Home Blog Page 346

StorONE touts Optane Flash Array

By
Chris Mellor
-

StorONE, the data storage startup, has crafted a super all-flash array by twinning Optane and QLC Flash SSDs in a 2-tier, 40TB, dual Xeon SP server chassis.

The Optane Flash Array (OFA), runs StorOne’s S1 Enterprise Storage Platform software and is intended as a replacement upgrade for first generation flash arrays. 

StorONE said the OFA, with its Optane-QLC combination, should cost less than a standard all-flash array but delivers more performance at equivalent capacity levels.

The design is akin to taking Intel’s H10 Optane+QLC flash drive and implementing its basic hardware scheme at array level. We described it earlier this month and can now provide more details.

StorONE S1 Optane Flash Array.

The initial product chassis contains two storage tiers; three Optane DC P4800X 750GB SSDs for performance and five Intel D5-P4320 NVMe 7.68TB QLC SSDs for capacity. 

Blocks & Files S1 Optane Flash Array diagram

The Optane drives form a performance tier and eliminate the need for cache in memory. They also stack up random writes and send them to the QLC tier as a small number of sequential writes which are optimised for the QLC NAND page size. This prolongs SSD endurance.

The S1 software promotes data from the QLC tier to the Optane tier when it identifies a read performance advantage. 

StorONE OFA supports configurations with 3, 4, 6 or 8 Optane drives and from 4 to 16 Intel QLC SSDs. Scaling Optane drives increases performance. Scaling the QLC SSDs increases capacity.

Performance

A StorONE Intel Optane Flash Array lab report details some performance numbers.

This OFA configuration, with its eight drives, delivers over one million read IOPS. Benchmarks include:

  • 1.05 million random read IOPS, 0.14ms latency
  • 310,000 random write IOPS, 0.6ms latency
  • 10GB/sec sequential read throughput
  • 2.8GB/sec sequential write throughput

How does this compare to a non-Optane StorONE all-flash array?

In September 2018 StorONE tested its S1 system using HGST SAS SSDs inside a Western Digital 2U24 chassis and reported 1.7 million random read 4K IOPS, delivered at less than 0.3 ms latency. The system provided 15GB/sec sequential reads, and 7.5GB/sec sequential writes.

StorONE’s George Crump told us the SAS SSD S1 array, with its 24 drives, exceeded the OFA’s bandwidth and IOPS numbers because of its ability to drive more IO with drive parallelism. However, the eight-drive OFA had lower latency than the SAS system.

We think a 24-drive OFA, with eight Optane drives and 16 NVMe QLC SSDs, would deliver a larger number of IOPS and greater bandwidth than the S1 24-drive SAS system. StorONE said its system delivers linear scalability, so we this configuration could deliver 3 million IOPS and 30GB/sec sequential read throughput.

Blocks & Files expects the S1 OFA to be made available in the next few weeks. We might ponder the idea of an all-Optane OFA. This would be suited for extreme performance use cases – but be hellish expensive.

Huawei, Pure and IBM enterprise storage sales up; Dell, Hitachi, HPE and NetApp are down

By
Chris Mellor
-

The Covid-19 pandemic sent enterprise storage revenues down in 2020’s first quarter but Huawei, Pure Storage and IBM went against trend and grew revenues.

The latest edition of IDC’s Worldwide Quarterly Enterprise Storage Systems Tracker reveals total storage capacity shipped in the quarter slipped 18.1 per cent Y/Y to 92.7EB. 

Sebastian Lagana, research manager at IDC, issued a quote: “The external OEM market faced stiff headwinds during the first quarter as enterprises across the world had operations impacted by the global pandemic. ODMs once again generated growth, taking advantage of increasing spend from hyperscalers – demand that we anticipate will remain solid through the first half of 2020.”

The 92.7EB of total storage capacity comprises server SANs, ODM (Original Design Manufacturers) sales to hyperscalers like Amazon and Facebook, and sales by OEMs to the enterprise external storage market.

ODM capacity shipped went down 20.2 per cent to 54.8EB, but revenues rose 6.9 per cent to $4.9bn. Enterprise external storage did the opposite; capacity rose 3 per cent but revenues fell 8.2 per cent to $6.5bn.

The total all-flash array (AFA) market generated $2.8bn in revenue, up 0.4 per cent. The hybrid flash array (HFA -disk + SSD) market generated $2.5bn, down 11.5 per cent. An industry source tells us the disk array portion of the enterprise external storage market declined 18 per cent.

Vendor numbers

There was a wide divergence in the fortunes of individual suppliers, as an IDC table comparing the first quarters of 2020 and 2019 shows: 

Dell continues to dominate in revenue and market share, but revenues fell 8.2 per cent in line with the market. We have charted vendor revenue growth rate changes to bring out the differences:

Are these changes significant? After all there has been minor revenue growth differences for years with little substantive effects over time as a chart of IDC Storage Tracker enterprise storage vendor revenues for the past 15 quarters makes clear.

Enterprise external storage vendor revenues from IDC’s Storage Tracker since 3Q 2016

Supplier positions

Dell is still top of the tree by a wide margin. Covid-19 has sent the market down but IBM experienced saw 3.8 per cent growth, driven by a mainframe refresh cycle drawing high-end DS8800 sales in its wake.

Huawei grew revenues at the fastest rate, up 17.7 per cent. It has dipped in and out of IDC’s top five vendor list and has shown a more jumpy curve than Pure.

