Skip to content

Technology

Analytics today is based on standard servers with some GPUs or TPUs thrown in for parallel processing that is required for neural network workloads. Cachengo has taken a different approach with purpose-built hardware and software that is designed to optimize neural nets and convolutional neural networks with massive parallelism, low power drain, high performance and tight security at a reasonable price.

In particular, the workflows that we are focused on are local Computer Vision based. By local, we mean that you don’t need a cloud or a data center to process your analytic workloads. You need a Cachengo.

The technology underpinnings for our approach rest on a management system that is designed for massive highly-parallel processing typical of analytic workloads with dedicated CPU, GPU/TPU, networking, and flash storage for each node, which we call a Symbiote. Cachengo management also provides the ability to easily introduce new applications, either from us, our customers or our partners through the Cachengo Market. This system give our customers the power to truly execute analytics at the source of the data with low latency for optimized edge AI.

We invite you to download our white paper to learn more about Cachengo technology.

Transportation

A lot goes on within a train or bus. Protecting and saving lives is just as important as getting passengers to their destination in a timely manner. For other computing companies, Mobile Edge Computing is merely academic. It’s okay, we love academics. But we’re actually doing what others have only dreamt.

Camera control, live stream vs video recording, sensor fusion, safety monitoring, gun detection… These are just a few of the things that Bento performs concurrently.

Don’t just settle for someone else’s imagination. Work with the winner of IDC’s Second Annual IDC Best in Future of Digital Innovation Awards.

Video Intelligence

Whether it’s real-time weapon detection, safety compliance or simply monitoring your livestock, Cachengo provides computer vision that is a complete video intelligence solution.

Other solutions require massive amounts of data to be streamed across the Internet to be analyzed. This is impractical on so many levels, chief of which is the response latency.

With Bento and Pizza, Cachengo offers massive on premise capabilities that allow your sensitive video data to remain on premise. Don’t trust your video to anyone’s cloud. Keep it!

Cachengo Portal – Cloud-based management tool for easy drag and drop management

Cachengo Market – Portfolio of applications

Cachengo Connect – Highly secure peer-to-peer Network SD-WAN perfect for Convolutional Neural Network jobs

bento-set-top-box-365
Bento - up to 8 Symbiotes
Pizza - 32 Symbiotes

Symbiote - Where the Drive is the Computer

Balanced compute and I/O
Deploy everywhere
AI/ML where data sits
2.5" Form-Factor
OPEX reduced 5x
CAPEX reduced 4x
CAPEX reduced 29,000:1
Symbiote includes CPU, GPU/TPU, Flash, and networking in a 2.5" form factor

Symbiote was inspired by the run-of-the-mill, industry-standard 2.5″ form-factor hard drive (HDD). We thought the packaging and size was great, if only we could marry it with something new. This epiphany led us to invent Symbiote – an ultra smart drive ideally suited for the edge!

What could be more symbiotic than infusing incredible computing power directly onto a drive? Others have tried to do similar things but have simply missed the mark by combining such capabilities with the drive’s internally accessible micro-controller. The old way of doing things relied on the system’s main CPU to offload pre-programmed functions to these additional resources.

This created a bottleneck for the already over-tasked CPU that was then forced to carry the additional burden of acting as a centralized arbiter of these computing resources. These earlier forms of smarter drives were also highly proprietary and did not embrace any notion of choice. Fast forward to 2020 and the latest crop of SSD companies are still doing the same thing.

Symbiote is something completely new and different. Each drive becomes a stand-alone computational platform and server. There is no notion of a centralized CPU to become a bottleneck, or to act as an arbiter. Each drive is therefore individually addressable over the network, and can act as an object storage device (OSD).

This stands in stark contrast with conventional network storage architecture, which is more than two decades old and still treats the entire multi-drive system as the OSD. This era of new technology requires a substantially new architecture to take advantage of it.

Unlike its predecessors, symbiote allows you to control how much capacity, whether you want to mix and match flash with hard drive technology, and even what type of software you wish to run. Choices! What a novel concept!

