Home Blog Page 177

Google’s Cloud Hyperdisk to bring SAN features to GCP

By
Chris Mellor
-

Google is introducing a SAN-like Cloud Hyperdisk persistent storage instance to its Cloud Platform, with separately tunable capacity and compute so compute-intensive VMs don’t have to be provisioned to get storage IOPS and throughput.

It says SAP HANA, SQL Server, data lake, modeling and simulation applications need a single storage entity providing object storage flexibility with file manageability and block access performance. The coming Google Cloud Hyperdisk block storage offering meets these requirements and will offer 80 percent higher IOPS per vCPU for high-end database management system (DBMS) workloads when compared to other hyperscalers.

Hyperdisk was previewed in September by Google, and is a next generation Persistent Disk type for use with Google Compute Engine (GCE) and Google Kubernetes Engine (GKE). In the preview Hyperdisk was mentioned along with the C3 machine series using Gen 4 Xeon SP processors and an Intel IPU customised for Google. The C3 virtual machines VMs with Hyperdisk will deliver four times more throughput and deliver a tenfold improvement in IOPS compared to the previous C2 instances.

Ruwen Hess, Product Management, Storage, at Google, presented Cloud Hyperdisk at an October 2022 Cloud Field Day event, and said: “We really thought about what does block storage for the cloud really, ideally, look like?”

“The feedback coming from customers was a lot around having it dynamically provisioned both performance wise and capacity wise, decoupled from instance, type and size. And with the ability to be dynamically adjusted over time, covering the full spectrum, all core workloads, and with each workload being able to change scale and other parameters, without suddenly getting efficiency or TCO checks.”

Because of this dynamic provisioning and instance decoupling, Google says, users don’t have to choose, aka over-provision, large compute instances to get the storage performance needed for Hadoop and Microsoft SQL Server data workloads.

There are three Hyperdisk types: Throughput, Balanced and Extreme, as a slide showed:

Smartphone camera shot of Youtube video showing Ruwen Hess presentation

The Throughput variant is designed to deliver up to 3GBps bandwidth for scale-out analytics workloads and be cost-efficient. The Extreme offering provides more than 4.8GBps of throughput with 300,000+ IOPS and will have performance and latency SLAs. It is for high-end SQL Server, Oracle and SAP HANA-class workloads, as Hess said: “Where you just need the utmost performance and less price sensitivity, more of a focus on the performance facilities.”

The Hyperdisk Balanced scheme fits between these bandwidth and IOPS-optimized variants, with around 150,000 IOPS and 2.4 GB/sec throughput. Hess said: “We expect the the bulk of workloads will find that Hyperdisk Balanced is well covered in terms of the price/performance profiles, and so forth.”

Hyperdisk pooling

In time Hyperdisk pools will be provided, one per Hyperdisk variant, with the ability for management at scale, thin-provisioning, data reduction and policy-based management. That means customers won’t have to directly manage every Hyperdisk.

Hess said: “Hyperdisk will exist both as fully provisioned as persistent disk as today. And then later in 2023, we will also offer storage pools where customers can optionally put these volumes into the pools, where the pools will benefit from thin provisioning, data reduction, and management by policies, where you have possibly hundreds of instances with hundreds of 1,000s of disks.” Customers will have both options.

He said: “A storage admin will manage the storage pool and much like the managing a SAN today, where you have total capacity, you have buffer capacity, you have thin provisioning, [and] over time, we will also expose, not only capacity pooling, but also performance pooling. So, if you have a lot of instances, and most of them are idle, but sometimes they spike a bit, you’ll be able to stick them into a pool and benefit further down the road.”

Provisioning example

Hess provided a comparison between provisioning GCP persistent storage today and how it will look in the future with Hyperdisk:

He outlined 6 steps needed to provision persistent storage for SQL Server today, wth manual resetting when changes occur, and contrasted that with 3 steps needed when Hyperdisk arrives. Today’s SQL Server set up requires, for example, a 64 vCPU instance whereas Hyperdisk will only need a 20 vCPU instance because it has decoupled storage performance from the VM instance type.

In more detail he said today’s scheme needed an 64-core N2 VM and 3.3TB of PD-SSD storage delivering 1.2GBps to provide 100,000 IOPS and 2TiB of capacity. Hess said: “You’re buying 3.3 terabytes to get to the 100,000 IOPS.” Switch over to Hyperdisk when it arrives and that changes to a 32-core N2 VM with 2TB of Hyperdisk Balanced capacity.

Because Hyperdisk can be dynamically tuned then it can be set up to provide 100,000 IOPS and 1.8GB/sec in office hours and 50,000 IOPS and 1GBps outside that period.

Hess said Hyperdisk will be developed so that: “over time, extending it into into workload-aware storage, where customers can set business level metrics that the system understands the workload level, say for SAP HANA rehydration plans and things like that.” 

He specifically said more user-visible metrics were coming: “We are actually working on exposing more metrics, especially around latency, for block storage in general.”

Hyperdisk is not yet listed in Google Cloud’s Persistent Disk webpage, and should be available early next year.

Google Cloud’s Persistent Disk webpage.