HPE and NetApp have swapped positions regularly as have Hitachi and IBM. That implies the latest figures do not indicate substantive changes in these vendors’ positions.

Nor did it affect Pure Storage – that much, with growth at 7.7 per cent. We see its comparatively long term growth trend relaxing a little in this latest Storage Tracker edition. Fewer customers stopped buying kit and services and the total AFA segment, Pure’s sole market, grew 0.4 per cent, while Pure grew 7.7 per cent. 

Pure told us the market in Japan shrank 4.3 per cent while Pure grew 37.5 per cent. In LATAM the market grew 19.9 per cent and Pure grew 99.9 per cent. The North America market shrank 9.2 per cent and Pure grew 11.8 per cent. 

Mission-critical computing and HCI: the time has come

By
Daniel Robinson
-

Sponsored Hyperconverged infrastructure (HCI) has been around for at least a decade, but adoption continues to grow apace, as shown by figures from research firm IDC which indicate that revenue from hyperconverged systems grew 17.2 per cent year on year for the fourth quarter of 2019, compared to 5.1 per cent for the overall server market.

Although it has become common for HCI to run general purpose workloads, some IT departments are still wary of using this architecture for the mission-critical enterprise applications on which their organisation depends for day to day business. Now, with technologies available as part of the Second Generation Intel® Xeon® Scalable processor platform, HCI can deliver the performance and reliability to operate these workloads, while providing the benefits of flexibility and scalability.

HCI is based on the concept of an appliance-like node that can serve as an infrastructure building block, enabling the operator to scale by adding more nodes or adding more disks. Each node integrates compute, storage and networking into a single enclosure, as opposed to separate components that have to be sourced and configured separately.

The real value of HCI is in the software layer, which virtualizes everything and creates a pool of software-defined storage using the collective storage resources across a cluster of nodes. This software layer facilitates centralised management and provides a high degree of automation to make HCI simpler for IT professionals to deploy and manage.

But that software-defined storage layer may be one reason why some organisations have been wary of committing to HCI for those mission-critical roles.

Doing it the traditional way: SANs

Enterprise applications, whether customer relationship management (CRM), enterprise resource planning (ERP) or applications designed for online transaction processing (OLTP), rely on a database backend for information storage and retrieval. This requirement will typically be met by a database system such as Oracle or SQL Server.

Traditionally, the database would run on a dedicated server, or a cluster of servers in order to cope with a high volume of transactions and to provide failover should one server develop a fault. Storage would be provided by a dedicated storage array, connected to the server cluster via SAN links. This architecture was designed so that the storage can deliver enough performance in terms of I/O operations per second (IOPS) to meet the requirements of the database and the applications using it.

But it means the database, and possibly the application, is effectively locked into its own infrastructure silo, managed and updated separately from the rest of the IT estate. If an organisation has multiple application silos such as this, it can easily complicate data centre management and hinder moves towards more flexible and adaptable IT infrastructure.

It also pre-dates the introduction of solid state drives (SSDs), which have a much higher I/O capacity – and much lower latency – than spinning disks. For example, a single Intel® 8TB SSD DC P4510 Series device is capable of delivering 641,800 read IOPS.

Partly, this is because of the inherent advantages of solid-state media, but also because newer SSDs use NVMe as the protocol between the drive and host. The NVMe communications protocol was created specifically for solid state media and uses the high-speed PCIe bus to deliver greater bandwidth than a legacy interface such as SAS while supporting multiple I/O queues. The NVMe protocol also ensures performance is not compromised by delays in the software stack.

Software-defined

With HCI, the database can run on a virtual machine, and the software-defined storage layer means that storage is distributed across an entire cluster of nodes. Every node in the cluster serves I/O and this means that as the number of hosts grows, so does the total I/O capacity of the infrastructure.

This distributed model also means that if a node goes down, performance and availability do not suffer too much. Most HCI platforms also now feature many of the capabilities of enterprise storage arrays as standard, such as snapshots and data deduplication, while built-in data protection features make disaster recovery efforts much easier.

With advances in technology, such as the Second Generation Intel® Xeon® Scalable processors, tends to feature more CPU cores per chip than earlier generations. This presents organisations with the opportunity to reduce the number of nodes required for a cluster to run a particular workload, and thus make cost savings.

But as the total I/O capacity depends on the number of hosts, such consolidation threatens to reduce the overall IOPS of the cluster. Fortunately, SSDs boast enough IOPS to counteract this, especially Intel® Optane™ DC SSDs, which are architected to deliver enough IOPS for the most demanding workloads. In tests conducted by Evaluator Group, a storage analyst firm, a four-node HCI cluster with Optane™ DC SSDs outperformed a six-node cluster using NAND flash SSDs under the IOmark-VM workload benchmark, with both configurations having a target of 1,000 IOmark-VMs.

Optimise the cache layer

It is common practice to implement tiered storage in HCI platforms. Inside each node in a cluster, one drive is treated as a cache device – typically an SSD – and the other drives are allocated as a capacity tier. In the past, capacity drives have been rotating hard drives, but these days the capacity tier is also likely to be SSD.

In this configuration, the cache tier effectively absorbs all the writes coming from every virtual machine running on the host system, which means it is critical to specify a device with very low latency and very high endurance for this role. In other words, you need a device that would not ‘bog down’ as those extra CPU cores are put to work.

