DEV Community: Sesame by ITRenew Engineering

Circular Data Center, Open Hardware and Open Compute Project

Sriram Ramkrishna — Wed, 16 Dec 2020 20:44:10 +0000

I hope everyone enjoyed the last post about open system firmware and the role that the circular data center plays in it.
Previous post: Circular data center - Working with Ecosystems.

One of the most important umbrella projects that the circular data center works under, and one of main enablers of the circular data center, is the Open Compute Project (OCP).

Open Compute Project History

OCP was initially started by Facebook as a way to share designs of modular data center components with the rest of the industry back in 2011.

The data center industry is dominated by a number of industrial giants each with their own proprietary set of technologies, built on a set of common component building blocks. Often, the same set of technology is implemented with a specific set of value adds or design constraints. The open compute project works by convincing the industry giants, and many of the key component providers, to work together by collaborating on common technologies without the overhead of maintaining many discrete designs or repetitive bilateral relationships. By working together we can reduce costs and duplicated efforts, while also building better technology.

OCP is separated into a number of projects. I'll focus on a couple important ones that we care about, because as a vendor that sells hyperscale technology they are integral to how our products are built.

Server

The server project is one of the more influential projects for the circular data center. This project provides the physical design for the racks and chassis documenting the specifications for electrical, mechanical, and debug/test framework.

All our products are based on the open rack compatible chassis and sleds. There is a difference between the open rack vs the standard rack sizes in which rack sizes are a little taller than standard sizes - so an important ecosystem to keep track of. But the advent of this project allows the circular data center to depend on the community of companies whose expertise is building rack hardware.

That doesn't mean we are sitting idle - our contributions in this space have been around re-envisioning OCP hardware in different form factors. The Sesame by ITRenew Fast-Start system for instance is an excellent example of rack scale form factor re-imagined into a desk side form factor, and we continue with form factors for edge. In this way we take advantage of the work done by the community and create new products and new uses for re-certified hardware.

Hardware Management

This is a relatively new group - which is why it is in incubation - but it focuses on hardware manageability. Examples would be a rack manager, the BMC (baseboard management controller), telemetry and so on.

This group is an important component of data center management and is an obvious choice for circular data center companies to be involved in.

Networking

There are a few interesting hardware and software networking open compute projects.

In the software space, some companies are collaborating with each other to create community supported networking operating systems. For now, the project to watch out for is SONiC/SAI. SONiC is a project incubated by Microsoft for Azure. This is a complete network operating system that is based on Linux that runs on switch hardware from many participating vendors and is already production worthy. For the circular data center, this means with the switches we get, we can deploy SONiC to get features like dynamic breakout on recertified hardware. Projects like SONiC allow us to not worry about end-of-life software on re-certified equipment. That shows the power of community!

SAI is the interface API that SONiC implements. This opens the possibilities of writing interesting software that interacts with the switch software. With the freedom to build, the SONiC community will have innovative solutions in the near future that I believe will outpace proprietary solutions.

Rack and Power

Another important project under the Open Compute Project is the Rack and Power group. This group focuses on rack standards, they analyze power distribution, the heating and cooling, air flow and other things that makes a rack an effective and scalable solution to house hyperscale servers. We rely on this team to provide the incremental equipment to make the best solution for a data center that reduces the impact of the by products of data centers like heat, energy consumption and what not.

Conclusion

All our products at ITRenew are based on open rack compatible chassis and sleds. Mechanical and electrical compatibility across generations of OCP hardware designs allow us to build rack solutions with a mixture of node types, including combinations of storage and compute elements. These shared specifications and details are also critical information for us as we work on design extensions that allow us to expand server use cases and re-use cases. Without open availability of detailed design files and specifications, the ideal of a circular data center would be much more difficult to achieve.

Furthermore, the software like the network operating system SONiC, allows us to run them on re-certified switches allowing them a new life in the data center.

Hyperscale size service providers often retire their data center IT equipment after 3 or 4 years of service. The application demands and business growth typically drive the need for the most advanced hardware. This hardware is very useful for many businesses and applications and repurposing the hardware earns non-trivial contribution towards the reduction of greenhouse gases (GHG). Today, a large portion of the total GHG emissions come during the pre-use phase, or manufacturing phase when raw materials are mined from the earth and converted into IT equipment. In fact, when IT equipment is installed into a data center that only uses renewable energy, all the GHG released into the atmosphere is during the pre-use phase.

