Unmatched Performance with MapR XD and Cisco UCS


Process, store, and ingest data faster than any other solution

Record-breaking speed

Together, Cisco and MapR hold ranks 1 and 4 for best 1-TB results in the TPCx-HS transaction processing and database benchmarks. They also hold ranks 1,2, and 5 for 10-TB results.

Cisco UCS M5 servers, utilizing the Intel Xeon processor, are able to provide world-record-setting performance:

  • 86% better performance
  • 50% better SAP Sales and Distribution performance
  • 78% faster Java application performance

Reduce time to deploy with MapR XD built-in analytics

Analyze data without having to move data out of the cluster once analytics are complete, reducing time to deploy and reducing set-up costs for dedicated clusters.

Increased efficiency

With advanced features such as an optimized shuffle algorithm, direct access to disk, built-in compression, and a framework written entirely in C++, which also increase performance, the MapR XD & Cisco platform overcomes bottlenecks in traditional architectures, processing 25% more data in 10% of the time.

Built for demanding applications

MapR Cloud-scale data store is built to handle new demanding applications and I/0 patterns that are commonplace with modern processing architectures like GPU and modern media architectures like NVMe and 3DXPoint. MapR XD is built to run extremely fast media types and provide the I/0 needed for real-time IoT operations, next generation intelligent applications, Augmented Reality/Virtual Reality driven research and applications as well as deep analytics. As performance continues to grow new technologies are constantly tested and validated.

Real-world results

Using the joint Cisco-MapR solution, Quantium cut query time by 92%, resulting in a 12.5x increase in overall performance.

Liasion found their computational power to be 80% greater than before they implemented MapR on Cisco UCS.

Management Science Associates (MSA) were able to sort through 1TB of data distributed across 8 nodes in under 16 minutes.

Handle more bandwidth with low latency

Platform responds to data processing challenges

Process massive workloads at high speeds with software that is designed to do so. The MapR Data Platform is geared towards prioritizing jobs and data to get the most out of big data.

MapR XD is POSIX-compliant and supports read/write operations on existing data (unlike standard HDFS), applications can access the data stored in MapR XD through two methods – NFS an POSIX. Utilizing the two results in faster data ingestion.

MapR XD can update in place without compactions to deal with, resulting in 5x greater performance
Cisco servers are based on the latest Intel Xeon scalable CPUs with up to 28 cores per socket, delivering significant performance with up to 1536 gigabytes of main memory.

Nexus 9000 Series Switches help prioritize jobs intelligently, maximizing effectiveness and increasing output. Using MapR XD running on UCS servers, MSA found they could spin up their servers in minutes, rather than hours or days.

Cisco UCS 6300 Series Fabric Interconnects

Offer higher capacity, higher port density, and lower power consumption with Cisco Fabric Interconnects. These efficient interconnects reduce costs while expanding the UCS networking portfolio.

  • Centralized, unified management with Cisco UCS Manager
  • Policy-based virtual machine connectivity with VM-FEX
  • Single management infrastructure frame for all applications

Rolling Upgrades

Decrease latency with rolling upgrades functionality, which ensures that relevant software updates are delivered across the infrastructure in an efficient and timely manner, without affecting performance.

MapR Direct Access Network File System (NFS)

Enable applications that write to and update data on Hadoop even while analysis is in progress. MapR XD Direct Access NFS enables random read-write capability, decreasing latency when jobs are running. Direct access NFS works seamlessly with non-Java applications such as MySQL, SAS, SAP and others.

Eliminate downtime and data loss automatically

No Single Point of Failure

Because of distributed name node functionality, the data lake continues to function even when one of its nodes fail. With information written across hundreds or thousands of nodes, information is easily re-synced in case of node failure.

With failover protection, the MapR Data Platform and Cisco Integrated Infrastructure for Big Data delivers 99.999% uptime.

Cisco UCS is designed with redundant components and uses model-based management to provision servers automatically, regardless of form factor. By simply associating a model with a resource through Cisco UCS Manager, your IT staff can consistently align policy, server personality, and workloads. These policies can be created once and used by IT staff with any level of experience to deploy servers. The result is improved productivity and compliance, greater availability, and lower risk of failures due to inconsistent configuration.

Point-in-Time Recovery

Easily recover lost data thanks to point-in-time saving. MapR XD technologies automatically takes snapshots to allow easy tracking and recovery of the data lake’s previous states.

Data Mirroring

Double UCS deployments with MapR XD data mirroring technology, ensure multiple copies of data are spread across nodes, clusters and racks. This mitigates the risk of any data loss and ensures data is reconstructed at a fast pace.