See who’s using Excelero
to accelerate their applications

Use Case

Leveraging NVIDIA GPUs in DGX-2 nodes, accelerate scientific research through machine learning and AI.
Use BeeGFS to share storage across cluster but provide the performance of local DGX-2 NVMe

The Challenge

Local DGX-2 storage is limited: The DGX-2 has only 30TB of local NVMe storage while a DGX-2 can process up to 14TB/s in a 16GPU Box. Copying data locally takes time and is inefficient.

Solution

STFC deployed a NVMe storage cluster with BeeGFS & NVMesh to support their GPU compute cluster. The storage cluster provides the required throughput with the option of future scale-out when required and enables seamless access and sharing of datasets for all machines on the network.

Giving GPUs access to unlimited NVMe resources boosts machine learning workloads

 

Leveraging BeeGFS & NVMesh to pump massive data streams into GPU Cluster

Case Study

Use Case

Leveraging NVIDIA GPUs in DGX-2 nodes, accelerate scientific research through machine learning and AI.
Use BeeGFS to share storage across cluster but provide the performance of local DGX-2 NVMe

The Challenge

Local DGX-2 storage is limited: The DGX-2 has only 30TB of local NVMe storage while a DGX-2 can process up to 14TB/s in a 16GPU Box. Copying data locally takes time and is inefficient.

Solution

STFC deployed a NVMe storage cluster with BeeGFS & NVMesh to support their GPU compute cluster. The storage cluster provides the required throughput with the option of future scale-out when required and enables seamless access and sharing of datasets for all machines on the network.

Improving GPU utilization 4 times

 

“Our clients cannot achieve the performance they need if GPUs are starved by slow storage or copying data to and from systems. Excelero NVMesh solved that problem for us.”
—Amine Kerkeni, Head of AI Product at InstaDeep

Case Study

Use Case

NVIDIA DGX enables Instadeep to run AI workloads for customers across industries such as Medical, Energy, Manufacturing …
The DGX is used by multiple scientists who run workloads for many different customers.

The Challenge

Local DGX Storage is too limited: the DGX only has 4TB of local storage while customers’ workloads require larger data sets.
Traditional external storage requires moving data to/from the DGX, which interrupts the workflow and is time consuming.

Solution

DGX systems connected over Mellanox IB fabrics to a Boston NVMe server powered by Excelero NVMeshprovides access to hundreds of TB’s of external high-performance storage.
DGX systems fully utilized to make perfect use of their resources and speed up InstaDeepcustomers deep learning workload

Improving GPU utilization 4 times

 

“Our clients cannot achieve the performance they need if GPUs are starved by slow storage or copying data to and from systems. Excelero NVMesh solved that problem for us.”
—Amine Kerkeni, Head of AI Product at InstaDeep

Case Study

Use Case

NVIDIA DGX enables Instadeep to run AI workloads for customers across industries such as Medical, Energy, Manufacturing …
The DGX is used by multiple scientists who run workloads for many different customers.

The Challenge

Local DGX Storage is too limited: the DGX only has 4TB of local storage while customers’ workloads require larger data sets.
Traditional external storage requires moving data to/from the DGX, which interrupts the workflow and is time consuming.

Solution

DGX systems connected over Mellanox IB fabrics to a Boston NVMe server powered by Excelero NVMeshprovides access to hundreds of TB’s of external high-performance storage.
DGX systems fully utilized to make perfect use of their resources and speed up InstaDeepcustomers deep learning workload

Turbocharging mission-critical data warehouse

With the new NVMesh-powered storage infrastructure, Bezeq achieved a 2x to 3x throughput improvement and cut query run times by up to 90%.

PRESS RELEASECASE STUDY

Use Case

Large-scale modeling, simulation, analysis and visualization.
Visualizes supercomputer simulation data on 128 monitors from a 128 node cluster.

The Challenge

Interactive work is generally small IOs.
Introduction of high-performance local NVMe SSDs create the problem of data locality.

Solution

NVMesh by Excelero enables NASA to create a petabyte-scale unified pool of high-performance flash distributed retaining the speeds and latencies of directly-attached media.

Turbocharging mission-critical data warehouse

With the new NVMesh-powered storage infrastructure, Bezeq achieved a 2x to 3x throughput improvement and cut query run times by up to 90%.

PRESS RELEASECASE STUDY

Use Case

Large-scale modeling, simulation, analysis and visualization.
Visualizes supercomputer simulation data on 128 monitors from a 128 node cluster.

The Challenge

Interactive work is generally small IOs.
Introduction of high-performance local NVMe SSDs create the problem of data locality.

Solution

NVMesh by Excelero enables NASA to create a petabyte-scale unified pool of high-performance flash distributed retaining the speeds and latencies of directly-attached media.

Bringing Medicaid Analytics into the future.

