Home > Storage > Data Storage Essentials > Storage Admin > The Next Generation PowerMax - Balancing Performance with Efficiency and Sustainability > A Distributed Compute Platform for Storage Operations
Like the “PowerMax Brick” from the previous generation PowerMax, the PowerMax 2500 and PowerMax 8500 are built on modular 3U building blocks called “nodes”. Nodes contain the primary compute elements (CPU and memory) of the next generation systems. Each next generation PowerMax system has at least two nodes or a “node pair”. Each node has dual Intel® Xeon® Scalable processors, 24 DDR4 DIMM slots, up to 32 front end connection ports for host connectivity. Both the PowerMax 2500 and 8500 platforms can support OS block, Mainframe, and File workloads within the same system.
Each next generation PowerMax node uses a new switched PCIe fabric which has been optimized for both efficiency and performance. Each node’s PCIe fabric uses a partitionable, multi-rooted, switched architecture. This architecture enables advanced PCIe functionality that increases system configurability, optimizes system resource utilization, and availability.
The new architecture has allowed for replacing multiple discrete physical PCIe switches with a single partitioned switch. This shrinks the total cost of ownership by reducing power consumption, decreasing board space requirements, and lowering system interconnect costs. Additionally, the unified switch complex reduces system development costs through hardware reuse that enables a single hardware platform to serve many end markets and customer price / performance points. This is one of the key enablers which allows the reuse of the same node architecture in both the PowerMax 2500 and PowerMax 8500.
Using a multirooted partitionable PCIe architecture enhances resiliency on the PowerMax. Resources associated with a failing node root are dynamically reassigned to the remaining operational roots. This architecture allows any number of the remaining functional roots to take control of the isolated system resources to reestablish service with no interruption and no data loss.
The 64-lane PCIe switches are partitioned so that there are dual x16 lane connections to the root complex of the peer node. Each node has PCIe access with the other nodes components such as its memory and I/O modules. This access allows for the augmenting of compute resources for a fixed set of I/O and peripherals. This provides an overall system performance boost by allowing additional compute resources from multiple nodes to help process I/O coming in from the hosts and going to the storage.
Each node plugs into the Node Pair Enclosure midplane using direct PCIe connections. These connections are new designs which provide higher efficiency and higher bandwidth to the node’s I/O modules, Back-End Module (BEM – PowerMax 2500 only), Fabric Attach Modules (FAM), and Management Switch Modules (MSM). The internode x16 connections from the 64 lane switches use the midplane. These internode midplane connections are called Compute Module Interconnects (CMI).
The use of well-known and proven technologies in the initial design of the PowerMax 2500 and 8500 PCIe architecture was deliberate as these technologies’ performance, interoperability, and power efficiency profiles were well established.
The next generation PowerMax nodes incorporate intelligent Baseboard Management Controllers (BMC) and variable speed fan technology to help drive thermal and power efficiency. Each next generation node uses four 40mm variable speed dual-fan modules that house 16 total fan rotors. In conjunction with the midplane venting and heatsink solutions, the fans provide cooling for the CPUs, DIMMs, I/O modules and a management switch module.
The variable speed fans are designed to adjust their rotational speed based on the system's cooling requirements. Instead of running at a fixed speed, they dynamically adapt to the heat load generated by the components. This intelligent control enables precise temperature regulation and optimized cooling efficiency. By operating at lower speeds when the system is not under heavy load, variable speed fans reduce noise levels and power consumption. On the other hand, when the system experiences higher workloads, the fans automatically ramp up their speed to provide effective cooling, ensuring reliable performance under the most demanding conditions.
The fans are controlled by an intelligent Baseboard Management Controller (BMC) module. The BMC constantly monitors the following node thermal parameters:
On-board Ambient temperature sensor located at the cooling fan exhaust. This provides the inlet temperature to the CPU Module prior to pre-heating of any components
DIMM Thermal Sensor data – This data is collected by the CPU and is accessed by the BMC
CPU Thermal Sensor data – This data is accessed via the PECI interface on the CPU by the BMC
Power Supply temperature data – This data is accessed by the BMC over the I2C bus
Using these internal and external ambient temperature values, the BMC will automatically adjust the fan speed using a fan control algorithm so that internal node temperatures are always within specified ranges.
The combination of using intelligent BMC and variable speed fans provides numerous benefits in terms of thermal management and cooling efficiency for the next generation PowerMax platform. These benefits include:
Increased operational efficiency: The BMC's proactive management and optimization capabilities contribute to improved overall system efficiency, leading to better resource utilization, and reduced operational costs.
temperature regulation: Variable speed fans adapt their speed according to the system's cooling needs, maintaining optimal temperatures for the components and ensuring their longevity and performance.
Energy efficiency: Variable speed fans operate at lower speeds during periods of low system activity, reducing power consumption and minimizing energy waste. This not only saves costs but also contributes to a greener and more sustainable operation.
In summary, the intelligent board management controller and variable speed fans provide greater internal node temperature control, which in turn provides improved reliability, optimized cooling efficiency, and enhanced energy savings. These technologies play a crucial role in making the system more power-efficient and sustainable.