Azure has an Elasticated SAN offering for large scale, IO-intensive workloads and top tier databases such as SQL Server and MariaDB. Amazon has a Virtual Storage Array (VSA) construct in its cloud in which compute is presented as VSA controllers and capacity expressed as volumes which can use NVMe drives. As a source on the industry said: “It’s a tacit admission by the hyperscalers that there is a gap in their storage portfolios in this space; specifically, high-performance sharable block.”

Solidigm roadmap promises 61TB SSD

By
Chris Mellor
-

Solidigm presented a detailed look of its near-term SSD roadmap at the Tech Field Day 2022 event, and tucked away in the portfolio is a 61TB drive.

There are actually two drives on the way, both built with Solidigm’s fourth-generation 192-layer QLC (4bits/cell) 3D NAND, and called Essential Endurance and Value Endurance. The Essential Endurance variant has a 4KB block size and 3.84, 7.68, 15.36, and 30.72TB capacities, and a maximum 32PB written (PBW) endurance. The Value Endurance has a 16KB block size and capacities of 7.68, 15.36, 30.72, and 61.44TB with a doubled maximum endurance of about circa 65PBW.

Solidigm slide
Solidigm slides from TFD 2022

Solidigm said TLC (3bits/cell) NAND is used in about 80 percent of flash capacity shipping today, with QLC a probable replacement. The QLC Essential Endurance drive is positioned as a TLC (3bits/cell) replacement with lower 4K random write performance, whereas the QLC Value Endurance drive is a lower cost TLC replacement for read-heavy workloads. It is less ideal for smaller mixed read/write workloads.

The maximum capacities of 30.72 and 61.44TB are greater than the current highest disk drive capacity available today: 26TB with Western Digital’s shingled drive and 22TB with its conventional HDD.

The endurance numbers are 65PBW – which comes down to 35.6TB/day over five years. Solidigm slides said most drives used only 15 percent at most of their rated endurance:

Solidigm slide

TFD host Stephen Foskett tweeted: “This QLC 30TB drive could sustain 70 years of sequential video writes, as discussed by @Solidigm at #SFD24.”

The coming drive’s performance was given as up to 113,000 random write IOPS and 474 MB/sec throughput, but we don’t know what type of interconnect was used. A Solidigm slide said this was 6x faster than a competing but unidentified QLC SSD and 24 percent better than a competing TLC SSD. The latency number of 0.563ms is also 84 percent better than the competing QLC drive. 

However, compared to Solidigm’s existing D5-P5316 QLC SSD, with its 30.72TB maximum capacity, the performance is lackluster. The P5316 offers up to 800,000 random read IOPS, 7GB/sec sequential read and 3.6GB/sec sequential write bandwidth across a PCIe gen 4×4 NVMe interconnect.

Solidigm said these new drives would grow its market as a TLC SSD and disk drive replacement:

Solidigm slide

They will come in U.2 (2.5-inch), E1.S, E1.L and E3.S formats and we expect them to arrive in the first half of 2023.

Storage news roundup – November 3

By
Chris Mellor
-
Storage news
Storage news
Christian van den Branden

Data protector Arcserve has appointed Christian van den Branden as CTO, leading the global research and development group, advancing Arcserve’s technology innovation, and driving the strategic direction of the product roadmap. Most recently, van den Branden served as CTO at Qualifacts, a healthcare technology company. Prior to this, he held leadership roles including head of engineering at ZeroNorth, XebiaLabs, SimpliVity, DynamicOps, and EMC Corporation. Prior Arcserve CTO Ivan Pittaluga is still listed on LinkedIn in that role.

Mike Jones.

…

Flash controller firmware startup Burlywood announced on November 1 that Mike Jones has become its CEO, with founder and original CEO Tod Earhart stepping aside to be CTO – a classic tech startup founding CEO move. Funny that: we had him in place in August as a Burlywood press release confirms.

…

Data Protector Commvault has announced the Metallic File & Object Archive. It’s Data Management as a Service (DMaaS) offering designed to help organizations deal with Governance, Risk, and Compliance (GRC) requirements. Commvault says it uses data insights, access controls, tagging, metadata search, audit trails, and reports to help manage compliance needs over a large amount of unstructured data. The service provides predictable cost modelling, actionable data insights, flexible bring your own storage (BYOS) options, and compliance-ready operations. It’s virtually air-gapped with logical segmentation of the access keys. More information here.

…

Data protector Druva has earned the ISO/IEC 27001:2013 certification by the International Organization of Standardization. This milestone, it says, proves the company’s commitment to data security and ability to meet the highest standards for security excellence. ISO/IEC 27001 includes more than 100 requirements for establishing, implementing, maintaining and continually improving an information security management system.

…

World Digital Preservation Day, is taking place today, November 3, 2022 – so-called by the Digital Preservation Coalition. The members are listed here, and don’t include any storage supplier. Steve Santamaria, CEO of Folio Photonics, said about the day: “Storage that is highly reliable, long-lived, easily accessible, and cost-efficient is crucial to any data preservation strategy. We have yet to see an ideal storage technology developed that strikes the right balance between these vectors. However, new technologies such as next-generation tape storage, advanced optical storage, and DNA storage are all currently being developed to sit at the center of data preservation strategies around the globe.”

