The Hardware Alternative To Softiron’s Public Cloud

Not everyone is happy with what hyperscalers are offering. But there are alternatives. One example is the provider Softiron, which implements an integrated, multi-client capable, and environmentally friendly approach with its “Hypercloud.”

Many security-conscious companies, managed service providers, or medium-sized public cloud providers, do not benefit from the infrastructural advantages of hyperscalers. They need other infrastructures to implement their business models optimally. It is not yet clear which hardware will prevail in the emerging edge market. However, edge locations still need a lot of flexibility.

Above all, they should be easy to maintain and scale since they are often located in inaccessible areas. An example of a hyperscale independent cloud infrastructure is “HyperCloud” by Soft iron. The manufacturer chooses a very unusual way of designing and manufacturing its hardware for its “intelligent cloud fabric.” It runs its own minimalistic and hardened operating system.

Simple And Multi Tenant

The head office designs adapt to the user and are manufactured close to the customer to save delivery costs and anchor themselves in the local economy. In addition to being close to the customer, the aim is also optimal recyclability of the components. The supply chain should be as short and transparent as possible.

Simple management, resilience (up to nine nines, 99.9999999 percent availability), simple and theoretically limitless scalability, and multi-client capability are other features claimed by Soft iron. The multi-client capability is expressed, for example, in the supplied customer-specific usage measurement of the individual resources with an interface to billing and customer- or group-specific access rules. It can be defined for each customer how much of which aid he is allowed to use.

ALSO READ: Digital Transformation Is An Ongoing Journey

Turn Key Installation

Soft iron Hypercloud is installed and ready for use by customers. Virtual machines, LXC containers, and multi-VM services can be provided via a virtualization layer. An infrastructure can consist of several cooperating on-premises zones and private and public clouds.

A single image provides stateless servers, storage, and networking on bare metal. For stateful applications, there are guest volumes that hold the data. Guest instances can be imported, uploaded, or created manually, but they can also be downloaded from a marketplace like an app marketplace. Other options include Docker Hubs, Linux containers, and operator-specific marketplaces. Working VMs can be saved via snapshot and integrated into the market.

Compute Node

Compute nodes have eight or 16 cores, 64 and 512 or 1024 (16 cores) GiB memory, and a maximum of two 10/25 Gbit/s interfaces. The power supply is available twice on request, and the power consumption is 110 or 165 watts. Heat dissipation is 375 and 563 BTU/hr, respectively. BTU (British Thermal Unit) is a caloric unit of measurement.

1 BTU is the heat needed to raise one liter of water by one degree. The value, therefore, indicates how much cooling is required as a result of the implementation of the system.

Storage

Storage nodes are available in four variants: Value (48 TB), Density (48 to 144 TB), Performance (53.78 TB) and Performance + (2.2 to 51.2 TB). The respective power consumption varies between 100 (value) and 230 watts (density), and the heat output is between 341 (value) and 784 BTU/hr. Data is placed directly from the compute nodes onto the media using a deterministic hash algorithm. There are features for journaling, caching, snapshots, and cloning.

Networking

The individual components fully mesh where possible, otherwise connected to hierarchical leaf-spine networks at 10, 25, or 100 Gbit/s. The control plane automatically configures the network. There are also cloud and hyperx-cloud management interconnects. The cloud interconnects module has 48 ports with 10 or 25 Gbit/s, plus four QSFP28 uplink ports with 100 GbE.

The power consumption is a maximum of 550 watts; the heat output is 1876 BTU. The management interconnects module manages 48 connections at 1 Gbit/s each. It has a 1Gbps uplink port, consumes 150 watts, and generates 512 BTU/hr of heat.

Widely Usable Reference Designs

The manufacturer specifies four reference designs: two minimal configurations with or without resilience and two other resilient arrangements with particular tasks – one geared towards performance and one geared towards large amounts of data.

The non-resilient minimum configuration has single points of failure in the compute and network area. The storage area is also redundant here so that no data is lost in the event of a loss. Areas of application are simple edge applications, crypto mining, development, and testing. 

The configuration includes a cloud and management interconnects module, a compute node with 64 GB RAM for eight VMs, and 48 ​​TB double redundant storage (three modules), which consumes six rack units. The minimal resilient configuration has at least two cloud interconnects, one management module, two computers, and three storage nodes. According to the manufacturer, it is the configuration for beginners who want to withdraw some or all of their workloads from the public cloud. The connection to the fabric is redundant for the compute nodes.

Data And Performance-Intensive Configurations

The resilient configuration for data-intensive purposes should feel like consuming cloud storage for files, blocks, and object storage. In an entry-level design, in addition to two redundantly connected compute nodes, you need two clouds, one management interconnect, and 42 storage nodes, each with 144 TB.

The performance-optimized version comes in the basic configuration with three particularly powerful NVMe storage nodes, each with 51.2 TB of RAM. They are combined with more enormous compute nodes with 1024 GiB RAM.

Carbon Dioxide Calculator

The “Carbon Calculator” on the Softiron website is interesting for environmentally conscious users. But it only works if all cookies are loaded. The calculator is available in a basic and an extended version, which can be accessed via the “Edit All Working Data” link. The basis is the amount of storage required and a country selection.

In the extended version, data such as the electricity price or the media usage time can be customized. Anyone who experiments with different countries recognizes that expensive electricity and few renewables in the electricity mix make Softiron particularly worthwhile.

It Helps Against Expensive Electricity.

An example, calculated with the simple version: With 1 PByte storage and storage on HDD in Germany in a data center with PUE 1.5, Softiron would save 94,616 tons of carbon dioxide over five years, according to the calculator. Almost 70,000 tons of carbon dioxide would continue to be generated. Electricity bills would shrink by around $83,400 over the same period, roughly the same amount of money in euros. Almost 53,700 euros would still have to be paid for electricity.

In Norway, with its primarily water-based electricity generation, all other things being equal, only 7,277 tons of carbon dioxide would be saved with around 13,700 tons of production. Soft iron would reduce electricity costs by around $24,000 to $15,459.