locations.
Start today and establish your strategic lead in 5G multi-access edge computing with Edgecell.