…

MRAM developer Everspin says its EMxxLX chips are now generally available with an xSPI serial interface, and is is the highest-performing persistent memory available today. It offer SRAM-like performance with low latency, and has a full read and write bandwidth of 400MB/sec via eight I/O signals with a clock frequency of 200MHz. The chip is offered in densities from 8 to 64 megabits and can replace alternatives, such as SRAM, BBSRAM, FRAM, NVSRAM, and NOR flash devices. Everspin says these are low-powered devices and deliver the best combination of performance, endurance, and retention available today. Data is retained for more than ten years at 85°C and more than 100 years at 70°C. Write speeds of more than 1000 times faster than most NOR flash memories are claimed.

…

Too many companies with no product news and nothing else to say run reports with findings that customers need their products. Today’s example is Kaseya with a Datto Global State of the MSP Report – “a report full of data and insights on topics that range from how MSPs run their business, the services they offer, and what is driving growth in today’s changing landscape.” The hybrid workforce is here to stay (you didn’t read it here first), ransomware and phishing are getting worse (you don’t say) and the biggest challenge facing MSPs is … wait for it … competition (as opposed to the challenges facing literally every other type of business in the world). Yawn. We’re told the most surprising finding is that the revenue generated from break-fix is increasing.

…

FD Technologies announced the appointment of Ashok Reddy as CEO of its KX business unit, reporting to Group CEO Seamus Keating. KX is a data analysis software developer and vendor with its KX Insights platform built upon the proprietary time series database kdb+ and its programming language q. The company was founded in 1993 by Janet Lustgarten and Arthur Whitney, developer of the k language. Investor Ireland-based First Derivatives (FD) increased its stake to 65 percent in 2014 and bought them out in 2018.

…

NetApp is going even greener. It has announced availability of new ways (BlueXP, Cloud Insights Dashboard, LCA reports, storage efficiency guarantee) for customers to monitor, manage, and optimize their carbon footprints across hybrid, multi-cloud environments. It has a commitment to achieve a 50 percent intensity reduction of Scope 3 Greenhouse Gas (GHG) emissions (those produced by an organization’s value chain) by 2030, and a 42 percent reduction of Scope 1 (emissions controlled or owned by an organization) and Scope 2 (electricity, heat, and cooling purchases) through adoption of a science-aligned target.

The Lifecycle Assessment (LCA) carbon footprint report uses Product Attribute Impact Algorithm (PAIA) methodology. PAIA is a consortium of information and communication technology peers, sponsored by MIT. The PAIA consortium allows NetApp to produce a streamlined LCA for its products that is aligned with leading standards including EPEAT, IEEE 1680.1 and the French product labeling initiative Grenelle.

…

Nutanix global channel sales boss, Christian Alvarez, is leaving for another opportunity after spending more time with his family. SVP Global Customer Success Dave Gwynn is taking over the position.

…

Pure Storage announced continued momentum within the financial services industry as global banks, asset management firms and financial technology organizations become customers. To date, nine of the top ten global investment banks, seven of the top ten asset management companies, and five of the top ten insurance organizations use Pure Storage products. Pure Storage’s financial services business reached an all-time high last year, growing more than 30 percent year-over-year in FY22.

…

In-cloud database accelerator Silk has promoted Chris Buckel from VP business development and alliances to VP strategic partnerships.

…

There are press rumors that Toshiba will finally be bought by a Japan Industrial Partners-led consortium for $16.1 billion in cash and loans. The consortium includes domestic Japanese companies Orix Corp and Chubu Electric Power Co, plus global investment firms Baring Private Equity Asia, CVC Capital Partners and others. A Bain-led consortium is also involved in the current (second) round of bidding.

Alation bags $123m in E-round investor funding

By
Chris Mellor
-

Data catalog and intelligence tech supplier Alation has raised $123 million in an E-round despite the generally uncertain environment for tech startups.

Customers want to know what data they have and then analyze it to seek out potential commercial opportunities and efficiencies, said Satyen Sangani, Alation CEO and co-founder

Satyen Sangani, Alation
Satyen Sangani

“The modern enterprise needs data in good times … and even more so in bad,” he said. “In one of the most challenging times to raise capital, especially for late-stage SaaS companies, we sourced the capital needed to continue to grow the business. While negative volatility has many pre-IPO companies slashing valuations and halting investments, we defied the odds, raised $123 million, and increased our valuation.”

Alation will use the cash to increase product research and development, build new products, and make acquisitions to expand its multi-tenant SaaS. The money will also be used to expand Alation’s global head count, currently standing at a little more than 700.

The funding, Alation says, follows five consecutive quarters of accelerated growth. The company was founded in 2012 and now has nearly 450 customers, including more than a quarter of the Fortune 100. It has passed the $100 million ARR mark. Total funding to date is $340 million and the business is now valued at $1.7 billion-plus, about 50 percent more than before. 

The latest round was round led by Thoma Bravo, Sanabil Investments, and Costanoa Ventures with participation from new investor Databricks Ventures. Other investors include Dell Technologies Capital, Hewlett Packard Enterprise (HPE), Icon Ventures, Queensland Investment Corporation, Riverwood Capital, Salesforce Ventures, Sapphire Ventures, and Union Grove.

