Edge Computing: Simplifying Complexity
Our Story
Jeffrey Ricker, CEO, Ricker Lyman Robotic
I am an engineer. I solve problems. I do not make problems; I solve them. My purpose is to make things less complicated, not more complicated. To make things easier, not harder. I do not want to protract a problem. I want to solve it quickly because my joy is in solving problems. The world will never run out of problems to solve. The faster I solve one problem, the faster customers will offer me more to solve.
I am not alone. There are many, many engineers like me. They love to solve problems too. They too like to make things simple and beautiful and elegant and helpful. We are building a company where such men and women feel at home, alive and unleashed.
We are a solutions company. That means we build software and hardware and provide professional services. The challenges companies are trying to solve can only be solved with a combination of software, hardware and professional services. Software is a huge part of the solution, but it is not the complete solution. We don’t say to our customers, “Here’s the software. Good luck with the rest of your problem.” Nor do we say, “Here is a cool piece of hardware. Good luck with whatever you intend to run on it.”
By focusing on the solution, and not just a particular aspect of the solution, we are able to achieve breakthroughs in simplicity for the customer. That is what you are looking at when you see Hivecell. It is the first example of such a breakthrough. The first in a continually growing series of such needed breakthroughs that we can only begin to imagine. Because that’s what you create when you unleash amazing engineers: the as-yet unimagined.
From the cloud to the edge
And here is where we have focused to solve problems: edge computing.
Because this is where we can achieve the most impact for our customers.
Because this is where customers that are successful are focusing their energy next.
In the media you only hear “cloud, cloud, cloud”, a chorus of effort to centralize all computing power with four cloud giants. But the companies that really use information technology for strategic advantage, those that are implementing artificial intelligence, machine learning, the internet of things, they know that the pendulum is already swinging. They know that they must move compute power back from the cloud and out to the edge. They know that they need compute power for advanced analytics at the source of the data or as close to it as possible.
There are four major drivers for this shift: bandwidth, security, latency and cost.
Edge computing
Cloud is not going away. Absolutely not. But the edge is where the exponential growth will occur over the next few years.
The amount of data in the cloud now is massive, but it is dwarfed by the data that is being generated at the edge. For example, an oil rig has 30,000 sensors generating data, but less than one percent of that data is currently being used for decision making. With the current bandwidth, it would take 12 days to move just one day’s worth of data from an oil rig to the cloud.
Even with 5G and all the other ongoing network improvements, there simply will not be enough bandwidth to move all the edge data to the cloud. Even if you could, the cost is impractical. You know, cloud providers charge you far less to upload data than to download it. Once your data is in their servers, you pay to get it back.
Secondly, there is a latency in moving the data from the edge to the cloud for decision making. With some AI applications, such as smart cars, that latency is simply not acceptable.
So the demand, the huge demand, is there for edge computing. But the complexity that companies face at the edge is daunting. How are they to deploy, manage, maintain, upgrade and secure all that compute power scattered all over the earth?
Hivecell
Well, that’s what we’re solving: removing the complexity of the edge. That’s what Hivecell is all about.
Hivecell is a small, low energy, inexpensive, software friendly server designed for edge computing.
It is not hardware; it is a solution. It is the software that companies need at the edge that happens to come in a cool, amazingly easy to use casing. By providing these complicated distributed software frameworks such as Kubernetes, Kafka and Hadoop on our own hardware, we cut the Gordian knot of complexity.
Problems with existing Hardware and Software
Software has undergone a profound revolution in recent years. Starting with Hadoop, software is now designed to run on multiple servers. It is called “distributed software”. All the new frameworks you hear about, microservices, containers, Kubernetes, Kafka, Tensorflow, they are all distributed software.
Software has changed, but hardware has not. They are still building to outdated form factors from the 1980s and 1990s. The 1U form factor used by servers dates back to the 1920s! Servers are built for when
Existing hardware was not designed for the edge. It was designed for the data center.
It’s a mess. And the software is no better than the hardware.
For instance, Cloudera has more than a thousand parameters that you can set to configure their software. Most of the their professional services are simply installing and configuring their software. It’s ridiculous. Instead of spending money solving business problems, they are spending all the effort just to get the software running.
Hivecell is the solution
Do you know how many parameters we have to set for Hadoop on Hivecell? One. The parameter is: how many Hivecells do you have. That’s it.
Why? Because it’s running on our hardware. Our engineers have already optimized it in our own lab. Cloudera cannot do that. The Kubernetes vendors such as Redhat cannot do that. They say to their customers, “Here’s the software. We don’t know what hardware you have, so you figure it out. Or pay us more to solve it for you.”
That may be a product, but it is not a solution. Hivecell is a solution. You want Hadoop at the edge? Here, it’s ready to go. Plug it in and start using it. You want Kubernetes or Kafka at the edge? Here it is, ready to go.
And we do that for a fraction of the price. You will pay twice as much for the same compute power in cloud in one year. That’s just the hardware. Our software support prices are also very competitive, in part because the vendors such as Cloudera and Redhat price for dozens of big data center servers, not thousands of small edge servers.
We are low power. Hivecell runs at 5 to 10 watts with a CPU load. It runs a mere 20 watts with a full GPU load. That’s when you are running intense machine learning or AI.
We scale linearly. You need more compute power or redundancy, just stack another Hivecell on top. No chassis to buy, no backplane. You don’t even need more cables.
We are an edge focused solution. As such, we compliment the other big data software and hardware providers. We make their solutions better, because we provide a easier and more cost effective solution to gather the valuable data at the edge for them to process; whether it be analytical software or data centers filled with GPU processors…the better the input the better the output, and that is what we provide.
Video: https://youtu.be/KbPPrv7RZS0
