At the Hewlett Packard Enterprise (HPE) annual Discover conference this week, the company announced a new product category called “Converged IoT Systems”. HPE described the category as “a fundamental shift in compute infrastructure”. The announcement was made in conjunction with the introduction of the new Edgeline IoT systems—basically new servers that integrate data acquisition, real-time analytics and control with the typical server computing, storage and networking.
Hewlett Packard Enterprise logo (Source: Hewlett Packard Enterprise)
Did HPE discover a new category? In my mind, HPE is heading exactly where the industry is heading: moving compute power and analytics closer to the source of the data. There are three main reasons for moving computing closer to the end devices.
- “Timeframe of Relevance”: Timeframe of relevance is my way of bundling together industry terms such as “latency”, “time to insight” and “time to action”. IoT is all about actions on data, but every application doesn’t need to be acted upon in the same way, or in the same time period. Some applications need real-time action, and others may only need daily or weekly actions. Thus the term “timeframe of relevance” is key to actual deployment of successful IoT solutions.
- Bandwidth: If you believe the forecasts, there are going to be billions of IoT end-devices by 2020 (and sooner). If all those devices are sending data on a regular basis, the stress on the current network infrastructure is going to be untenable. By processing data closer to endpoints, the amount of data sent upstream can be greatly reduced, minimizing the load on network infrastructure.
- Cost: Sending data costs money.
So given that there are real advantages to processing data closer to the actual device, let’s take a look at where data analytics could possibly occur.
- Devices: There are many applications where analytics can be done on the actual end-device. Many IoT devices have on-board compute power (think an ARM Holdings-based or Intel processor) to handle analytics, allowing the data to be acted upon immediately. This immediate action will not only provide real-time control, but it also reduces the data traffic on the network upstream. A simple example is monitoring a camera. If you think about a video camera set up to count cars at the intersection of two roads, you don’t really need to watch video of the cars going by, you only want the count and timing of cars. In addition, sending video data upstream will suck up a lot of network bandwidth. The smart and more efficient choice is to process the data on the device and just send the car count upstream.
- On premise: As we have seen already, there are a number of vendors that are looking to do analytics at the “edge”. This can take a couple different forms. First, gateways can be made smarter (i.e., add “server like” functions to a gateway). We’ve seen this with IoT gateways out of Dell, Cisco Systems and HPE where storage and compute power, along with analytics software have been added to an industrial gateway. The second is to leave the gateway alone and move a server on premise—such as HPE just announced with the new Edgeline EL Converged IoT Systems. These systems combine datacenter-class server functions including security, device and system management, along with compute power with data capture and control to allow analytics to be run closer to the actual device. A typical application for this type of local control is asset management, where you need the location of everything you have, but often not in immediate or real-time. In addition, in this type of application the data will often need to be sent up-stream to be integrated with ordering and inventory systems, etc.
- The cloud: I think this one is pretty self-explanatory, and more traditional. Collect all the data at the device and move it to the cloud for processing. This is great for applications where real-time is not critical and the data needs to be integrated with other systems and sources of data.
Depending on your data needs, and your existing infrastructure for both networking and data, you can potentially use any or all of the above—and large enterprises will.
So has HPE developed a new category? The answer is really beside the point. The industry is moving towards an Internet of Things architecture that allows for data analytics in a timeframe of relevance. HPE’s Converged IoT Systems, no matter what you call them, provide enterprises the means to deliver on analytics where and when they need it. This puts HPE squarely in line with where I believe the IoT architecture is headed.