Databricks is, of course, a startup as well, founded in 2013. It has raised $3.4 billion, with $2.6 billion coming from two rounds in 2021 alone.

There are more than 100 joint Alation-Databricks customers. Databricks and Alation reckon the two companies will accelerate joint go-to-market initiatives based on product integrations with Databricks’ Delta Lake and Alation’s Unity Catalog. This is intended to deliver data discovery, governance, and end-to-end lineage experiences for data engineering, data science, and analytics use cases on Databricks’ Lakehouse Platform.

Cohesity: Don’t rely on ransomware guarantees

By
Chris Mellor
-

Cohesity has suggested that ransomware recovery failure guarantees – such as the ones that Druva and Rubrik offer – are not worth the paper they are written on.

This is because the most common attack vector for ransomware – phishing (employee error) – is excluded.

The customer also generally has to abide by some definition of best practices, which is a recipe for legal dispute. They don’t cover ransom payments either.

Tim Robbins, Cohesity
Tim Robbins

Tim Robbins, Cohesity chief legal officer, told us in a briefing that he’d read the Druva and Rubrik ransomware guarantee documentation. He said: “The first thing that stands out to me is … there’s some right upfront exceptions and these definitions … exclude most forms of ransomware attack – particularly foreign intrusion.”

Our understanding is that if you download software in your environment thinking it’s a harmless application and it turns out it hides ransomware, you are not covered. Additionally, these companies don’t pay the ransom – they pay recovery costs.

A look at the Rubrik and Druva warranty agreement documents bears this out:

Cohesity is critical of such exceptions
Druva ransomware exception from its resiliency guarantee agreement
Cohesity is critical of such exceptions
Equivalent Rubrik exclusions from its Recovery Warranty Agreement

Robbins said these definitions “exclude the things that are most dangerous and most common – for example, losing your credentials, turning your credentials over to the bad guys. That’s the number one cause of ransomware or foreign third-party intrusion into your network.”

Our understanding is that general cyber security companies do not offer ransomware guarantees against phishing and equivalent attacks because they are so difficult to prevent. Robbins agreed.

A disadvantage of ransomware guarantees is that they can put the customer and supplier into an adversarial relationship instead of a cooperative partnership. Robbins characterized this as “a blame game, rather than a team recovering when the chips are down.”

Then he revealed that Cohesity had offered to match these guarantees in some customer deals. “We’ve offered them. We said, if you really need it, to check the box or something, we would match these guarantees. But we tell them why they’re not worth it. We think they’re negative value. And we’ve never had a customer ask us for one.”

Druva and Rubrik

A Druva spokesperson said: “While cyber crime is on the rise and customers are looking for any advantage they can get, they are right to be discerning about the value of recent ‘guarantee programs’. Many programs are 1) singularly focused on specific threats and/or 2) are loaded with legal jargon to exclude, disqualify, or otherwise limit payout. The reality is many of these guarantees are first and foremost marketing campaigns in disguise. Fortunately for Druva customers, none of these conditions apply to our new Data Resiliency Guarantee.

“In fact, our guarantee includes five service-level agreements (SLAs) that cover the ransomware limitations and exclusions that Cohesity references. We also cover four other common data risks including human risk (ie accidental deletion), operational risk (ie data rot/corruption), application risk (ie backup reliability), and environmental risk (ie uptime). More importantly, Druva’s standard customer agreement already includes many of these clauses.

“In other words, our customer commitment, and the technology to support it, has been in place and proven out over years. The Data Resiliency Guarantee simply formalizes our commitment with a financially backed guarantee. As for abiding by best practices, we do require customers to use the features we built to protect their data, and to use industry standard best practices. We do not believe this puts us at odds with our customers.

“Finally, given Druva is 100 percent SaaS, it is in our best interest to help our customers successfully combat ransomware. A well-structured, straightforward guarantee gives customers the confidence that we are putting skin in the game to work with them, rather than making it ‘their problem’.”

When asked about Cohesity’s points, a Rubrik spokesperson replied: “We don’t have anything to share on the topic at the moment.”

Solidigm CEO’s departure takes staff by surprise

By
Chris Mellor
-

Rob Crooke, the CEO of Solidigm – SK hynix’s acquired Intel SSD business – has suddenly departed.

Crooke ran the Intel SSD business when it was sold to SK hynix for $9 billion in January. SK hynix set it up as a separate business unit and branded it Solidigm with Crooke as the CEO.

Rob Crooke, former Solidigm CEO
Rob Crooke

Now SK Hynix has confirmed a new leader: “Noh-Jung (NJ) Kwak, current co-CEO of SK hynix, has been appointed Interim Chief Executive Officer. This change follows the departure of Rob Crooke, who served as Solidigm’s CEO since the company’s launch. We are grateful to Rob for his leadership and important contributions in establishing Solidigm. Our team members remain focused on delivering an unmatched product portfolio and best-in-class service to our customers.”

Solidigm told us that “Woody Young, a current member of the Solidigm Board, has been appointed President of Solidigm.” Young is an ex-investment banker.

In addition we’re told: “The Solidigm Board has commenced a search for a new CEO to lead Solidigm. In the interim, we are confident that this senior leadership team is very well-positioned to lead Solidigm and to continue to serve the best interests of our customers.”