Intel® Optane™ SSDs fit the bill here, because Intel® Optane™ is based on different technology to the NAND flash found in most other SSDs. Current products such as the Intel® Optane™ SSD DC P4800X series have a read and write latency of 10 microseconds, compared with a read/write latency of 77/18 microseconds for a typical NAND flash SSD.

In terms of endurance, Intel claims that a half terabyte flash SSD with an endurance of three Drive Writes Per Day (DWPD) over five years provides three petabytes of total writes. A 375GB Optane™ SSD has an endurance of 60 DWPD for the same period, equating to 41 petabytes of total writes, representing around a14x endurance gain over traditional NAND.

The capacity tier of the storage serves up most of the read accesses and can therefore consist of SSDs that have greater capacity but at lower cost and endurance. Intel’s second generation of 3D NAND SSDs based on QLC technology is optimised for read-intensive workloads, making them a good choice for this role.

Furthermore, IT departments can use the greater efficiency of Intel®  Optane™ SSDs to make cost savings by reducing the size of the cache tier required. Intel claims that the cache previously had to be at least 10 per cent of the size of the capacity tier. But with the performance and low latency of Intel® Optane™, 2.5 to 4 per cent is sufficient. This means a 16TB capacity tier used to require a 1.6TB SSD for caching but now customers can meet that requirement with a 375GB Intel® Optane™ SSD.

Boosting memory capacity

Another feature of Intel® Optane™ is that the technology is byte-addressable, so it can be accessed like memory instead of block storage. This means that it can expand the memory capacity of systems, boosting the performance of workloads that involve large datasets such as databases, and at a lower cost compared to DRAM.

To this end, Intel offers Optane™ DC Persistent Memory modules, which fit into the DIMM sockets in systems based on Second Generation Intel® Xeon® Scalable processors. The modules are used alongside standard DDR4 DIMMs but have higher capacities – currently up to 512GB. The latency of the modules is higher than DRAM, but a tiny fraction of the latency of flash.

These Optane™ memory modules can be used in two main ways; in App Direct Mode, they appear as an area of persistent memory alongside the DRAM and need applications to be aware there are two different types of memory. In Memory Mode, the CPU memory controller uses the DRAM to cache the Optane™ memory modules, which means it is transparent to applications as they just see a larger memory space.

In other words, App Direct Mode provides a persistent local store for placing often accessed information such as metadata, while Memory Mode simply treats Optane™ as a larger memory space.

VMware, whose platform accounts for a large share of HCI deployments, added support for Optane™ DC Persistent Memory in vSphere 6.7 Express Patch 10. In tests using Memory Mode, VMware found it could configure a node with 33 per cent more memory than using DRAM alone. With the VMmark virtual machine benchmark suite [PDF], VMware said this allowed it to achieve 25 per cent higher virtual machine density and 18 per cent higher throughput.

In conclusion, HCI might have started out as a simpler way to build infrastructure to support virtual machines, but advances in technology now mean it is able to operate even mission critical workloads. With Second Generation Intel® Xeon® Scalable processors and Intel® Optane™ DC SSDs, HCI can deliver the I/O, low latency and reliability needed to support enterprise applications and their database back-ends.

It can also potentially deliver cost savings, as the greater efficiency of Intel® Optane™ storage means that fewer drives or nodes may be required to meet the necessary level of performance.

Sponsored by Intel®

NetApp loves Iguazio’s AI pipeline software

By
Chris Mellor
-

Each month, NetApp’s Active IQ handles up to 10 trillion data points, fed by storage arrays deployed at customer sites. Data volumes are growing.

Active IQ uses various AI and machine learning techniques to analyse the data and sends predictive maintenance messages to the arrays. The service has a lot of hardware capacity at its disposal, including ONTAP AI systems, twin Nvidia GPU servers, fast Mellanox switches and ONTAP all-flash arrays. But what about orchestrating and manage the AI data pipeline in those arrays? This is where Iguazio comes in.

NetApp is using Iguazio software for end-to-end machine learning pipeline automation. According to Iquazio, this enables real-time MLOps (Machine Learning Operations), using incoming data streams. 

The Iguazio-based Active IQ system has led to 16x storage capacity reduction, 50 per cent reduction in operating costs, and fewer compute nodes, NetApp says. Also new AI services for Active IQ are developed at least six times faster.

Unsurprisingly, NetApp people are enthusiastic. Shankar Pasupathy, NetApp chief architect for Active IQ, said in a prepped quote: “Iguazio reduces the complexities of MLOps at scale and provides us with an end-to-end solution for the entire data science lifecycle with enterprise support, which is exactly what we were after.”

NetApp has now partnered with Iguazio to sell their joint data science ONTAP AI solution to enterprises worldwide. Iguazio also has a co-sell deal with Microsoft for its software running with Azure, and a reference architecture with its software on Dell EMC hardware. It is carving out a leading role as an enterprise AI pipeline management software supplier.

Let’s take a closer look at the Active IQ setup.

Hadoop out of the loop

NetApp needs real-time speed, scalability to cope with the massive and growing streams of data, and the ability to run Active IQ on-premises and in the cloud. It also wants Active IQ to learn more about customer array operations and get better at predictive analytics.

NetApp’s initial choice for handling and analysing the incoming Active IQ data was a data warehouse and Hadoop data lake. But the technology was too slow, too complex and scaling was difficult.

Active IQ dashboard.

The fundamental issue is that real-time processing involves multiple stages with many interim data sets and types, and various kinds of processing entities such as containers and serverless functions.

