HPE Expands GreenLake

Comments

This is a significant launch for HPE covering a domain they wish to play a key role in, capitalizing on the GreenLake success and recognition.

Unified DataOps, wow it seems that every word is important here. It is supported by 3 products and/or solutions named Data Services Cloud Console, Cloud data services and Alletra. Thus it is focused on infrastructure management but here a few remarks we'll develop later.

HPE recognized that the cloud model is a must, it's never too late, we're almost mid 2021. Users expect to be agile and elastic and wish to consume every IT resource as-a-service with a pure software model, charged per unit of time, incrementally like the cloud. In fact, the model should be independent of the infrastructure spanning potentially private and public data centers and clouds, resources could potentially reside anywhere.

We notice once again that the firm consolidates on acquisition as Aruba and Nimble InfoSight served as roots for these new service offerings. With credible AI embedded services, the engineering team has realized some clever developments that will simplify the modern data center operations targeting long and complex tasks. The idea of programmable data center or infrastructure makes some progress. There are some interesting features such intent-based provisioning that combines the global knowledge of the storage environment, applications workloads, templating, learning phase and historical informations. At scale users also require high automation levels to deliver more reliable services faster.

Data is used everywhere because infrastructure data is the fuel of these solutions but promoting a new data vision, we have expected different products or not only these ones. As said, it is a major launch for the company to regain some credibility in the data center and cloud-based domain. HPE continues to release block storage and again nothing on file and object storage. We can even ask ourselves if HPE really believes in file and/or object? For such solutions they choose to rely on third party software such CTera Networks, Cohesity, Qumulo, Scality, WekaIO, IBM Spectrum Scale and Nutanix, if we include them in the storage domain, all good proven products, no doubt about them. We really hope HPE will change and decide to be serious - to own something - in these areas with some acquisitions. Some of these names will fit pretty well in the HPE portfolio and address immediately holes. But again it has to be done carefully as PolyServe and Ibrix acquisitions have been disasters.

The company declared that cloud-native applications are paramount for enterprises, they support of course - it's a minimum CSI, but it continues to not have their own object storage solution for such needs as they chose to support Scality again in such environments. This is a bit paradoxal, and, at the same time, consistent with its choice for classic object storage needs. But as an investor for several years in the company via its Pathfinder investor force, we all wonder why HPE didn't make the move and offer a price to acquire the software vendor. Probably some divergence about the number of dollat digits... But notice that HPE didn't make the move for Portworx or Hedvig, even with some investments in them.

HPE Alletra storage array line is a surprise as the storage group already released some time ago the Primera line. The description of the 2 models - 6000 and 9000 - are a bit fuzzy apart being a full NVMe array in 4U chassis with 24 and 48 drives respectively. In such dimension, the 9000 series offers 656TB raw for 2,000TB effective after compaction, compression, de-dupe and thin provisioning coupled with RAID-6 (10+2 NVMe drives associations) data protection service. Host interface could be FC or Ethernet and 2 cache sizes are available: 1,024Gb for the 9060 or 3,072GB for the 9080 model.

We understand they represent the new generation of Primera and Nimble arrays respectively. With the 9000 model, the firm claims to deliver 100% availability, which is ambitious, and at the same time dangerous. It is based on an all active multi-node model leveraging a real massively parallel architecture. This high-end model delivers 2 million IO/s per chassis. The 6000 model is positioned to deliver 6 nines of availability.

But HPE had to react and innovate against the competition. At the storage strategy level, we wish to insist that outside block, we don't see the company playing a serious role as they don't own key software. We really expect to see some other developments as HPE promotes unified data management and it should go beyond block. Block storage is good but for a company like HPE, this is clearly not enough.