The company reiterated: “All of us at Solidigm are grateful to Rob for his leadership and important contributions in establishing Solidigm.”

Noh-Jung Kwak, Solidigm
Noh-Jung Kwak

A person close to the situation told B&F that Solidigm employees were informed at the same time as official statement was released. “Rob was simply gone,” they said.

In Crooke’s time as Solidigm’s CEO the company upgraded its datacenter PCIe 4 connected SSDs, released the P44 Pro gaming SSD. It also started promoting the penta-level cell (5bits/cell or PLC) NAND concept as the follow-on to today’s QLC (4bits/cell) flash. 

Woody Young, Solidigm
Woody Young

Crooke also was involved in the decision to keep Solidigm’s NAND technology separate from SK hynix technology for a period, with eventual fusion assumed to take place some years in the future. It is more costly for a company to have two NAND production technologies.

SK hynix recently reported a 7 percent revenue drop in its third 2022 quarter to ₩11 trillion ($7.7 billion) with net profit falling 67 percent year-on-year to ₩1.10 trillion ($771 million) and is reducing cap-ex investments.

It may be that a strategic difference over Solidigm’s technology independence from SK hynix and capital spending contributed to Crooke’s departure, however, this just speculation.

Micron is also feeling the squeeze from the global economic situation and has announced capital spending reductions.

Kioxia brings SSDs to AI and Arm environments

By
Chris Mellor
-
3D rendering of AI brain

Kioxia has extended the market for its SSDs by getting them qualified for Arm servers, and by devising an AI image classification method using SSD data.

It seems counterintuitive, but apparently AI models can get dumber as they learn. Kioxia says that when existing neural network models are retrained with fresh data, leading to revised parameter weights, knowledge acquired in the past can be lost, leading to deterioration of classification accuracy. It calls this “catastrophic forgetting.” 

One possible fix is unpalatable: retrain the neural network from the beginning, which can require huge amounts of time and energy costs. The other fix is obvious, and conceptually simple: have the existing neural network knowledge – image data, labels and image feature maps – stored in SSDs, and refer to this when the model is being retrained with fresh data.

Kioxia graphic
Kioxia graphic

Using this method, knowledge can be updated by adding newly obtained image labels and feature maps to the stored data, and image classification can be maintained more accurately. 

Kioxia says there are two benefits. First, by using the data referred from the storage when the neural network classifies images, the basis for the classification results can be visualized, which is expected to improve the explainability of AI and allow the selective modification of knowledge sources. Second, by analyzing the referred data, the contribution of each stored data item can be evaluated according to the frequency of references.

This technology was presented on October 25 at the oral session of the European Conference on Computer Vision 2022 (ECCV 2022) in Tel Aviv. The paper title was “Revisiting a kNN-based Image Classification System with High-capacity Storage”. More information can be found here and here.

Kioxia hopes to apply it to other neural network problems areas.

Kioxia getting Arm-powered

Kioxia and Ampere have qualified Kioxia’s CD6, CM6, and XD6 Series SSDs with platforms based on Ampere’s Arm CPU-based Altra and Altra Max Cloud Native Processors. The CD6 and CM6 SSDs have PCIe 4 interfaces and a U.3 format. The XD6 has an EDSFF E1.S format and a PCIe 4 interconnect as well.

Kioxia XD6

Sean Varley, senior director of product marketing at Ampere, said: “We’ve worked with Kioxia to certify its broad offering of NVMe SSDs with our Ampere Altra family of platforms for our mutual customers, including the top hyperscale cloud customers and leading ODM/OEM server manufacturers.”

HP’s latest generation of ProLiant servers include x86-powered models and Ampere processor products as well. Other SSD manufacturers can be expected to follow Kioxia in getting their SSDs similarly qualified.

Supermicro partnering with GRAID and Qumulo

By
Chris Mellor
-

Supermicro is partnering with GRAID and Qumulo in two separate deals to have customers buy its servers fitted with GRAID’s GPU-powered RAID cards or used to run Qumulo’s scale-out filesystem software.

Update: GRAID erasure coding capability added. 7 Nov 2022.

These deals give GRAID a new channel to market for its SupremeRAID card, and Qumulo a value-oriented filesystem node.

Supermicro and GRAID Technology are co-operating in customer deals. Supermicro EMEA President Vik Malyala said: “We are thrilled to team with an innovative and sustainable company as GRAID Technology.” GRAID’s card “offloads RAID and NVMe data durability tasks from the CPU, freeing server resources for compute intensive workloads. Customers will quickly see the benefits in applications ranging from VMs and Containers to AI and HPC. It’s a very effective addition to high performance storage based on Supermicro’s flexible Building Block Solutions.”

It is claimed that the combination of Supermicro and SupremeRAID will offer customers an easy-to-configure option for data protection with a lower cost of deployment and outstanding NVMe and NVMeoF performance.

GRAID CEO Leander Yu spoke of Supermicro and GRAID having “a true competitive advantage in the NVMe and NVMeoF marketplace where together we will have an unbeatable combination.”

Supermicro now gets added to the existing list of GRAID’s technology and go-to-market partners: 

Qumulo and Supermicro

Scale-out filesystem supplier Qumulo has announced it’s supplying a hybrid-NVMe SSD and disk drive product from Supermicro that optimizes performance and price in a single platform. 

