Dell Sets Standard for Unlocking Potential of Data With PowerScale Storage

Representing new Isilon generation on new hardware

This is a Press Release edited by StorageNewsletter.com on June 18, 2020 at 2:20 pm

Jump right to comments

News summary

PowerScale storage systems bring together the best in Dell EMC server hardware and storage software to help customers manage file and object data in core data centers, edge locations and public cloud
PowerScale OneFS operating system introduces new software capabilities, including enhanced data reduction technology, S3 object access and support for Ansible, Kubernetes and OpenShift
Dell EMC DataIQ software helps organizations discover, understand and act on unstructured data across private and public cloud storage
Supported by Dell Technologies On Demand with flexible payment options to acquire storage capacity as needed

Dell Technologies, Inc. announces PowerScale, a family of storage systems engineered with storage software and server hardware to set a standard for how organizations capture and capitalize on unstructured data, such as documents, images, videos and social media content.

“The amount of unstructured data enterprises store as file or object storage is expected to triple by 2024, and there are no signs of it slowing,” said Dan Inbar, president and general manager, storage. “In this data era, businesses need a simple, seamless and cost-effective way to store and use unstructured data to innovate, create differentiation and bring products to market faster. The Dell EMC PowerScale family provides the foundation companies need to unlock the potential of their data, no matter where it resides, and use it to drive meaningful business impact.”

Unlocks the potential of unstructured data
PowerScale runs on the next generation of OneFS, the operating system known for powering Isilon. The family features 1U PowerEdge-based PowerScale nodes and existing Isilon all flash, hybrid and archive nodes running the PowerScale OneFS 9.0 operating system.

PowerScale delivers the performance that customers need to handle demanding AI, analytics, IoT, digital media, healthcare and life sciences workloads. It can deliver up to 15.8 million iIO/s, and all-flash PowerScale nodes are up to 5x faster than its predecessor. Enhanced inline data reduction makes the platform up to 6x more efficient.

Simplicity at any scale
PowerScale can start small and grow to massive, petabyte-scale while remaining easy-to-use:

Scale without disruption: PowerScale clusters can scale from 11TB raw capacity to 60PB and millions of file operations without disruption or costly downtime for customers. Nodes can be added to either PowerScale or an existing Isilon cluster in just 60s.
Intelligent automation: With smart scale-out capabilities, PowerScale distributes resources effectively so that customers can get the most performance out of a cluster.
Resilient and efficient: Through flexible failover policies, it delivers up to 85% storage utilization across a cluster and can sustain multi-node failures.
Programmable infrastructure: With support for a number of leading management and container orchestration frameworks, such as Kubernetes, Ansible and OpenShift, customers can streamline application development and reduce deployment timeframes.

Intelligent insights
Dell makes it easy for customers to understand their data and storage infrastructure health through software included with PowerScale:

Put data to work: The introduction of DataIQ software helps companies extract business value from unstructured data, typically uncategorized and found in siloes throughout businesses. It breaks down data siloes by delivering a single view of file and object data across Dell EMC, third-party and public cloud storage. Users can gain better control over their data, ensure the right teams have access to it, and make the most of their investment by ensuring data is stored on the right tier within their storage environment.
Proactive health monitoring: CloudIQ infrastructure monitoring and analytics software combines machine learning and human intelligence to provide customers with real-time performance and capacity analysis as well as historical tracking for a single view of Dell EMC infrastructure.

Any data, anywhere
PowerScale supports a variety of file protocols and customers can easily deploy it to meet their infrastructure needs:

Any application: PowerScale OneFS 9.0 features broad multiprotocol support, including new S3 support for apps relying on object storage. Additional support for protocols including NFS, SMB and HDFS, allow customers to run a large number of traditional and modern applications without compatibility concerns.
Deployment flexibility: PowerScale is easily deployed in core data centers, in edge locations or as part of a multi-cloud strategy. The 1U footprint and minimum cluster size of new all-flash and NVMe PowerScale nodes make it an option for edge deployments.
Multi-cloud support: PowerScale for Multi-cloud can directly connect to all major public clouds as a managed service – for customers looking to move or deploy demanding apps in the cloud. Google Cloud customers looking to save time and management complexity can choose Cloud PowerScale for Google Cloud, a native cloud service that combines the performance, scale, and consistent experience of PowerScale with the economics and simplicity of Google Cloud.

Flexible consumption with Dell Technologies On Demand
With Dell Technologies On Demand, PowerScale customers can respond to workload spikes and new service requests with elastic capacity and cloud economics. Several flexible pay-per-use choices with short-and-long term commitment options are available, including a one year term for flexible consumption.

Availability: PowerScale OneFS 9.0, PowerScale nodes and DataIQ are now available.

Keith Bradley, IT manager, Nature Fresh Farms, said: “When you buy our tomatoes, cucumbers and peppers, what you don’t see is the sheer amount of data created by taking that produce from the greenhouse to the grocer. Nature Fresh Farms is a data-driven operation that needs storage that can scale-out without causing costly technology refreshes. That’s why we’re an Isilon customer today, and why we’re so excited about PowerScale. New nodes at the edge that integrate seamlessly with our existing clusters will be a gamechanger.”

