HPE enhances GreenLake for File Storage

HPE is upgrading GreenLake for File Storage to deliver a sevenfold increase in capacity and throughput, aiming to overtake Dell PowerScale and Pure Storage FlashBlade systems in early 2024.

On-premises GreenLake for File Storage was announced in April this year, and uses Alletra MP storage hardware twinned with VAST Data software. This has a disaggregated shared everything (DASE) architecture, and Alletra MP has one or more 2RU controller chassis (two CPU nodes per chassis) talking across a 100 GbitE NVMe fabric to one or more 2RU capacity (JBOF) nodes. These come with up to 20x NVMe SSDs (7.58 TB, 15.35 TB) and four Kioxia XL storage-class memory drives (800 GB, 1.6 TB encrypted) for fast metadata operations. HPE is adding support for 30 TB SSD and 400 Gbps Nvidia Quantum-2 InfiniBand connectivity support.

HPE storage marketer David Yu blogs: “These new enhancements can accelerate AI data aggregation and data preparation and speed AI model training and tuning, including for GenAI and Large Language Models (LLMs).”

HPE Alletra MP Diagram
HPE Alletra MP Diagram

Yu says the 30 TB SSDs plus Quantum-2 connectivity and some more but unspecified enhancements “can increase the capacity density and throughput by seven times” compared to “the shipping product featuring 307 TB 2RU storage shelves using 15 TB NVMe SSDs and delivering aggregate read throughput of up to 200 GBps.”

VAST Data has just released v5.0 of its software and this may be included in the coming HPE enhancements.

GreenLake for File Storage can now scale up to 250 PB effective capacity, assuming a 3:1 data reduction ratio, meaning 83.33 PB raw, and provides >700 GBps read throughput with >200 GBPs write throughput.

Yu says these enhancements will enable GreenLake for File Storage to deliver up to 2.3x the read throughput per unit of raw capacity, and also 2.3x the raw capacity, of Dell PowerScale and Pure FlashBlade systems.

GreenLake for File Storage’s Quantum-2 switch support means it gets remote direct-memory access (RDMA) and GPUDirect technology. Yu says this enables it “to accelerate AI workloads and outcomes by boosting performance in the model training and model tuning stages, including faster checkpointing.”

Yu says the product has automatic discovery, onboarding, and configuration of new devices for file share creation via a single cloud console, and more than 50 data services, including file sharing, lifecycle management, hybrid cloud backup, and disaster recovery. It also supports non-disruptive upgrades. 

The GreenLake for File Storage 30 TB SSDs and Quantum-2 connectivity are available to order now. The additional enhancements will be delivered around the first 2024 quarter.

Read a GreenLake for File Services tech overview here and watch a demo here.