This complexity means careful data handling is required. Get it wrong and IO requests multiply and multiply some more, overwhelming the fastest storage media.

AI pipeline

Iguazio’s software speciality is organising massive amounts of data and metadata in such a way as to make applying AI techniques in real time possible. Its software provides a data abstraction layer on which these processing entities can run.

An AI pipeline involves many stages:

  • Raw data ingest (streams),
  • Pre-processing (decompression, filtering and normalization)
  • Transformation (aggregation and dimension reduction) 
  • Analysis (summary statistics and clustering) 
  • Modeling (training, parameter estimation and simulation) 
  • Validation (hypothesis testing and model error detection) 
  • Decision making (forecasting and decision trees) 

There are multiple cloud-native applications, stateful and stateless services, multiple data stores, and data selection and filtering into subsequent stores involved in this multi-step pipeline. 

NetApp and Iguazio software

Iguazio’s unified data model is aware of the pipeline stages and the need to process metadata to speed data flow. The software runs in a cluster of servers, with DRAM providing an in-memory metadata database, and NVMe drives holding interim data sets.

Iguazio data model concept.

NetApp uses Trident, a dynamic storage orchestrator for container images integrated with Docker and Kubernetes and deployed using NetApp storage. Iguazio integrates with the Trident technology, linking a Kubernetes cluster and serverless functions to NetApp’s NFS and Cloud Volumes Storage. Iguazio is compatible with the KubeFlow 1.0 machine learning software that NetApp uses.

Iguazio detects patterns in the data streams and relates them to specific workloads on specific array configurations. It identifies actual and potential anomalies, such as a performance slowdown, pending capacity shortage, or hardware defect.

Then it generates actions, such as sending alert messages, to the originating customer system and all customers with systems likely to experience similar anomalies. It does this in real time, enabling NetApp systems to take automated action or sysadmins to take manual action.

Zerto provides disaster recovery for containerised apps

By
Chris Mellor
-

Zerto is branching out from disaster recovery of virtual machines to offer general backup services. The company will also cover cloud-native applications with its continuous data protection (CDP) and journalling technology.

The company announced the plans, along with a roadmap for its IT Resilience Platform, today at the ZertoCON virtual customer conference.

Specifically, Zerto has announced Zerto for Kubernetes (Z4K – our acronym) to protect applications running on Amazon Elastic Kubernetes Service (Amazon EKS), Google Kubernetes Engine (GKE), Microsoft Azure Kubernetes Service (AKS), Red Hat OpenShift, and VMware Tanzu.

Gil Levonai.

Zerto’s Gil Levonai, CMO and SVP of product, said in prepared remarks: “With the clear shift towards containers based application development in the market, we are looking to extend our platform to offer these applications the same level of resilience we have delivered to VM-based applications.

“While next-gen applications are built with a lot of internal availability and resilience concepts, they still require an easy and simple way to recover from human error or malicious attacks, or to be mobilised and recovered quickly without interruption. This is where Zerto can help.”

Zerto for Kubernetes

Z4K protects persistent data and can protect, move and recover containerised applications as one consistent entity, including associated Kubernetes objects and metadata. It features continuous journaling for Kubernetes apps, including Persistent Volumes, StatefulSets, Deployments, Services, and ConfigMaps. This journal can provide thousands of recovery checkpoints.

It has always-on replication to provide protection and recovery of Kubernetes persistent volumes within and between clusters, data centres or clouds.

Entire applications with their component containers can be recovered as entities in an ordered way. Z4K can instantiate a full running copy of an entire application in minutes from any point in time for recovery from data corruption or ransomware without impacting production or for testing.

There are automated Z4K workflows for failover, failover test, restore, rollback restore, and commit restore. The Z4K software is managed through a kubectl plug-in and native Kubernetes tooling.

Competitive positioning

Z4K is not provisioning storage to containers – unlike Kasten and Portworx which also offer  containerised application protection. Deepak Verma, Director, Product Strategy at Zerto, told Blocks & Files it runs as a native K8s application which provides CDP-based journaling for persistent volumes running on the cluster nodes plus the backup, restore, and DR orchestration necessary to execute any desired use cases for resilience of K8s applications.

Verma said: “Kasten, judging from public documentation, appears to be relying the snapshot mechanism of the cloud platforms to capture data, with a minimum timeframe of five mins. Zerto for K8s on the other hand is providing CDP for persistent volumes at a 5-10 second interval and capturing all the necessary attributes of a K8s application in a consistency group to recreate locally or remotely down to a very granular point in time.”

In summary, Z4K has “less complexity and vendor lock-in than Portworx and more granular RPOs and having point-in-time consistency across multiple containers compared with Kasten”.

CPU load

Blocks & Files pictures a Kubernetes system executing thousands of containers over a few hours. Z4K is app-aware and so the host server has to run these containers and also do the granular continuous journaling for Z4K, with its thousands of checkpoints. We asked Zerto what is the burden on the host server processors of this additional load?

Verma said: “Zerto for K8s is built upon the same core journaling and CDP technology and intellectual property that Zerto has successfully deployed and improved for the last 10 years in very large VM environments.

“The relative overhead for most environments has been less than 10 per cent CPU and memory at the host level during busy times of high change rate. The predictable scaling of which is very well understood as well. Since we are still at the alpha stage, we have not run extensive performance tests, but do not expect the technology to be much different than what we currently utilise as guidelines.