How does all this relate to the Open Compute Project? OCP has embraced the theory of a circular economy (CE) where IT equipment is repurposed rather than recycled. Repurposing may occur within a company or may go through refurbishment and deployed by a 2nd or even 3rd tier business. To facilitate the repurposing, many of the OCP projects are taking steps to make it easier to repurpose the IT equipment and for the new owner to achieve ongoing business continuity (firmware refreshes, security features, etc.) OCP offers a marketplace to help businesses find efficient and scalable CE equipment. Companies like ITRenew is also a critical link in this supply chain.

So to conclude this blog post the OCP Project and its members recognize the importance of the circular economy and its far reaching impact both for its positive impact on carbon footprint but also as an practical demonstration of the benefits of OCP as touched on in previous blog posts. In turn, the circular data center spends a lot of time providing OCP solutions into markets that would not normally see this, thanks to the myriad of solutions that target second and third tier markets.

That's all for this month. Next month - I will write about the openBMC project,and maybe some of the projects we are working on that y'alls might find interesting.

This blog post by reviewed by the kind folks at Open Compute Project staff, and ITRenew comrades.

Circular Datacenter - Working with Ecosystems

Sriram Ramkrishna — Wed, 21 Oct 2020 23:24:42 +0000

This is part two of a series of blog posts that talks about the circular datacenter as an intro primer of what we do. Part one is here where we left with a cliffhanger on how openness is important to the circular data center.

We talked about how sustainability and reducing our carbon footprint are important for a healthy planet. As humans we have a responsibility to take care of the planet, our home amongst the cosmos.

Now let's talk about business, because that's important as well. Making the sustainable choice is great, but we can’t sacrifice the stability of the data center. Even though the cloud environment provides you with a seemingly fault resistant environment - you're still vulnerable.

We'll get back to that in a bit.

Let's talk about firmware. Firmware is very low-level software that runs on hardware that is used to initialize and load the OS bootloader and provide a priori knowledge about the hardware configuration.

Hardware manufacturers maintain firmware in order to fix software bugs, work around hardware problems, and fix security exploits. All of these keep your hardware useful, reliable and dependable; As long as your hardware is relatively new.

Traditionally, silicon vendors tend to keep firmware as a secret, because it provides competitors, ostensibly, with the knowledge of the internals of the hardware.

Original ODMs and OEMs don't have a choice as their license doesn't generally allow them rights to redistribute the source provided by the Independent BIOS vendor (IBV). This also means that the hardware vendor must support the firmware for the life of the product in most cases. A microprocessor could have a product support of 20 years, but smaller hardware manufacturers could provide substantially fewer years of product support.

Once a new generation of hardware comes out - support for the old firmware starts deprecating.

In fact, over time firmware tends to move to "sustained long term support" where only critical or security bugs are fixed. A skeleton crew is staffed to support it as the hardware company starts pushing its clients to purchase the new hardware and upgrade.

If your hardware is coming from sustainable sources and it's already four to five years old - how well supported is the firmware for that hardware? How long before support is sunset, leaving these machines without support?

To answer the supportability question, we need to look at a project called Linuxboot. In short, if we replace firmware with the Linux kernel, we can move to a more sustainable long-term support model. Given the previous attitude by hardware manufacturers - why would they move to this?

There is nothing specifically magical about firmware - companies end up re-implementing the same code paths all the time. By open sourcing the firmware - the community can handle the maintenance of the firmware and continue it indefinitely. Leaving companies to only implement the hardware feature specific portions - reducing engineering time and time to market.

But more interestingly, that maintenance cost is spread across interested parties. So, companies only spend a fraction of their manpower to support older hardware - plus critical bug and security fixes will be quickly addressed. So overall it's a win-win from a practical point of view.

Plus, we leverage the resources of the large community that surrounds the Linux kernel.

This is great news for the circular data center since that means we can use Linuxboot to continue to support older hardware far longer than a singular company would be interested in doing and without paying the big bucks to that hardware company it would require to keep that product supported.

