Greetings! In today’s blog I would like to share the architecturally driven performance benefits HyperFlex demonstrates as seen in Enterprise Strategy Group (ESG) findings, and invite you to our HyperFlex first year anniversary on March 22: More Apps, More Simplicity, The New HyperFlex.
I joined Cisco last year as the VP of Product Management to lead strategy for Data Center Compute, Storage and Hyperconverged Infrastructure (HCI) with particular focus on the roadmap and market execution for HyperFlex. Over the last year, we have proudly shared Cisco’s differentiated architectural approach aiming to deliver the best foundation for hyperconverged infrastructure (HCI) in your data center. Recently, tests conducted by ESG not only demonstrate the performance benefit that you can see today, but also validate the importance of this approach as you migrate more critical workloads to HCI.
Dynamic Data Fabric: Engineering HX Data Platform on Cisco Unified Compute & Fabric
Cisco HyperFlex is engineered on UCS, providing a unified, low-latency and high-bandwidth networking fabric that is at the core of this scale-out architecture. Networking that is thought through and engineered is key for delivering a high-performance, resilient and highly available HCI. The foundation of UCS provides a networking architecture that is fast and robust, yet simple to deploy and manage. In addition, UCS provides policy based configuration and automation for a multitude of critical items such as server firmware, QoS policies and VLAN definitions for different data types that have been traditionally manual and error-prone. This approach not only delivers rapid deployment and expansion, but also ensures optimized utilization of all the compute and storage resources in the hyperconverged cluster.
The HX Data Platform is a log structured filesystem written from the ground up specifically for hyperconvergence. It is designed for scale-out and for optimal utilization of resources across the cluster. It uniquely pairs the underlying network foundation with a low overhead software data path to deliver consistent low latency and enable higher VM density. The HX Data Platform is also flash optimized, sequentializing SSDs writes in order to reduce flash wear while delivering extremely fast, low latency performance. Since this platform provides flexible and independent scaling of compute and/or capacity resources, customers can easily grow their environment while optimizing TCO for their specific needs.
Together, the Cisco Dynamic Data Fabric approach – engineering the HX Data Platform on Cisco UCS compute and fabric – provides end-to-end and complete hyperconvergence, allowing you to achieve significantly better infrastructure utilization with predictable and leading performance.
What it means for you
3X higher density: You can expect to do a lot more; running more applications and VMs on HyperFlex than you would on competing platforms at similar latency. In the chart below you see HCIBench demonstrating the number of virtual machines that can be supported by the cluster with a 5ms write latency constraint. The optimized performance of HyperFlex results in significantly more virtual machines on a similar hardware configuration, so you can dramatically improve the TCO in your environment.
High performance with low latency: As customers migrate more workloads to HCI, the importance of a high, consistent performance increases. Here, you can see an HCIBench test running 140 VMs. Cisco HyperFlex delivers significantly better latency performance, providing customers the confidence to mix business applications into your HCI environment.
Predictable IOPS delivery: Another important consideration moving forward is predictable performance. Transitioning to All-Flash configuration can help reduce latency in any solution, but you may still have performance variability, inhibiting your ability to deploy applications. In ESG’s testing using HCIBench, you can see stable and predictable performance across all virtual machines, while competing solutions provide unacceptable level of IOPS for some of the VMs in the cluster.
These results are not an accident but the direct outcome of our Dynamic Data Fabric approach providing customers high, consistent and predictable performance. With this foundation in place, HyperFlex provides a comprehensive data center migration path to HCI – utilizing a single management policy across your environment, providing cloud-like resource elasticity and allowing you to expand your HCI infrastructure to more use cases and applications.
Source : http://blogs.cisco.com/datacenter/the-new-hyperflex-more-performance-more-workloads-more-efficiency