“Part of the success of Zerto has been its very efficient CDP mechanism. For production application we view this as a minor overhead to provide the level of RPOs and RTOs that customers have come to expect from Zerto.”

Z4K will be available in an early adopter program later in 2020 and goes on general release next year.

Roadmap

Zerto said it is decoupling operational recovery from backup because continuous journaling eliminates the need for snapshot-based backup. 

Levonai said: “Historically, top-tier, customer-facing applications would be protected with multiple data protection and disaster recovery solutions while lower-tier applications would be protected with high RPO backups only, or not protected at all. Zerto is levelling the playing field by applying its CDP technology to each of these … applications, transforming the backup market.”

The company plans to extend its IT Resilience Platform with a mix of one-to-many replication, local and remote journaling, long-term repositories, and short-term and long-term retention policies. The aim is meet various SLA needs at various cost levels.

The company will offer in-cloud protection and DR for AWS, Azure and GCP. There will be tiering to cloud object stores, with built-in data reduction technology, for long-term retention.

Zerto will develop new workflows to simplify file or VM restores back to production from a local or remote journal. Additional roadmap features include added VM auto-protection, encryption and security, automatic VRA lifecycle management for maintenance, features for managed service providers, new analytics functionality and an improved licensing mechanism.

Pure Storage delivers Purity six appeal

By
Chris Mellor
-

Pure Storage today launched the sixth generation of the Purity FlashArray OS. The upgrade includes extended disaster recovery, integrated file support and NVMe-over-Fabrics with RoCE for Cisco UCS and VMware.

Purity 6.0 adds a DirectMemory Cache, multi-factor authentication, increased volume management and simpler quoting for Pure1 management software. A Cloud Block Store is in beta test on Azure.

James Kelly, senior systems administrator at Chapman University, a private university in Orange Country, California, said in a prepped quote: “The unified SAN and NAS capabilities of this new FlashArray OS represent a game-changer for our highest-performance file-based workloads that otherwise need to run in all-block environments. It offers us a great way to cost-effectively run VDI or performance-critical file-based applications right alongside our key enterprise and research workloads.” 

Purity 5.0, announced in June 2017, delivered initial NFS and SMB file services support and synchronous replication. There was also a demonstration of end-to-end NVMe over fabrics to Cisco UCS servers using a 40Gbit/s RoCE v2 Ethernet link at the v5.0 launch.

Active DR

Pure is offering “active disaster recovery” built on new continuous replication technology. It uses active-passive replication for geo distances, and provides a near-zero RPO (three to four seconds). It is bi-directional, there are no journals to manage and Pure says it provides a fast recover/failover time. Customers have test failover, failover, reverse and failback functionality.

Pure offers a wide range of replication options according to availability requirements and price points: synchronous, active-active, replication with ActiveCluster, snapshot-based asynchronous replication, and now continuous replication.

There are validated designs for VMware Site Recovery Manager and applications such as Microsoft SQL, Oracle, SAP and MongoDB. 

Files

In April 2019, Pure Storage acquired Compuverde, a Swedish storage software developer. Purity 6.0 incorporates Compuverde’s file access technology, which sits alongside the existing block protocol above the data reduction and encryption layers in the Purity software stack. This makes FlashArray a unified file and block storage platform.

Block and file data benefit from global deduplication and compression of the shared storage pool in FlashArray. 

iPhone screen grab of Pure slide.

Alex McMullan, Pure’s  VP & CTO, International, told us that the current FlashBlade file support is aimed at high-performance file applications such as big data and machine learning environments. FlashArray files is suited to user-stye file services that don’t need the scale-out capability of FlashBlade.

He said FlashArray files will get near-synchronous replication. Unstructured data on Purity’s new file services can be protected with Veeam and CommVault backup offerings. 

Other features

Pure announced the third generation FlashArray//X R3, which can be fitted with an Optane cache, in February. Purity 6’s DirectMemory Cache tries to satisfy read requests from this cache and delivers 50 per cent improvement in read latency – down to 125µs. Customers can add Optane capacity in 3TB or 6TB packs of 750GB DirectMemory Modules.

The NVMe-RoCE (RDMA over lossless Converged Ethernet) has a validated design for Cisco UCS servers. Pure will offer NVMe/TCP support but does not say when.

Purity 6.0 hits the streets on June 18. Evergreen Storage subscription customers receive all Purity 6.0 features, with no additional licenses or added support costs. Cloud Block Store for Azure should be delivered later this year.

Snowflake ‘preps $20bn IPO’

By
Chris Mellor
-

Snowflake is prepping an IPO this year that could value the data warehousing startup at up to $20bn.

The company has already submitted a confidential IPO filing with the US SEC, according to the Financial Times, citing unnamed sources.

Snowflake has raised $1.4bn in VC funding, including a $479m G-series round earlier this year which priced the company at $12.4bn. Salesforce was co-lead investor.

Snowflake CEO Frank Slootman said at the time that the company was poised to turn cashflow positive this year, revenue grew by 174 per cent last year, and would soon top $1bn. Snowflake claims more than 2,000 customers.

He told the San Francisco Business Times that $12.4bn valuation could be higher following an IPO: “The reason is that our growth trajectory is so fierce and our addressable market is so large. When companies grow so fast, as Snowflake has, the valuation may seem like a big number now but not later. When I was with ServiceNow (as CEO), the valuation was $2.5bn when we went out and now it has a $65bn valuation.”