 

Double database query performance and lower overall TCO

CASE STUDY

Use Case

The Medicaid Data Warehouse was designed to enable faster and more insightful analytics by providing instant access to over a decade’s worth of data to over 3,000 simultaneous users. MDW data is stored in an Oracle RAC system; Oracle Business Intelligence provides the analytics functionality.

The Challenge

• The MDW holds double the volume the initial architecture was designed for. As a result, just 50% of queries could be completed in under 1 minute.
• The new MDW architecture needed improved query performance, facilitate further growth and lower overall MDW TCO.

Solution

The CMA Mosaic solution deployed by NYS DOH features an Oracle RAC cluster with 8 Dell R740xd storage nodes running Excelero’s NVMesh software. The solution enables NYS DOH to increase the percentage of ad hoc queries completed in under 1 minute to 98%. 99% are completed within 3 minutes.

Bringing Medicaid Analytics into the future.

Double database query performance and lower overall TCO

CASE STUDY

Use Case

The Medicaid Data Warehouse was designed to enable faster and more insightful analytics by providing instant access to over a decade’s worth of data to over 3,000 simultaneous users. MDW data is stored in an Oracle RAC system; Oracle Business Intelligence provides the analytics functionality.

The Challenge

• The MDW holds double the volume the initial architecture was designed for. As a result, just 50% of queries could be completed in under 1 minute.
• The new MDW architecture needed improved query performance, facilitate further growth and lower overall MDW TCO.

Solution

The CMA Mosaic solution deployed by NYS DOH features an Oracle RAC cluster with 8 Dell R740xd storage nodes running Excelero’s NVMesh software. The solution enables NYS DOH to increase the percentage of ad hoc queries completed in under 1 minute to 98%. 99% are completed within 3 minutes.

4K Streaming @ 60fps to 50+ Workstations Concurrently

“Now, with innovations like the shared NVMe storage such as Excelero provides, Technicolor can enjoy a hardware-agnostic approach, driving down costs and enabling flexibility for tomorrow while not sacrificing performance.” – Amir Bemanian, Engineering Director at Technicolor.

PRESS RELEASESOLUTION BRIEF

Use Case

Any-K production & post-production.
Streaming, editing and reel presentation to simultaneously active workstations.

The Challenge

4K & 8K video is changing requirements for storage performance and capacity.
Workstations need high BW & low latency performance for streaming, editing, and reel presentation.
Customers need future-proof solutions that will support 10k.

Solution

With just 6 standard servers powered by 3D NAND, NVMesh enabled 4K streaming @ 60fps to 50+ workstations concurrently.
The solution gives twice the performance of competing solutions at 50% of the cost and requires a much lower data center footprint.

4K Streaming @ 60fps to 50+ Workstations Concurrently

“Now, with innovations like the shared NVMe storage such as Excelero provides, Technicolor can enjoy a hardware-agnostic approach, driving down costs and enabling flexibility for tomorrow while not sacrificing performance.” – Amir Bemanian, Engineering Director at Technicolor.
PRESS RELEASESOLUTION BRIEF

Use Case

Any-K production & post-production.
Streaming, editing and reel presentation to simultaneously active workstations.

The Challenge

4K & 8K video is changing requirements for storage performance and capacity.
Workstations need high BW & low latency performance for streaming, editing, and reel presentation.
Customers need future-proof solutions that will support 10k.

Solution

With just 6 standard servers powered by 3D NAND, NVMesh enabled 4K streaming @ 60fps to 50+ workstations concurrently.
The solution gives twice the performance of competing solutions at 50% of the cost and requires a much lower data center footprint.

High-performance metadata tier
for Research Data Store

 

Leveraging extreme efficiency and unlimited scalability of NVMesh to supercharge Research Data Store

PRESS RELEASECase Study

Use Case

The Research Data Store (RDS) provides robust, reliable storage services to efficiently manage large volumes of research data throughout its life-cycle.
A high-performance metadata tier with scale-out NVMe needs to guarantee extreme performance for the entire cluster.

The Challenge

The Research center knows how much storage they need now, but growth is unpredictable so the metadata tier needs to be scalable as needed.
Research funds are limited so ICL needs to ensure the highest ROI for their NVMe storage.

Solution

Excelero’s NVMesh software enables and extremely fast metadata tier that spans multiple Dell R740XD servers. ArcaStream’s PixStor combines the different storage tiers into a single integrated solution, leveraging Mellanox Spectrum and ConnectX technologies.

High-performance metadata tier
for Research Data Store

 

Leveraging extreme efficiency and unlimited scalability of NVMesh to supercharge Research Data Store

PRESS RELEASECase Study

Use Case

The Research Data Store (RDS) provides robust, reliable storage services to efficiently manage large volumes of research data throughout its life-cycle.
A high-performance metadata tier with scale-out NVMe needs to guarantee extreme performance for the entire cluster.

The Challenge

The Research center knows how much storage they need now, but growth is unpredictable so the metadata tier needs to be scalable as needed.
Research funds are limited so ICL needs to ensure the highest ROI for their NVMe storage.