Qumulo CTO Kiran Bhageshpur said: “We are thrilled to offer our customers more platform choices. Qumulo’s intelligent caching technology, which ensures that most reads and all writes are NVMe-first for best performance, takes advantage of the new Supermicro NVMe + HDD platform to maximize performance at an affordable price point.”

Don Clegg, SVP Worldwide Sales at Supermicro, said: “Our continued collaboration with Qumulo  delivers industry-leading storage solutions for customers demanding higher performance and lower TCO.”

The actual Supermicro product isn’t specified but we are told these “new Hybrid-NVMe nodes provide high-density storage in just 1U of rack space.” Scanning Supermicro’s website product pages reveals storage servers like the 1U SSG-610P-ACR12N4H and SSG-610P-ACR12N4L which would fit that description: 

Supermicro servers

These have gen 3 Xeon CPUs, 12 x 3.5-inch drive bays, and 2 or 4 SDD bays. Twelve 20TB HDDS would give them 240TB of raw disk capacity and 10 would provide 2.4PB, plenty enough for Qumulo nodes.

An all-flash Supermicro system fitted with GRAID’s card and running Qumulo’s software ought to provide a screamingly fast scale-out filesystem package. Qumulo implements erasure coding, but GRAID already has that, so the possibility is obvious.

VAST Data to ‘revolutionize’ data lake market

By
Chris Mellor
-

VAST Data claims to have has grown its business significantly faster than the general storage industry and says it will revolutionize the data lake market by onboarding OEMs in the coming months.

More than a dozen customers spent more than $10 million on its Universal Storage product in the first half of 2022, in dicating a potential $120 million run rate from them alone. Its Annual Recurring Revenue (ARR) at the end of the period was 4.2x higher than a year ago. Three customers have now exceeded more than $100 million in commitments for Universal Storage since first buying the product. VAST has been cash flow positive, on average, for the last six quarters.

Renen Hallak, VAST Data
Renen Hallak

Renen Hallak, VAST co-founder and CEO, said in a statement: “Organizations are waking up to the fact that they need more utility from their data than legacy data platforms will allow them to realize… The platform revolution is only just beginning, and we can’t wait to show everyone our next acts which will be unveiled over the coming months.”

The announcement says VAST’s momentum will continue as it ”doubles down on the all-flash cloud, revolutionizes the data lake and onboards world-leading OEM partners, global system integrators and key ISV partners.”

Board member Gary Reiner said of VAST’s growth: “We are even more excited about VAST’s near-term plans to continue innovating, expanding further into the cloud with remarkably advanced data management capabilities.”

So VAST will be adding data lake and data management capabilities, and OEM partners that are world-leading, without saying if they are in the IT hardware systems, data lake or data management business. Should we be thinking of Databricks, Dremio, and Snowflake, or traditional system OEMs like Dell and HPE? Or both even, and where does the public cloud play in this?

Co-founder and CMO Jeff Denworth has written a blog in which he says “VAST wants to be a killer of categories in the future” with the URL linking to a Blocks & Files story about VAST working on the Thinking Machines vision. It appears we were getting close to what VAST is doing with data science infrastructure software development. Stay tuned.

How simple, stateful storage and Kubernetes streamlines developer productivity

By
David Gordon
-

By iterating on standards, HPE CSI Driver and storage approach smooths application dev lifecycles

SPONSORED FEATURE Containers have become the wellspring of application developer productivity gains, especially in the cloud-native world. However, owners of stateful applications now have the same capabilities at their fingertips, all without the storage-based limitations of years gone by.

It’s worth taking a look back at the evolution of Kubernetes and storage if only to show how far companies like Hewlett Packard Enterprise (HPE) have advanced. Plus, it gives us a sense of the progress made toward allowing users to instantly provision storage and clone production-level datasets while steering performance and capacity during runtime for peak reliability that yields better application response times, especially during periods of high congestion.

No more storage bottlenecks

For application developers and release engineers, storage was long considered a bottleneck which hindered successful Kubernetes deployments. While this might have been the case in the early days when various plugins offered little or no data management or dynamic provisioning capabilities, times have changed.

It used to be challenging to incorporate vendor-specific storage drivers into Kubernetes. And while there were plenty of efforts to that effect, they did not deliver the broad functionality that standardization initiatives generally provide. This is one reason why Google and others worked to develop the Container Storage Interface (CSI) specification, which allows third parties to write a driver within their own domain that could present standard constructs onto Kubernetes.

CSI provides a thin shim between the container orchestrator and the backend storage systems that is unique for each vendor. Now that it is mature and widely recognized by developers, customers can pop to the CSI documentation and see all the drivers for various storage interfaces. While this is certainly helpful, it can still be a bit of a jumble to sift through, however.

HPE has laid its engineering and storage systems expertise on top of this vendor-neutral effort to allow users to tap into a container storage provider (CSP – a REST API specification that defines how provisioning, mounting and deallocation workflows are invoked from a host client) which is able to handle all the data management operations on a range of storage infrastructure.