Slootman replaced CEO Bob Muglia in May 2019. At that time Blocks & Files suggested Slootman’s track record of acquisition and successful IPO-based growth must have been attractive to Snowflake’s investors looking for a great exit.

HPE updates Primera and Nimble arrays

By
Chris Mellor
-

HPE has refreshed the Primera upper mid-range array and Nimble lower-mid-range arrays. Let’s take a closer look.

The Primera array gets NVMe support via Primera OS v4.2. HPE said all-flash Primera already delivers 75 per cent of I/O within 250 μs latency. Now Primera with all-NVMe supports twice the number of SAP HANA nodes at half the price. HPE has not provided an IO latency number for an all-NVMe Primera array.

Primera nodes and C=controllers

Primera has been given an AI sub-system so it can self-optimise array resource utilisation in real time.

A Primera array, with the Peer Persistence feature, can now replicate to three global sites for extra and transparent data access reliability if metro-scale disasters occur. The array also gets near-instant asynchronous replication over extended distances. This has a one minute recovery point objective.

A replication target site can be in the public cloud and replication can provide data for test and development and analytics.

Primera automation has been optimised for providing storage to VMware VVOLs and containers. The container support comes via a Kubernetes CSI plug-in and the VVOL support includes disaster recovery via Site Recovery Manager. Nimble arrays already support VVOLs and a CSI plug-in.

Nimble Storage

Nimble features include six nines availability and two-site replication. They also have asynchronous replication on-premises or to the cloud for extended distances. The arrays get global three-site replication via NimbleOS 5.2.

Storage class memory (SCM) is now supported and cuts response time in half for the all-flash Nimble array, according to HPE. It says SCM is too expensive to use as a storage tier for Nimble customers. A relatively small amount of SCM is used instead as a cache to speed IO. It provides an average sub-250 μs latency3, at near the price of an all-flash array.

InfoSight system management has been extended for Nimble arrays with cross-stack analytics for Hyper-V. These identify abnormal performance issues between storage and Virtual Machines, and under-utilised virtual resources. 

Availability 

Primera OS 4.2, is available worldwide in Q3 2020 at no additional charge for customers with valid support contracts. Primera All-NVMe is available for order direct and through channel partners. The Primera  CSI Driver for Kubernetes 1.1.1 is now available for Primera. 

HPE Nimble Storage 1.5TB Storage Class Memory Adapter Kits are available now for Nimble Storage AF60 and AF80 All Flash Arrays. Nimble 3 Site Replication is available as part of NimbleOS 5.2 release for any customer with an active support contract. InfoSight Cross Stack Analytics for Hyper-V is available next month for any Nimble customer with an active support contract. 

HPE launched the Primera arrays a year ago as a massively parallel product line, intended to replace the 3PAR array. They have a 100 per cent availability guarantee. Nimble and its array technology was acquired by HPE in March 2017. Both arrays are available as a service through HPE’s GreenLake.

InfiniteIO extends file tiering to Azure

By
Chris Mellor
-

InfiniteIO has added an Azure Blob backend, to make its file metadata accelerator more suited to Microsoft users.

There are now four public cloud tiering targets: Amazon Glacier, Azure Blob, Google Cloud Platform and IBM’s Cloud Object Store.

Mark Cree, InfiniteIO CEO, said in prepared remarks: “Enterprises can slash their storage spend today with InfiniteIO and Microsoft Azure technology without sacrificing speed or performance for their private and public cloud initiatives.”

InfiniteIO processes NFS or SMB filesystem access metadata in a dedicated system. A typical NFS request could involve seven metadata requests across the NFS network with a single data transfer. InfiniteIO handles the metadata requests locally with its own Application Accelerator and so speeds up network file access.

The software moves files to cloud backend storage and remaps file requests seamlessly from on-premises file access users or applications. 

On-premises S3 object storage targets are accessed via InfiniteIO’s ‘Hybrid Cloud Tiering’ .

InfiniteIO Hybrid Cloud Tiering

The Infinite Insight app scans files in on-premises NAS arrays and identifies those that have not been accessed recently. These files are shunted to object stores on-premises and in the public cloud, to release primary filer store capacity.

Read more about the InfiniteIO technology and Hybrid Cloud Tiering in a white paper (registration required).

The mainstream all-flash array is an enterprise dead-end, these three startups say

By
Chris Mellor
-

Three startups aim to revolutionise and rescue the on-premises storage array from the twin assaults of hyperconverged infrastructure (HCI) and the public clouds.

They are Infinidat, StorONE and VAST Data and each company is gaining enterprise customers who agree that their technology supersedes all-flash arrays in marrying capacity scaling and performance.

All three claim the classic all-flash array, even when extended with NVMe-over-Fabrics speed, is a dead-end. Ground-up re-written storage software is needed to reinvigorate shared storage array technology.

Each company is a technology outlier and has responded differently to the twin needs of storing large and growing volumes of primary data cost-effectively while providing fast, flash-level access.

Yanai Hallek Naor
From left, Moshe Yanai (Infinidat), Rennen Hallek (VAST Data) and Gal Naor (StorONE)

Infinidat

According to Infinidat, all-flash arrays are prohibitively expensive at petabyte scale and its DRAM caching makes its disk-based arrays faster than all-flash arrays, even when they are boosted with Optane 3D XPoint drives.

Infinidat eschews a flash capacity store, relying instead on nearline disk drives with data prefetched into a memory cache using a Neural Cache engine with predictive algorithms. A thin SSD layer acts as an intervening store between disk and the DRAM. There are three controllers per array, which enables parallel access to the data.

