Periodic Reporting for period 1 - Fabriscale (Fabric management for the software defined data centre) Reporting period: 2016-09-01 to 2016-12-31 Summary of the context and overall objectives of the project Fabriscale creates software, called a fabric manager (FM) that controls computer networks. Think of a FM as software similar to the air traffic control service found at all airports. Air traffic control is responsible for organising and expediting the flow of air traffic, while Fabriscale’s fabric management software is responsible for configuring and optimising the communication going on in a computer network. Fabriscale specializes in smart algorithms that simplify network configuration, management and routing, which improves network performance, network reliability and network monitoring beyond what existing products can provide.Whenever a link or node fails in a compute cluster, an FM needs to reconfigure switches in the cluster such that the nodes in the cluster remain connected. If done incorrectly, reconfiguration can lead to degraded performance, causing compute jobs to slow down or fail, which can cost more than € 9 000 per minute[1]. This is especially devastating for large clusters with thousands of nodes and the severe consequences are that critical medical, industrial and societal services hosted in the cloud are interrupted.The goal for phase one was to complete a feasibility study for Fabriscale products with the following main objectives:* Complete a analysis of the worldwide market for fabric managers.* Finalise agreements with two pilot customers that can be involved in product demonstration activities (e.g. testing, piloting, validation, performance verification).* Identify potential partners and patent obstacles in the US and European market for fabric managers.* Analyse FM value chains and develop refine the model for the US and European market.[1] ‘Cost of Data Center Outages’, Ponemon Institute LLC, Jan. 2016 [Online]. Available: http://www.emersonnetworkpower.com/en-US/Resources/Market/Data-Center/Latest-Thinking/Ponemon/Documents/2016-Cost-of-Data-Center-Outages-FINAL-2.pdf Work performed from the beginning of the project to the end of the period covered by the report and main results achieved so far The main work performed in the project was the market study, the partner search and the business model development. In the market study we analysed the worldwide market for fabric managers and identified critical pain points for customer through direct interaction with potential customers and through a survey of 60 IT companies together with IDC. Our talks with customers show a great interest in Fabriscale products and the survey results show that 59% of all respondents considered existing tools average, poor or very poor. Which show that there is space for significant business value play for Fabriscale products. We also gain a better understand of the the market size with detailed numbers for our primary markets in the US and in Europe, as well as an outline of the market in Asia.The pilot and partner search focused on pilots present i Europe in order to reduce travel costs and overhead in the pilots planned for 2017. We have identified three partners for our InfiniBand technology and two for Ethernet. All partners have observed pain points in line with the results from the market survey and we believe they will be excellent early evaluators of our products.The business modelling activity focused on identifying the needs of different customer segments, refine the business model in order to maximise revenue for our products, and identifying the right partners for distribution of our products. The main results for this activity is an ongoing negotiation with two US distributors about a re-seller agreement in the USA. Progress beyond the state of the art and expected potential impact (including the socio-economic impact and the wider societal implications of the project so far) A trend in computing is that all larger computing systems are located in a data centre. These systems include traditional HPC clusters and infrastructure for private and public clouds. A common challenge for this type of large systems is that the larger they become the more challenging it becomes to balance the load and handle failures. The FFM provides excellent load balancing of the network in order to achieve high performance and fast recovery from link and switch faults in order to reduce downtime. These features gives facility owners the flexibility to decide whether to increase revenues by adding more revenue-generating compute jobs to their system, or decrease costs and energy consumption by decommissioning unneeded computing nodes and network equipment. These feature and performance capabilities can be summarize as follows:* Increase network throughput of a system by up to 40% compared to the current state of the art.* Reduce recovery time from several minutes to less than a second (up to 400x improvement depending on system size).* Reduce energy consumption if the rise in efficiency is used to decommission unneeded equipment, which leads to a reduction in CO2 emissions through reduced electricity usage.* Ease the management of data centers by a unique monitoring system and integration of the Fabriscale Fabric Manager with de-facto standard resource managers. Company logo