The drive is your new analytics engine

Today’s complex analytics require data– and lots of it. If we are talking about machine learning (ML), artificial intelligence (AI), or just some type of augmented reality (AR), the need for fast storage is substantial. Latency and throughput become much more critical in order to deliver the data to where the analytics are processed.

The whole genre of ‘Computational Storage’ was created to address the needs of these new workloads. But analytics processing (also referred to as neural networks) is very different from offload. While other solutions incorporate offload capabilities, they cannot implement complex neural networks. This is because they do not possess graphics processing units (GPUs).

Our Symbiotes feature embedded GPUs and can perform sophisticated neural networking functions locally, right where the data is stored. What this means is you can ingest the data where it will permanently reside at the lowest CAPEX and OPEX possible, perform your analytics, and then merely send the results to wherever your central indexing resides.

This approach is not only cheaper, but is faster, less complex, and also simply smarter. As we continue to usher in 5G and Edge capabilities, we have a responsibility to be ecologically conscious. Processing data at the drive significantly reduces this network backhaul!

Networking - A new type of SD-WAN

Most SD-WAN technologies simply focus on connecting or extending networks. At Cachengo, we needed to do much more than that. We needed to provide a way to connect to potentially millions of Symbiotes, which are almost exclusively deployed behind secure firewalls and can be configured as object storage devices (OSDs), or even as application servers.

We wanted to handle all of our communications without a dependency upon Secure Shell (SSH). It is with this idea in mind that we created Cachengo Connect. Cachengo Connect is a messenger and a protocol at the same time, making it a unique product that allows us to seamlessly connect to our devices, no matter where they sit.

We can use Cachengo Connect in combination with Cachengo Portal to quickly deploy applications. Anything we can do via SSH can be done via Cachengo Connect, but in a fully-trackable, secure, and auditable fashion.

The worst thing you can do for your edge computing latency is to connect everything up to a proxy. We efficiently and securely connect all of our managed devices, regardless of whether they sit behind different firewalls. Proxies are inherently bad for scaling for numerous reasons— the biggest is that they ultimately create bottlenecks.

Imagine wanting to go from point A to point B, but having to go through point Z to do so, only to find that there is major construction going on at point Z. Why go through this experience if you don’t have to? Connecting devices up via VPN tunnels is equivalent to throwing a proxy in the middle of all of your traffic. You can do it, but you don’t have to.

In addition to the bottleneck potential, proxies are inherently bad because they also provide a risk for man in the middle attacks.

With a man in the middle attack on a proxy, any vulnerability can be exploited to create an opportunity for data to be hijacked, manipulated, or otherwise compromised. Furthermore, any connected devices or services can also be attacked through these types of exploits.

Besides the proxy concerns, there is also a concern for exposing endpoints to the public. Ever hear of DoS or DDoS? DoS stands for denial of service. When you expose a service on the internet it becomes extremely vulnerable to attacks. This is because the service must broadcast itself to whatever is trying to find it and utilize it. People with malicious intentions can discover these exposed endpoints and disrupt such services by overloading them with requests until the underlying resources crash.

With our Secure Routes we can connect many endpoints without the use of VPN tunnels and without exposing resource endpoints. Imagine being in a building with many offices. Now, remove all doors, windows, and even hallways, but you want to go from one office to another. While such a scenario might wreak havoc to someone suffering from claustrophobia, it is actually ideal when it comes to network security. This is how we connect devices efficiently and securely. Of course, you are free to continue to utilize your VPNs if you like... but we wouldn’t recommend it.

Management

Management outside the data center might seem difficult, but the environment demands an easy-to-deploy and manage solution. Cachengo Portal is that management solution. This cloud-based Tool enables easy provisioning of the Symbiotes and addition of new applications from the Cachengo Market which could include object storage from MinIO or containers with Kubernetes.

Object Classification Demo

Peer Groups vs. Open Firewall Ports

Our one-touch approach means that large scale deployments are simple and easy- even though the Symbiotes Operate as independent elements, perfect for highly parallel workloads like analytics and Computer Vision.

Schedule a Demo

Please fill out the form below
Copyright © 2024 · Cachengo, Inc.