However, there is no passivity on the part of companies in the circular data center business. They play an active part in this ecosystem because they may want specific products supported - that means they hire firmware developers and contribute code, tests, and other resources upstream as part of the ecosystem. Given that it is all open source, contributions are all merged upstream and community supported - thus we socialize the costs of support (separate from the cost of initial engineering) - thanks to the strong community around Linuxboot.

While the bulk of this post has been around firmware - that isn't the only ecosystem that the circular data center invests in.

In part three, we'll talk about other ecosystems specifically projects that are part of Open Compute Project.

Acknowledgements:

Photo by Markus Spiske on Unsplash

Big thank you to the Open System Firmware community for reviewing this post.

Let's talk about the Circular Data Center!

Sriram Ramkrishna — Tue, 29 Sep 2020 21:23:44 +0000

This is the first post of a series of blog posts from the engineering team from Sesame by ITRenew Inc. Open Source, the Linux kernel, userspace and the open data center, is an important part of what we do at Sesame Engineering. Not only is the ethos of “openness” important, it is critical to the success of the circular data center. Let’s talk a bit about what the circular data center is for context and then talk about how openness both in software and hardware is so important today. There isn't a tag for hardware on dev.to so we'll have to make one! I've used 'ocp' for Open Compute Project which drives open hardware in the datacenter.

Circular Data Center

Never heard of the circular data center or maybe the circular economy? You’ve come to the right blog post! To truly understand, let’s talk about supply chains.

The life cycle of any component of a data center, be it a network gear, or racks of compute and storage nodes - starts of course from their manufacturers. Data center hardware products are manufactured from a supply chain of silicon, copper, gold, steel and rare earth metals and then are fab’d ostensibly in a foundry. The resultant product is then sold to companies and organizations.

Now if you’re a first tier company like Google, Facebook, or others the moment these data center components land in your facility - the clock starts ticking and the value of the of the now company asset starts depreciating. Over a period of four or more years, the residual value of the asset is such that it is time to buy the next generation of data center assets.

What happens to the assets after they are replaced? Some get re-sold, or donated or find themselves in a dumpster in the U.S. or in a foreign country.

Seems like a waste doesn’t it? These components are top computing equipment four or five years ago. First tier companies even have a large volume that could end up as e-waste.

Some companies though - they contract companies like ITRenew - who takes these products and decompose them into its essentials - CPU, memory, fans, and motherboards if they are compute and storage nodes. The drives are also taken and securely wiped. ITRenew has one of the most mature methodologies to do this. The racks are recycled. Some products are taken as a whole like routers and other network gear.

Still with me?

In the aggregate, we’ve now created a supply chain of components that we can now re-use to build hyperscale servers at only a fraction of the original cost of what they were four years ago. In fact, a company like ITRenew Inc can process over 18,000 or more components a month!

Now, these servers that might have ended up in a dumpster; find a new life by being re-certified for other markets. While they may no longer be attractive to first tier markets, they would be a boon to second tier markets, educational institutions, and government.

You might say - what’s the big deal? Glad you asked! The dirty secret is that the only difference these days between say microprocessors from 4 years ago and now, is the number of cores. Processors are not relatively getting faster instead; they are doubling.

First tier companies though, find the number of cores doubling to be a very attractive for them because they get more computing per floor tile - meaning they get greater density in a data center allowing them to use the space more efficiently and save on buying and expanding existing data centers. This can be prohibitively expensive - in the billions of dollars to maintain because of costs of power and HVAC.

However for second tier and third tier markets - the computing power is more than enough for most needs and do not require the large growth in computing needs that first tier companies like Google require. In fact, second and third tier growth will increase much more slowly - allowing for sustainable horizontal scaling.

So not only does circular data center make fiscal sense, it leads to a sustainable use of resources while lowering the carbon foot print. With re-use, we can defer manufacturing for longer and not manufacture more. If you purchased new equipment you’d have already increased your carbon footprint without even turning on the equipment. An analysis by ITRenew - has shown that re-use can slash carbon footprint by 25%. That’s not peanuts.

So, let’s end it here. Let that percolate a bit and if you have questions on what I’ve been talking about - happy to answer. My next blog post will be about openness and it’s importance in the circular data center.

NOTE: This is a new platform for us so if I have inadvertently violated any community norms through use of tags - please let me know. Looking forward to hear your feedback.