Infinidat’s monolithic InfiniBox array provides block and file access to the data and there is technical background information in a Register article. The Neural Cache engine monitors which data blocks have been accessed and prefetches adjacent blocks into DRAM.

According to a Silverton Consulting report, the cache metadata uses LUN block addresses as an index and “can maintain information on billions of objects, accessing and updating that metadata in microseconds”. It enables more than 90 per cent of reads to be satisfied from memory.

Infinidat said it had deployed more than 6EB of storage to its customers in February – an exabyte greater than six months previously. It also said this 6EB was more than the top eight all-flash array (AFA) vendors had shipped in 2019.

Competitors include Dell EMC’s PowerMax, IBM with its DS8800 product, and Hitachi Vantara’s VSP. Blocks & Files expects Infinidat growth to continue.

StorONE

StorOne, Israeli startup came out of stealth in 2017, after spending six years developing its own storage software stack.

The basic idea about marrying capacity and performance is expressed through the company’s TRU (total resource utilisation) software. This can deliver the rated throughput of a storage drive, whether it is a disk drive, a SAS or SATA SSD or an NVMe SSD. StorONE said its storage software stores data as blocks which could be accessed either as blocks, files or objects in a universal store. 

StorONE’s basic pitch is that its software make all types of storage hardware go faster. For example, StorONE delivered 1.7 million IOPS in a 2-node ESXi server system using 24 x 3.84TB WD SSDS in a JBOD as the data store. The company claims a typical all-flash array would need four times as much controller and storage hardware to deliver that performance.

More Optane Flash Array background can be gleaned from a video on Vimeo.

StorONE OFA video.

At launch, the TRU system scaled up to 18GB/sec of throughput and 15PB of capacity, making this enterprise-class in scalability. StorONE is now developing S1: Optane Flash Arrays, using Optane SSDs as a performance tier and Intel QLC flash as the capacity store. It says the combination with its own S1:Tier software “delivers over one million read IOPS, and over 1 million write IOPS”. 

StorONE has not revealed much customer information, at time of writing. However, the company has made its pricing transparent and markets its technology as significantly cheaper than the competition. CEO and founder Gal Naor wrote in April 2020: “It is not unusual for us to provide a five-year total cost of ownership price to a customer that is less than what they are paying for 18 months of maintenance.”

The company is also developing an all-Optane array for extreme performance workloads. This is described in an unpublished StorONE Lab report which we have seen. In testing a three-drive system the S1 Enterprise Storage Platform delivered between 85 per cent to 100 per cent of the raw performance of the physical Optane drives:

  • 1.15 million random read IOPS at 0.035ms latency
  • 480,000 random read IOPS at 0.55ms
  • 6.8GB/sec sequential read bandwidth at 0.3ms
  • 3.1GB/sec sequential write bandwidth at 0.5ms

Our sense is that StorONE is making steady progress, while trying to demonstrate it can accelerate any storage hardware configuration without focusing on any particular market.

VAST Data

VAST Data already had some customers for its enterprise storage arrays when it came out of stealth in February 2019. Initially, the arrays used TLC (3bits/cell) flash and then transitioned to denser and lower cost QLC NAND.

VAST Data’s basic pitch revolves around extra efficient data reduction and erasure coding. The way it reduces the number of times it writes to its flash drives means the array cost and endurance are better than ordinary all-flash arrays. Customers can store all their data on flash, access it via file and object protocols, and get flash speed at disk array prices, the company claims.

Customer progress has been rapid, according to Jeff Denworth, co-founder and VP for products. He told a press briefing this week that year one sales, starting February 2019, eclipsed first year sales of Data Domain, Isilon, Pure Storage and Nutanix. The average customer spend was $1.02m and VAST claimed dozens of customers across four continents. Denworth said the company has $140m cash in the bank and sees a “clear path to breakeven”.

VAST focuses on the enterprise flash capacity market, which means it spreads less thinly across its potential customer base than StorONE. It has also produced customer references, unlike StorONE.

VAST Data customer quotes from briefing deck

Comment

Denworth told the press briefing that the NVMe-over-Fabrics array startups had failed to live up to expectations. But how will VAST, Infinidat and StorONE prosper?

Blocks & Files notes that no mainstream storage array vendor has so far made moves to extend or adapt their technology in response to Infinidat, StorONE or VAST Data. That is unsurprising, as each startup has developed its software stack from the ground up. Adding similar caching or erasure coding or data reduction technologies by the mainstreamers with their different software stacks would be difficult.

That implies that, if one or more of the three are successful enough, then an acquisition might be on the cards.

AIMES ‘normalises’ healthcare AI with Pivot 3- controlled GPUs

By
Chris Mellor
-

Specialist GPU servers could be overkill when crunching relatively small amounts of data for analytics purposes. So why not use ordinary servers boosted with GPU accelerator cards and deployed via HCI instead? This is what AIMES, a specialist tech provider for the UK’s National Health Service, does with its Health Cloud service.

AIMES is using GPU-enhanced hyperconverged infrastructure appliances (HCIAs) from Pivot3 to provide AI-based applications to NHS clinicians and researchers.

NHS trusts

The NHS is organised into regional NHS Trusts. These are interconnected by the closed, secure and private broadband Health and Social Care Network (HSCN) which safeguards the exchange of sensitive NHS data between trusts and their managed service providers.