Ulrich Betzler, senior storage architect, Steinbuch Centre for Computing at Karlsruhe Institute of Technology, said: “New S3 protocol support in PowerScale OneFS is highly impressive and exceeded our performance expectations. PowerScale is a formidable platform, and we think it will play a critical role in our IT infrastructure moving forward.”

Amita Potnis, research director, Enterprise Infrastructure Practice, IDC, said: “Organizations seeking out ways to manage the exponential growth of unstructured data must understand that this is not just a storage issue, it’s a holistic data management challenge. PowerScale allows companies to easily capture unstructured data, and when combined with new DataIQ software, companies can better understand how to make that data can work for them. PowerScale’s software-defined architecture allows the platform to easily handle today’s unstructured data challenges while not losing sight of what the future might hold.”

Mark Graham, vice president, Partner Business Development, Trace3, said: “PowerScale’s ease of use, innovative new data management capabilities and flexible deployment options – from edge to core to cloud – represents a massive opportunity for the unstructured data storage market and the channel. Trace3 is eager for our customers across all our market segments to see what it can do for their data centers.”

Resources:

Comments

It's a strategic announcement for Dell EMC.

First, the title of the press release include the words Sets New Standard for Unlocking the Potential of Data. It's surprising and we're sure our readers will share our point of view when you read this product announcement and the analysis below.

In a nutshell Dell EMC confirms his follower status and no longer an innovator like it was in the past when we see and compare what other vendors release on the market. But did Dell EMC ever innovate in the file storage space? Except with evolutions following acquisitions like Exanet acquired by Dell in 2010 for $12 million, EMC on his side acquired CrosStor Software (formerly Programmed Logic Corp.) in 2000 for $300 million, Rainfinity in 2005 for approximately $100 million and Isilon in 2010 for $2.25 billion. And were EMC Celerra and HighRoad, later MPFS, real innovations or just me too products? Even for Centera, it comes from an external growth with the acquisition of FilePool in 2000 for approximately $50 million. Investor in Exablox and Elastifile, Dell EMC even let others acquirers bought these companies having understood it won't go anywhere. The surprise is Dell EMC has and owns everything to innovate in that file storage space at least at the same level but its big company status, inertia and lacks of decision, hesitations probably, even internal politic games... can freeze directions and ideas and delay decisions. In fact, it illustrates once again that innovation comes from small agile players and not big irons. Let us explain.

With that launch Dell EMC releases an association of hardware and software, a bit like what they did when they announced PowerStore a few weeks ago. In other words, the equation seems to be PowerXXX=hardware+software and here PowerScale means Dell servers+OneFS 9.

The vendor realizes that it needs now a convergence between its file and object offerings as users asked for this and competition already delivers it. Today its object storage product - ECS (Elastic Cloud Storage) - exposes S3 and NFS among others and PowerScale, historically a NAS platform under the Isilon brand, adds S3 now. Never too late but late. It means also that Isilon brand is dropped - look at the photo at the top of this article - as it was already the case when Dell EMC announced OneFS for Google Cloud recently. No surprise Isilon was absorbed 10 years ago and the name is famous.

PowerScale, the hardware
Let's look at the new hardware that supports PowerScale, it means here the 1U boxes, model F200 and F600, based on PowerEdge servers. The first thing is the absence of performance numbers (IO/s and throughput) for these 2 machines in the hardware specification sheet here while we find numbers for previous system line named Isilon F800 and F810 in the same document just one paragraph below.

In term of processor, no real details are given except single and dual socket Intel processors respectively for F200 and F600. As of today Intel doesn't supports PCIe Gen 4 with Xeon, it means that these machines can't run this level without swapping components if they will receive a new Xeon line or others. There is a challenge here as Intel prepares Optane PCIe Gen 4 but no processor yet to support it. At the same time AMD releases EPYC processors already supporting PCIe Gen 4 and guess what Dell has some PowerEdge servers in his line - R6515R and R6525R - built with these AMD EPYC products respectively with 1 or 2 processors. Tough to understand the logic here.

We also don't understand why PowerStore accepts Optane and even full Optane appliance and PowerScale not. And these 2 servers mentioned above can support up to 10 or 12 2.5" NVMe SSDs more than the F200 and F600 can, respectively 4 and 8. And the F200 supports flash drives without a NVMe connectivity, only the F600 receives NVMe SSDs. It seems that the F200 and F600 have more networks capabilities than these other models, a key element to consider in a scale-out mode and especially with a share-nothing architecture therefore with natural inter-node messaging and data exchanges. Definitely the F600 node is an interesting compromise and 1U systems are a good approach for the edge.

This release maintains the same limitation of 252 nodes or 63 chassis of 4 nodes each for variable capacity maximum, the full flash F models can reach 58PB and even 60PB for the hybrid and archive models.

Like PowerStore, the message around programmable datacenter and management integration is the same with support of Ansible, Kubernetes and OpenShift.