This shift means that Kubernetes deployments are no longer restricted to the more ephemeral world of stateless applications. Stateful application developers can also capture the full promise of container-based development with better data management, higher efficiency, secure multi-tenancy and far faster development times, even for traditional applications going through modernization.

Pushing productivity

Most applications that deploy on Kubernetes and run in containers are derived from stateful applications of one sort or another (databases, for instance). As storage flexibility has increased, this means the next challenge is to integrate production datasets into test workflows, especially as higher productivity leads to faster software releases, which drives higher quality deployments.

All of this also means testing needs to improve with the addition of production-like data in the dev/test and staging environment.

That is where the functionality of the HPE CSI Driver for Kubernetes comes in. Stateful application owners can integrate that level of data early on in test cycles using instant zero-copy clones of production instances in Kubernetes and other functions inherent in their enterprise storage. With the ability to clone and attach that data to basically any type of container infrastructure where the driver is available, the bar for application development in containers has been raised.

Although vendor-neutral, the HPE CSI Driver has support for HPE’s entire storage portfolio, which is especially useful as large enterprises go on their application modernization journeys.

The HPE CSI Driver’s power is in allowing the ability to dynamically provision persistent volumes which can be set by system and storage admins. The focus of HPE’s driver is to provide a consistent experience across private cloud environments.

Control runtime storage performance/capacity

HPE CSI Driver for Kubernetes – when attached to primary storage like the all-NVMe HPE Alletra cloud-native data infrastructure (powered by HPE GreenLake), HPE Primera or HPE Nimble storage systems – means users can instantly provision storage and clone production datasets while controlling volume performance and capacity during runtime to minimize operational disruption and downtime. Users can also deploy multiple Kubernetes clusters on the same infrastructure with top-line storage infrastructure backing production-level data for full developer productivity.

When storage is kept distinct from the Kubernetes cluster, existing clusters can be wound down and redeployed without changing the data, which remains shielded and can be made active again as soon as a new cluster is provisioned. In short, the development environment can be deployed and recovered quickly, without too much concern over the safety of the data—a major advantage for test-to-production projects.

Combining these external storage lines delivers an ideal environment for customers who want to scale compute, network, and storage individually. In addition to snapshots and clones, users can also get sophisticated multitenancy so you can run production, staging, and test/dev on the same storage by compartmentalizing all the storage resources to support a particular workload, and then you can scale that back or grow as the need for capacity or performance changes.

Existing HPE customers won’t need to know the details of the various HPE storage systems, and they can replicate data between the Alletra 9000 and Primera and between the Alletra 6000 and Nimble systems, for even better scaling and flexibility of performance.

HPE hyperconverged solutions such as HPE Alletra dHCI (disaggregated hyperconverged infrastructure) or HPE SimpliVity data virtualization platform can give customers the flexibility of using storage provided by the infrastructure layer CSI drivers, with the option of using HPE CSI Driver for Kubernetes for richer data management for stateful applications.

Easy storage provisioning for Kubernetes clusters

But perhaps the most important aspect of the HPE CSI Driver is that storage managers can delegate control and command and conversely, application developers writing containerized microservices applications don’t know – and don’t have to care – that enterprise-grade storage with features such as zero copy snapshots are making their stateful storage more scalable and resilient. They can do it in a self-service manner, just like they could provision virtual machines and stateful storage on server virtualization hypervisors in the days before containerization.

All of which leaves the application developers to do what they do best, and the storage managers to make sure the scale is there to support applications as they move from test/dev through qualification and into production. All on enterprise-class, scalable storage.

“Containers are designed for portability, so any organization using them must create an architecture that allows for data to be wherever it needs to be, whenever it needs to be,” says Michael Mattsson, Technical Marketing Engineer at HPE.

“Containerization, and Kubernetes in particular, helps businesses move away from imperative operational models that require multiple teams to perform tedious tasks on a regular basis. It allows them to adopt an agile declarative model where there’s a clean separation of concerns, along with a high degree of automation and abstractions that make sense for running a high-performance technology company,” he adds.

Developers already have enough to think about in getting stateful applications into production as quickly as possible – the HPE CSI Driver might at least mean that having sufficient underlying storage resources ready and available isn’t one of them.

Sponsored by HPE.

Backblaze: Buying less reliable disk drives can make financial sense

By
Chris Mellor
-

Backblaze has published its quarterly disk drive reliability stats and explained how buying less reliable drives can be cheaper in the long run.

There is little change from last quarter’s report: the average annualized failure rate (AFR) in the quarterly table was 1.64 percent measured across 226,697 drives, with the worst performers being Seagate’s 4TB (4.38 percent), 12TB (4.96 percent), and 14TB (9.2 percent) models. A Toshiba 4TB drive had an 8.25 percent AFR, but Andy Klein, Backblaze principal cloud story teller, pointed out: “The high AFR (8.25 percent) is due to the limited number of drive days in the quarter (8,849) from only 95 drives.”

Backblaze hard drive failure rates

“Three drives had zero failures this quarter: the 8TB HGST (model: HUH728080ALE604), the 8TB Seagate (model: ST8000NM000A), and the 16TB WDC (model: WUH721816ALE6L0). For the 8TB HGST, that was the second quarter in a row with zero failures. Of the three, only the WDC model has enough lifetime data (drive days) to be comfortable with the calculated annualized failure rate (AFR).”

