As the technology world stares down exascale computing the need for faster and faster interconnect solutions is critical for delivering the requirements of super low latency and high throughput. Mellanox offers end-to-end InfiniBand and Ethernet products aimed at the cloud and Big Data. We caught up with Eyal Gutkind, Senior Manager, Enterprise Market Development of Mellanox, to learn more.
insideAI News: What products does your company offer the Big Data sector?
Eyal Gutkind
Eyal Gutkind: Mellanox Technologies offers analytics and file systems accelerations for Big Data deployments, using Mellanox 40Gb/56Gb Ethernet and InfiniBand solutions. With Remote Direct Memory Access (RDMA), Mellanox enables an efficient data transfer mechanism that allows applications, such as IBM BigInsights, to increase customer data ingress rate by more than 35 percent and analytics data processing by 50 percent compared to other data transport solutions.
With a novel data analytics approach, Mellanox is providing double performance to data Extract, Transform, Load (ETL) solutions based on Hadoop Map Reduce. The code to enable such performance gains is part of the Hadoop community code. With acceleration of the networking stack, Mellanox is quadrupling the number of clients served on a single Memcached server, a dominating key-value caching service for large-scale web applications.
Big Data and Cloud storage solutions leverage the practically unlimited bandwidth provided by Mellanox’s End-to-End solutions. The complete solution includes SwitchX-2 switching systems both for 40Gb/56Gb Ethernet and InfiniBand, coupled with ConnectX-3 Pro and Connect-IB Network Adapters. Software Defined Storage solutions such as Ceph, IBM GPFS, Lustre and others, take advantage of the unbounded bandwidth as well as the smooth data transfer of RDMA, resulting in lower CPU and power consumption per gigabit transferred. This translates to real dollars in the customer OPEX and CAPEX budget by saving on HVAC infrastructure, physical space and monthly energy bills.
To summarize, Mellanox Big Data solutions help our customer to increase data value with more efficient data analytics and storage.
insideAI News: Who are your customers in this market currently and who do you foresee your customers being in the future?
Eyal Gutkind: We are seeing an exponential growth in applications that generate and consume data. There are an increasing number of usage models and applications that utilize this amount of data for making better business decisions. As the amount of data increases, there is a greater need to move it faster and more efficiently, and Mellanox solutions are the best fit for that. PayPal is an example of a company that takes advantage of the Mellanox interconnect solutions for real time fraud detection. There are supermarket chains that leverage Mellanox capabilities in order to analyze customer habits and to run the right programs. Mellanox has many other customers in several industries, including national security, healthcare and transportation. We expect that in the future, more applications around the Internet of Things will leverage Mellanox solutions.
insideAI News: Mellanox was recently involved in the Massachusetts Open Cloud Project designed to run Big Data and other exciting projects for the state. How did your company get involved in this initiative?
Eyal Gutkind: Mellanox solutions power the fastest supercomputers in the world, and deliver higher performance and more efficient cloud infrastructures. We also allow for quicker data analytics. As such, Mellanox was invited to join the Massachusetts Open Cloud Project to lend our expertise to the project.
insideAI News: What did you learn from this project?
Eyal Gutkind: As the Massachusetts Open Cloud Project is still in its early days, we expect to continue to collaborate with the vendors and end-users to enable better cloud and big data solutions. We already contribute our core technology and plan to also invest both on the hardware and software level as the project moves forward. We also recently announced CloudX, an efficient cloud platform that is being built via a large eco-system where Mellanox contributes fast interconnect solutions. We do expect that CloudX will be part of this project as well.
insideAI News: Are there more big projects like this on the horizon?
Eyal Gutkind: Definitely, there are more projects and collaborations worldwide involving Exascale computing, cloud computing, data analytics and more. As these projects become public we will be happy to share.
insideAI News: What can we expect from Mellanox as Exascale comes into focus?
Eyal Gutkind: Mellanox has been paving the road to Exascale for several years now. We continue to lead the interconnect technology market and continue to enhance its capabilities in order to enable larger scale and higher performance at lower power consumption. We recently announced new ways for reliable data communication that for the first time are independent of system size, and are set for the Exascale machines. We work closely with our partners and customers on the various Exascale projects around the world and will continue to drive the technology further.
Speak Your Mind