According to IDC, the world will create 163 zettabytes of information by 2025. The world is moving fast and demands more capacity. The storage industry is moving to NVMe SSD storage, but there’s only so much silicon in the world to fulfill these orders.
I had the opportunity to attend Storage Field Day 17, hosted by Tech Field Day. The last presentation was with NGD Systems and how they want to change how the industry treats large data by introducing computational storage with their In-Situ ecosystem.
Increase Watts / Terabyte
Capacity will need to increase in our SSDs and NGD Systems has figured out how by reducing the watts per terabyte while increasing capacity. This is very tempting from cloud providers who are looking to reduce environmental cooling costs while increasing their storage and not increasing the footprint of the data center.
NGD Systems is improving on hardware and capacity, but they are also developing a platform around their product and in doing so is creating some very unique solutions. One of the problems cloud providers face is computing large data sets. It’s very expensive to have a host CPU dedicated to moving data from the storage unit, process and spit out an outcome.
Relate this to a company network share that has terabytes of files in it and you’re trying to perform a search. It can take a while to load in the data, let alone perform a search.
To help ease the host CPU, NGD mounted an ARM processor to the SSD. This mounted ARM processor is a quad core 64-bit processor. Bringing the compute to the storage. I like to think a computer…in a computer, or a physical micro-service. It makes perfect sense! Why not offload the searching or computation of large datasets to a smaller processor. A single ARM processor isn’t going to do much with a large dataset, so NGD Systems has developed a way to parallelize these same devices so you can throw a potential 256 ARM processors at a dataset from a single host.
Now you can offload your network share search to the ARM processor on the drives and perform the search from the storage itself. Instead of moving chunks of data to the host CPU, scanning it, and repeating.
If that wasn’t revolutionary enough, NGD Systems offers a licensed software package they call In-Situ. In-Situ is a suite of Open Source products to provide an entire ecosystem to interface with these computational storage devices. NGD Systems realized if they want adoption, they need to provide tools developers are familiar with.
The In-Situ suite provides the ability to now make the drives discoverable on the network. A great example is a distributed file storage spanned across a multitude of these computation storage devices from a number of servers. This now allows not just the data, but now services to be offloaded from the main CPU.
One of the tools used by NGD Systems is utilizing Docker Containers. Docker provides a repeatable environment that’s portable and consistent across multiple platforms. NGD Systems provides the foundation for developers to start building containers that can be deployed to the computational storage devices. This provides fast adoption from the developer community space.
One of the other uses cases for NGD Systems is using their technology in IoT devices or Edge computing. Both device types have reduced compute and storage in them today. Add in an NGD System with the In-Situ suite, you’ve just added additional compute and the ability to port your containerize application to an entirely different platform without re-writing the app.
I enjoyed listening to NGD Systems and their approach to today’s capacity problem. Other storage vendors talk about speed and getting to the data as quickly as possible. Sometimes speed isn’t always the answer. NGD Systems has found a different approach to today’s capacity problem.
Introducing computational storage to the industry with a great software ecosystem has unlocked a lot of great potentials. Cloud providers can redirect their efforts and re-think their compute engineering efforts. Instead of each server having dual sockets, get away with one and reduce software licensing costs as well as reduce costs to cool and power dual socket systems.
The In-Situ ecosystem provides the ability to move containers to the edge. Maybe now our edge devices got even smarter and we can tap into the devices unused CPU usage and add it to the overall compute infrastructure. There are so many devices that have CPUs in them, routers, switches, modems, phones, IoT devices, cars etc that if we offload their daily CPU usage and it becomes idle, those are additional devices to draw cycles from.
NGD Systems has opened up a lot of doors as well create new ones that we can see yet. It will be fun to watch them and their future.
Delegate follow up with @NerdBlurt @deepstoragenet and @jhoughes