Edge Computing- An insight and consideration on real deployment

In typical cases, IoT devices send several messages per seconds on cloud to get updated IoT device behaviour and to update its status on cloud. This poses multiple deployment challenges:

1)    Excessive WAN traffic, increasing network and processing requirements on cloud for data processing

2)    Since it involves WAN routing, RTT could be slow and can have delayed response

3)    Limiting solution design to handle when number of IoT devices increases

4)    Limitation with respect to bandwidth requirement of core network and backhaul

Edge Computing has advantages in terms of faster response time for the command/stimuli/request from IoT/field devices and reduced WAN traffic.

An Edge Computing Node is typically at proximity of IoT device on which computing has been offloaded. In such deployment case minimal additional computing is required on IoT devices. The IoT devices can be connected to Edge Computing Node using some access technologies.

Also, with this deployment option minimal SW logic change/additional framework required on IoT device to pre-process stimuli and reduce WAN traffic.

However, at the same time there are various challenges involved with Edge Computing.

The challenges with Edge Computing are as follows

1) Physical Deployment of Edge Computing nodes

2) Business Logic Decomposition & Solution Design

3) Communication Infrastructure & Setup

4) Communication Security

1)    Physical Deployment

IoT device functions has impact in terms of RTT with Edge Computing Node at Access Network. Given that various access technologies- WIFI, Bluetooth, Zigbee etc. has varying RTT, bandwidth, throughput and power consumption requirement of IoT devices supporting varying distances. Farther the distance of IoT device from Edge Computing Node, the weaker would be wireless signal affecting power consumption and data throughput.

This implies that it is better to have the Edge Computing Node close to IoT end devices.

At the same time deploying Edge Node with each IoT device is not feasible from deployment, cost and maintenance perspective.

Better option would be that a single Edge Computing Node should serve the multiple IoT devices. E.g.

·       An IoT Gateway can be extended as Edge Computing Node for multiple IoT devices

·       A home router can be extended as Edge computing Node for IoT devices in a home

·       A Dwelling Unit router can be extended an Edge Computing Node along with triple play- voice, video, data for set of devices in multiple homes

·       Geo-correlated or Neighbor Gateway

In order to support Edge Computing deployment from multiple vendors, such nodes can support multi-tenant infrastructure managed from WAN cloud.

Vendor can deploy lightweight Edge Computing software as on Cloud where platform shall be available as service (PasS). However, HW deployment on Edge cannot grow in same ratio as in the Cloud because Cloud Computing is geography independent, but Edge Computing is very much geography dependent.

2)    Business Logic Decomposition & Solution Design

Traditionally the Cloud computing software has been designed considering RAW data message from IoT devices. The messages involve both- the management messages (for device self-operation) and data/control messages (for device field deployment operation).

On Edge Computing, typical requirement is to pre-process the data messages. Typical pre-processing involves

      i.         Message suppression and filtering

     ii.         Messages aggregation

    iii.         Message transformation

    iv.         Message sanitization

     v.         Others?

Application specific business logic partitioning is required in such a way that

A)    At Cloud, below points should be considered

      i.         Most of cloud component and data pipeline require least changes

     ii.         Cloud scalability is considerably improved

    iii.         Cloud security is not compromised

B)    At Edge Node, below points should be considered

      i.         Should be able to filter/suppress message based upon requirement

     ii.         Should support access technologies used by IoT devices

    iii.         Should support messaging protocol used by IoT devices

    iv.         Should support messaging protocol used by Cloud solution

     v.         Should not alter the message authorization

    vi.         Should not alter the message identification and its source

  vii.         Should assume limited storage so that it will avoid keeping offline data

 viii.         Should not perform lengthy operations to become bottleneck

    ix.         Should support controlled and on-demand computing

     x.         May be equipped with load distribution with capability to distribute messages to next level upstream edge computing nodes

    xi.         May assume always ON broadband data connection towards cloud

C)    At IoT device, below points should be considered

      i.         Should be agnostic to upstream message handler- Edge or Cloud node

     ii.         Should have defined and constant field in message, in the beginning of message, to differentiate the message type- control or data message

    iii.         Should be more dependent upon transport protocol (e.g. TCP, SCTP) ack instead of application level ack

3)    Communication Infrastructure & Setup

Depending upon the domain and deployment (range, line of sight, bandwidth, point-to-point/mesh), IoT device and Edge Node can use one of the available access technology

- Wifi (short range and long range)

- Bluetooth

- ZigBee

- LPWAN

- Z-wave

IoT devices can be moving or stationary depending upon their domain and deployment, e.g. home automation IoT devices are stationary whereas wearables, automotive involve moving IoT devices. The device movement aspect creates the dependency on access technology. Also, the range of movement further complicate the dependency on access technology which varies from few foot to several kilometres (short range WIFI to long range WIFI to WiMAX).

Using cellular for communication at access layer can be benefitted with Traffic Offload otherwise it may not be much solving the Edge computing need as it will involve WAN routing which is as good as cloud computing, killing the benefit of edge computing by increasing the round-trip time.

In non-rural/developed geographical region, fixed broadband-based Edge node could be deployed providing support of application specific access technology. Also, deployment of additional HW should be easy.

However, to support edge computing in rural and geographically remote area, Industrial Router supporting LTE / 5G on WAN and multiple access technology can play a crucial role, providing good platform of Edge computing deployment and routing.

For the dynamic HW requirement on Edge node deployed in remote area, kind of distributed computing over connected nodes, in a kind of LAN environment, can be employed.

To support HW on demand at Edge node, cloud platform management can be deployed on Edge node.

In next set of articles we will discuss about

• Role of Distributed database in Edge Computing
• Role of Wifi Mesh in Edge Computing