You know how much we love ARM and Raspberry Pi. The first because it is a great datacenter technology that enables us to deliver on our promise of efficiency and flexibility while giving our customers more freedom of choice. As for the second, it is essential to ease the adoption of our technology, build PoCs quickly on a low budget… and it is fun to play with at home.
Today we are taking it a step further with the launch of the Raspberry Pi Program.
Why the Raspberry Pi Program?
OpenIO SDS is open source software that can be installed on commodity servers to create an object storage infrastructure capable of scaling from 1TB to 1000s of PBs. Its innovative and lightweight design, coupled with Conscience technology (for dynamic load balancing), are at the base of its flexibility and ease of use.
SDS's flexibility and scalability allowed us to build products like the Nano-node. An ARM server (1 2-core CPU, 1GB of RAM, 2 2.5Gb/s Ethernet ports, 32GB SSD) packaged in a very small form factor and connected to a single, large-capacity storage device (HDD or SSD). It also enabled us to partner with companies like Ambedded Technology.
This type of solution has several advantages when compared to other object storage architectures: it offers the smallest failure domain, lower power consumption, unmatched flexibility in system management, and improved parallelism in data access.
OpenIO is continuing to develop SDS on ARM for several reasons. The most important is to do R&D and test next-generation solutions while showcasing the benefits of our innovative architecture. At the same time, since our software is open source, we want to make it available to the widest audiences and grow the community around SDS. This is also why we support Raspberry Pi and make it part of our strategy.
How does the Raspberry Pi program work?
SDS on ARM is not only a product we support in production for our customers, but also a community project that leverages the Raspberry PI hardware. The goal is to grow the OpenIO community by doing something interesting, fun, and rewarding.
The first thing to do is to join our community on Slack, where you can find information and interact with other members. Then you can start learning about OpenIO SDS and Grid for Apps to discover the possibilities they offer. If you come up with an idea about building a project let us know; we are here to help.
Anything goes: a how-to explaining an integration with a third-party solution, a plug-in to integrate SDS with other systems (like this for example), documentation, your home IoT project, projects based on Grid for Apps, and more. And if you don't have a Raspberry Pi yet, you can work on a Docker container or with a VM. Publish the content on the internet and let us know on Slack or via a DM on twitter. If it has value for the community we will help you in promoting it and reward your work with a Raspberry Pi kit. The box will include a complete Raspberry Pi in a special case with our logo and all the accessories you will need to install OpenIO SDS on it, and a bit more.
ARM CPUs are becoming increasingly relevant in the datacenter. The characteristics that have led to this success in mobile computing (power and space efficiency) are now at the base of its success in the modern datacenter. ARM is not the only alternative to x86; specialized coprocessors like GPUs and FPGAs are becoming more relevant too for the same reason: the search for better overall efficiency.
Hyper-scalers and large enterprises are looking at alternative server architectures because they want to squeeze more resources per square meter, but, at the same time, they need to consider power consumption and infrastructure sustainability. ARM makes it possible to have many more cores per rack unit while using less power, and this is perfect for the high number of parallelized tasks performed by applications that scale horizontally, such as most web apps and modern storage architectures. To achieve this result, hardware is not enough; software must be designed to scale quickly and seamlessly with embedded resiliency and availability mechanisms to sustain any sort of failure and maintain the system manageable at any scale.
OpenIO SDS and Grid for Apps, through their lightweight and innovative design, are optimized to run on x86 and ARM CPUs, providing freedom of choice to our customers and enabling them to choose the hardware platform that best meets their needs.
IoT, Grid for Apps and ARM
Our vision is to provide not only object storage but also an integrated computing framework to help process data directly on the storage infrastructure. By doing so, it is possible to offload specific tasks to the storage system and improve overall efficiency without needing to build complex infrastructures. This approach, usually referred to as serverless computing, is very interesting in the data center, but is even more appealing in IoT environments.
OpenIO SDS and Grid for Apps provide the same level of functionality at the core and at the edge, making it possible to build a seamless, end-to-end data plane. Simple applications, triggered by events, can perform local data processing without needing any additional components. For example, data filtering, integrity checks, or even image recognition or manipulation can be performed with a goal of enriching data and metadata before replicating it to a central repository. Thanks to this type of functionality, it is possible to choose the right pieces of information to transmit, saving bandwidth by sending only relevant data while making sensors smarter and more secure.
The Grid for Apps core framework is open source software and comes with SDS. It's available on the ARM platform and it is perfect for building enhanced IoT infrastructures that can respond to the most demanding requirements in terms of efficiency, resiliency, and security.
OpenIO SDS on ARM, with Ambedded
Beyond Raspberry Pi, OpenIO SDS on ARM is also a reality in production environments. We started two years ago with OpenIO SLS4U-96, a specific hardware appliance used for very dense multi-petabyte environments, but we have also received a lot of inquiries for smaller installations. We think we now have the right solution to start much smaller (starting at 50TB) and grow from there, seamlessly and granularly, when needed. A solution that is perfectly tailored for local ISPs and enterprises that need to start with object storage just for a single application, or with an initial limited investment.
Ambedded Technology and OpenIO partnered to bring the most affordable and efficient object store solution to the market. Ambedded Mars 200 is an innovative product providing a cluster-in-a-box approach. Each Mars 200 is one rack unit tall and has 8 ARM-based micro servers connected together on redundant 2.5Gb/s switches with 10Gb/s uplinks, perfect for running OpenIO SDS. The customer can start with a small capacity (48TB usable) and grow from there, linearly scaling with the number of nodes, capacity, and performance depending on their needs.
The solutions is available through Ambedded and OpenIO resellers all around the globe starting at 0.008 $/GB/month including a 3-years standard support subscription.
Freedom of choice is a key aspect when it comes to long-term strategies in IT. Supporting ARM alongside x86 gives more options to end users, and, at the same time, OpenIO’s investment in the open source community protects the customer against any form of lock-in.
SDS on RPI is an opportunity to look at OpenIO’s open source technology, play with it, and test its ability to integrate with third-party solutions.