Solution

Excelero’s NVMesh software enables and extremely fast metadata tier that spans multiple Dell R740XD servers. ArcaStream’s PixStor combines the different storage tiers into a single integrated solution, leveraging Mellanox Spectrum and ConnectX technologies.

Pooling NVMe Within GPFS NSDs enables new Science use cases

 

80 pooled NVMe devices give 250GB/s of throughput and 20M random 4k IOPS

PRESS RELEASECASE STUDY

Use Case

Large-scale modeling, simulation, analysis and visualization.
Visualizes supercomputer simulation data on 100s of compute nodes.

The Challenge

Finish check pointing faster and start running the job.
Achieve performance of 250GB/Sec at the lowest price.

Solution

NVMesh by Excelero enables SciNet to create a petabyte-scale unified pool of high-performance flash distributed retaining the speeds and latencies of directly-attached media.

Pooling NVMe Within GPFS NSDs enables new Science use cases

80 pooled NVMe devices give 250GB/s of throughput and 20M random 4k IOPS

PRESS RELEASECASE STUDY

Use Case

Large-scale modeling, simulation, analysis and visualization.
Visualizes supercomputer simulation data on 100s of compute nodes.

The Challenge

Finish check pointing faster and start running the job.
Achieve performance of 250GB/Sec at the lowest price.

Solution

NVMesh by Excelero enables SciNet to create a petabyte-scale unified pool of high-performance flash distributed retaining the speeds and latencies of directly-attached media.

10x lower IO latency than Ceph through all layers in the virtual machine

Excelero’s NVMesh features a Cinder block storage driver that was designed to provide low-latency access to distributed, high-performance NVMe flash storage. Thanks to its unique RDDA technology and client-side archi- tecture, NVMesh enables 10 times lower IO latency and 20 times higher IOPs performance than Ceph.
PRESS RELEASECASE STUDY

Use Case

High-performance block storage for OpenStack.
New workloads such as machine learning and artificial intelligence are creating new opportunities for service providers but also bring unseen performance requirements for OpenStack Clouds.

The Challenge

Ceph is a popular storage solution for OpenStack, but the layered architecture has a big impact on the performance and makes the solution less attractive for new low-latency workloads.

Solution

TeutoStackCloud runs NVMesh on SuperMicroservers from Boston Limited along with MellanoxConnectX-4 and ConnectX-5 NICs running 2x25Gb Ethernet to each client node. It uses OpenStack Cinder to manage storage though the NVMesh Cinder driver.

10x lower IO latency than Ceph through all layers in the virtual machine

Excelero’s NVMesh features a Cinder block storage driver that was designed to provide low-latency access to distributed, high-performance NVMe flash storage. Thanks to its unique RDDA technology and client-side archi- tecture, NVMesh enables 10 times lower IO latency and 20 times higher IOPs performance than Ceph.
PRESS RELEASECASE STUDY

Use Case

High-performance block storage for OpenStack.
New workloads such as machine learning and artificial intelligence are creating new opportunities for service providers but also bring unseen performance requirements for OpenStack Clouds.

The Challenge

Ceph is a popular storage solution for OpenStack, but the layered architecture has a big impact on the performance and makes the solution less attractive for new low-latency workloads.

Solution

TeutoStackCloud runs NVMesh on SuperMicroservers from Boston Limited along with MellanoxConnectX-4 and ConnectX-5 NICs running 2x25Gb Ethernet to each client node. It uses OpenStack Cinder to manage storage though the NVMesh Cinder driver.

Pooling NVMe enables new Science use cases at NASA

“By reducing random access latency to large data sets, humans can interact with the computing environment in a more natural fashion”

PR - THE REGISTERCASE STUDY

Use Case

Large-scale modeling, simulation, analysis and visualization.
Visualizes supercomputer simulation data on 128 monitors from a 128 node cluster.

The Challenge

Interactive work is generally small IOs.
Introduction of high-performance local NVMe SSDs create the problem of data locality.

Solution

NVMesh by Excelero enables NASA to create a petabyte-scale unified pool of high-performance flash distributed retaining the speeds and latencies of directly-attached media.

Pooling NVMe enables new Science use cases at NASA

“By reducing random access latency to large data sets, humans can interact with the computing environment in a more natural fashion”

PR - THE REGISTERCASE STUDY

Use Case

Large-scale modeling, simulation, analysis and visualization.
Visualizes supercomputer simulation data on 128 monitors from a 128 node cluster.

The Challenge

Interactive work is generally small IOs.
Introduction of high-performance local NVMe SSDs create the problem of data locality.

Solution

NVMesh by Excelero enables NASA to create a petabyte-scale unified pool of high-performance flash distributed retaining the speeds and latencies of directly-attached media.

Learn How Excelero Fits Your Business

EXCELERO'S SOLUTIONS