OneFS 9, the software
In term of data services we continue to find FlexProtect, SmartPools, SmartLock, SmartQuotas, SmartConnect, SmartDedupe, SnapshotIQ, SyncIQ, InisightIQ and CloudPools. SmartDedupe has made some progress with more reduction capabilities. The company continues to support NDMP, an archaic method. Nothing really changes here, striping is done at the file level and protection continues from N+1 to N+4. Very tough to make changes at this level as it impacts lot of other codes and behaviors of the system, so 85% of storage efficiency, very good ratio, means something like 22+4, 22 data fragments and 4 parities fragments, or 17+3, in a Reed-Solomon schema.

The good decision was to add a S3 engine to expose the datastore via this protocol. This datastore can be exposed at the same time via NFS or SMB and others thus limiting extra copies, consumed space and loads on the machines and of course avoiding potential data divergence. This is already offered by Qumulo with the MinIO gateway embedded for more than a year and by Vast Data as well. Among top file storage vendors and especially NAS ones, Pure Storage can't do that with its FlashBlade product. It's important also to mention that NetApp prepares to add S3 on all AFF, FAS and ONTAP Select running ONTAP 9.7.

With 8.2.2, a few months ago the file size limit moved up from 4TB to 16TB and it's still the case with 9.0 but it appears to be a bit small with FlashBlade 64TB file size limit, VAST Data made tests up to 128TB for a multi EB limitation on paper and Qumulo shows 9EB on paper as well. This OneFS 16TB limit is a perfect illustration of a design made for 2000 to 2010 needs thus limiting file size for today's applications and requires juggling multi files (today but already true for a few years). To put things in perspective NFSv3 limitation is 8EB also on paper. At the same time AWS S3 supports 5TB object size with a limitation of 5GB by a single PUT.

In term of cloud, here is what top file/NAS storage vendors can support, Dell EMC being limited to GCP:

Companies	On-Prem	AWS (Marketplace presence)	GCP (Marketplace presence)	Azure (Marketplace presence)
Dell	Y	-	Y (Y)	-
NetApp	Y	Y (Y)	Y (Y)	Y (Y)
Pure Storage	Y	-	-	-
Qumulo	Y	Y (Y)	Y (Y)	Planned
Vast Data	Y	-	-	-

With this launch, the storage giant unveils DataIQ now trying to get serious with data analytics. But we're not sure, let's read these few more lines.

Hopefully it changed ClarityNow name and release something different, users expect a solution from Dell EMC by Dell EMC to deliver data insights. We're still wondering why the firm made this strange acquisition, DataFrameworks, getting acquired in the summer 2018, that didn't help sales with a non native analytics capability. If you click on ClarityNow link on the product page, you're redirected to the DataIQ page and even better click here and see where you land (DataFrameworks website redirects you to the same DataIQ page). Wow is it the same data analysis engine? Competition will love that.

Beyond DataIQ, we find also InsightIQ for performance monitoring and reporting and CloudIQ for a cloud-based infrastructure monitoring with machine learning and advanced analytics. Of course, all these three tools support now this new PowerScale product fueled by the OneFS 9 release.

One of the reference on the market is Qumulo, designing their internal file system around this analytics need and capability, confirming that adding such element later is tough and delicate.

As PowerScale introduces new hardware with OneFS 9 plus a few other things, Dell EMC needs to think about a new architecture, still scale-out, but something that goes beyond shared-nothing that finally is a limiting factor in the linear scalability of the cluster. With NVMe-oF, share-everything architectures are now possible with a complete disaggregation of the compute and the storage layers to offer independent scalability, from a capacity model with addition of storage nodes for same number of access nodes or the reverse, a performance model with addition of access nodes for same number of storage nodes. Of course dual growth is possible but it's easy and common. Shared-nothing with OneFS is 20 years old and even if Isilon did some innovative designs - they introduced for instance erasure coding for primary storage (high-end NAS in fact) many many years before people in object storage started to promote it - current technologies and developments should invite Dell EMC to reconsider some architecture choices.

Designed 20 years ago and acquired 10 years ago, Isilon is a bit of the past. It would be a good idea for the company to target companies like Vast Data which represents the future of file storage, with several key design choices based on technology trends - Flash QLC..., persistent memory (SCM), NVMe-oF (TCP for instance), shared-everything, cacheless -, but below $3 billion, no deal will happen for sure and the team is not ready to sell. In fact, it's just a question of price and keep this in mind Dell Technologies Capital is an investor in Vast Data. We'll see but definitely Dell EMC needs a huge refresh of its file storage product line. For NVMe/TCP, it already invested in an interesting company, Lightbits Labs, leader in the NVMe/TCP development, and Dell has a joint offer with them displayed here.

So now imagine an access layer with a series of 1U systems each connected via NVMe/TCP to a set of storage nodes full of Optane and QLC drives, sound familiar, it starts to be a good disaggregated model. But from that you need a new software on the access and storage nodes to leverage internal parallelism, add data reduction, bring new data placement schemes and data protection...

At the end, what is really new (for Dell EMC) here?