The Backblaze drive fleet is getting older. Klein noted: “The AFR for Q3 2022 was 1.64 percent, increasing from 1.46 percent in Q2 2022 and from 1.10 percent a year ago. As noted previously, this is related to the aging of the entire drive fleet and we would expect this number to go down as older drives are retired and replaced over the next year.”

In terms of lifetime stats, a couple of Seagate drives have greater than 3 percent annual failure rates; a 16TB one with a 3.3 percent AFR and a 14TB one failed more often with a 5.47 percent AFR. The average quarterly AFR was 1.41 percent across 226,000 drives.

Buying less reliable HDDs

Having accumulated all this data over many quarters, Klein posed the question: “Why would we continue to buy a drive model that has a higher annualized failure rate versus a comparably sized, but more expensive, model?” The answer comes from studying the relationship between acquisition costs and drive failure rates.

He supplies a theoretical table showing the acquisition costs, annual replacement costs, and failure rates for three 14TB hard disk drives to answer his question:

Blackblaze hard drive costs

The three drives are priced as low, medium, and high with AFR values of 1.5, 1, and 0.5 percent respectively. They share the same replacement labor cost, $300. There is no drive replacement cost as “all drives are returned for credit or replacement to the manufacturer or their agent.”

The table lists the total cost of the drives over five years, factoring in the total lifetime drive replacement cost, which is added to the total cost of each drive model.

Even though, with the low-end model, more drives fail and the total drive replacement cost is higher than for the middle and top-end models, the acquisition cost is low enough for the total drive cost to be lowest of the three. The last line in the table shows this comparison.

Klein then asked: “How much failure can we tolerate before our original purchase decision is wrong?” Using the values in the table he computes that:

  • Model 1 and Model 2 have the same total drive cost ($1,325,000) when the annualized failure rate for Model 1 is 2.67 percent
  • Model 1 and Model 3 have the same total drive cost ($1,412,500) when the annualized failure rate for Model 1 is 3.83 percent.

He said that modeling drive costs and failure rates this way is important when buying thousands of drives a year: “The need for such a model is important in our business if you are interested in optimizing the efficiency of your cloud storage platform. Otherwise, just robotically buying the most expensive, or least expensive, drives is turning a blind eye to the expense side of the ledger.”

Commvault delivers 10 quarters of growth

By
Chris Mellor
-

Against a background of storage media suppliers announcing downturns, Commvault reported its tenth growth quarter in a row, beating analysts’ earnings and profits estimates.

Revenues were up 5.8 percent year-on-year to $188.1 million in its second quarter of fiscal 2023 ended September 30, and profits rose to $4.5 million from $1.73 million.

Businesses may be buying fewer disk and solid state drives, but they need to protect the data they have. CEO and president Sanjay Mirchandani said: “Our fiscal Q2 record results and double-digit constant currency growth reinforce that customers see the value of Commvault’s software and SaaS solutions. We believe our comprehensive data protection portfolio has never been more important in today’s increasingly difficult world.”

Annual recurring revenue (ARR) represented 84 percent of Commvault’s earnings. It was 79 percent a year ago Q2 and 75 percent a year before that.

Revenue segments:

  • Software and products revenue: $82.8 million, up 10 percent
  • Service revenues: $105.2 million, up 3 percent
  • Total recurring revenue: $158.2 million, up 12 per cent year-on-year
  • Annualized recurring revenue: $604.4 million, up 11 percent

There were 173 large deal transactions compared to 163 a year ago. The average dollar amount of these was approximately $346,000, an 11 percent increase from the prior year. Large deals accounted for 72 percent of Commvault’s software and products revenues, $59.6 million, and 32 percent of Commvault’s revenues for the quarter.

Commvault revenues by quarter
Ten quarters of growth

Commvault reported $49.8 million in operating cash flow, up 90.8 percent annually. Total cash at quarter end was $262.5 million compared to $267.5 million as of March 31, and it has no debt. It said earnings per share were $0.57 compared to $0.48 a year ago. 

Customers stored 3EB of data in the cloud, a 2.5x increase since Q2 of fiscal 2022.

Mirchandani said in an earnings call that the exec team had returned Commvault to responsible growth. Customers, he said, wanted both software and SaaS approaches to data protection and Commvault supplies both via it’s Metallic SaaS. There were more than 2,500 Metallic customers at quarter end, 72 percent of them new to Commvault. There were fewer than 2,000 at the end of fiscal 2021.

Metallic surpassed $75 million in ARR this quarter, up 50 percent since the start of the fiscal year, and half of the Metallic customers had also bought another Commvault product. It is a land-and-expand strategy and Commvault wants to accelerate this. Mirchandani said he expects every customer over time to need both Metallic and Commvault’s Complete Data Protection software product: “That’s our playbook. That’s how we go to market.”

The outlook for the next quarter is $203.5 million plus/minus $1.5 million, up just 0.54 percent year-on-year. CFO Gary Merrill said it was taking longer to close new business, and Mirchandani agreed, adding that closing got harder across the quarter, particularly in Europe. Merrill said this longer time to close was factored into the outlook.

1...176177178...453Page 177 of 453

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