An MSP can provide shared IT facilities which regional trusts can use, instead of operating their own duplicated IT department and facilities. One such MSP is AIMES (Advanced Internet Methodologies and Emerging Systems), which is based in Liverpool and originated as a spinout from Liverpool university.

Glenn Roberts, business development director at AIMES, tells us NHS trusts typically do not want to use the big US public cloud providers because of data sovereignty and security issues. They also want to expand the clinical side of their hospitals and not the admin side, which includes local IT services.

My AIMES is true

AIMES provides managed tier 3 data centre services to commercial customers over the general internet, and also to NHS trusts using the NHS HSCN networking facility. One focus area is to provide Trusted Research Environments (TREs) to its NHS customers inside a Health Cloud.

These customers such as clinicians, could provision and decommission research environments with the same speed and flexibility as the public cloud while providing researchers with secure and high-availability virtual access complying with NHS data privacy regulations.

Pivot3 precursors

AIMES is Pivot3’s customer for its GPU-enhanced systems and makes Pivot3 capabilities available through its Health Cloud infrastructure, which it set up in 2015.

Glenn Roberts

Roberts told us AIMES initially provided Health Cloud TREs using a set of servers accessing a Dell EMC EqualLogic (PS) SAN, and then a Compellent (SC) SAN but experienced problems with reliability and upgrades. (The SC SAN is still in use with general virtual servers.) AIMES also experimented with a SAN in the cloud, an Azure pod and Hyper-V, but that did not fulfil its needs either.

Pivot3

After checking out Dell EMC’s VxRail and HPE’s HCI offerings, AIMES switched to Pivot3 HCI systems, In 2018, as a way of fixing these issues. Roberts said: “Pivot3 ticked all our boxes. It’s rare that a technology does exactly what it says on the tin.”

AIMES has five clustered Pivot3 Acuity nodes, two with NVMe SSDs to provide better general performance, and two Lenovo-based systems with GPU cards to deliver faster analytics capability. The non-GPU nodes are based on Dell EMC PowerEdge servers.

Roberts is particularly pleased with Pivot3’s scalability and the quick and easy way its Acuity software recognises and adds nodes such as the two GPU systems to the cluster. That reduces AIMES’s own admin effort.

HCI AI

Roberts told us the GPU nodes help with running predictive AI applications looking at the likelihood of depressive episodes in mental health patients. The AI model code monitors clinical data and sends alerts to clinicians if certain early marker patterns are detected in a patient’s data.

He also told us: “We’re doing AI in a project with Barts [Health NHS Trust] for cardiac imaging.” A patient has a heart MRI scan and the AI code looks at a series of right ventricle contraction images from the scan. It discerns a measure of heart health from them that is up to 40 per cent more accurate than a human assessment.

This use of AI running on Pivot3 HCI systems represents a “normalisation of AI” where ordinary servers are fitted with GPU cards to run accelerated AI code. Roberts said this is big data-stye AI analytics with relatively small amounts of data and it’s not necessary to use separate GPU systems such as Nvidia’s DGX-type GPU servers.

Commvault gives way to Starboard, shakes up board

By
Chris Mellor
-

Commvault has yielded to Starboard Value by appointing three new members to the board. The data management vendor has also set up a new operating committee.

Starboard Value announced in March that it held 9.3 per cent of Commvault stock. In April it told Commvault it wanted six proxy directors on they board. Commvault accepted one of Starboard’s picks – Todd Bradley, CEO of Mozido. Presumably, they both agreed about the other new members; Alison Pickens, formerly COO of Gainsight, and Arlen Shenken, Citrix CFO.

The new directors replace three long-standing members of the 11-strong board; co-founder Al Bunte, Frank Fanzilli and Daniel Pulver.

Jeff Smith, Starboard CEO, said in a canned statement: “We appreciate the collaborative dialogue we have had with Commvault’s Board and leadership team. Commvault is an outstanding company. We believe the expertise provided by these new directors and the focus of the Board’s Operating Committee will help improve Commvault’s profitable growth, return on investment, and enhance value creation.”

Commvault chairman Nick Adamo also issued a quote: “We are pleased to add three highly qualified directors who bring both targeted experience and diversity to the Board.”

Operating committee

Bradley, Schenkman and existing director Charles Moran sit on a new operating committee that will oversee Commvault’s budgeting process and work with management to establish margin targets and a balanced capital allocation policy no later than December 31, 2020.

Adamo said: “We … expect the new Operating Committee to build on the progress Sanjay Mirchandani has made since being appointed CEO, including important changes to our operating priorities, improvements to our go-to-market strategy and investments in Commvault’s technology differentiation.”

Commvault is migrating from its traditional on-premises soultion sales to a subscription based cloud-centric model. But progress has been slow. In response to faltering revenue growth it appointed a new CEO, Sanjay Mirchandani, in February last year. Three financial quarters later and it had not returned to growth, hence Starboard’s interest in shaking up the company.

1...345346347...438Page 346 of 438

ABOUT US

Blocks & Files is a storage news, information and analysis site covering storage media, devices from drives through arrays to server-based storage, cloud storage, networking and protocols, data management, suppliers and standards. We publish news, views, opinions and analysis of storage industry events and developments, aiming to keep storage users, developers, consultants, marketeers, salespeople, support staff, analysts and everyone else involved in IT storage up to date with what’s going on.

FOLLOW US

© Situation Publishing, 2018-2024