Data Processing Units (DPUs) help organisations build denser, faster, more efficient and cost-effective IT infrastructures, with the goal of providing an overall lower total cost of ownership.
This is the thrust of a GigaOm Sonar report, written by analyst Enrico Signoretti, looking at early-stage emerging technologies. DPUs are hardware accelerators, usually installed on commodity x86 servers, to offload specialised compute tasks such as security, storage and networking. They can be implemented as ASICs, FPGAs or proprietary Systems-on-Chip (SOCs) using Arm or specially-developed processors.
The Sonar report evaluates suppliers’ products looking at their performance, programmability, power efficiency, longevity and cost. It also checks key characteristics needed for enterprise adoption: architecture, drivers, APIs, ecosystem, management and support.
The included suppliers are Fungible, Intel, Kalray, Marvell, Nvidia, Pensando and Pliops. They are placed in a triangular 2D area to represent their relative positions as Challengers or Leaders. The three axes defining the space are technology, strategy and roadmap. The three axes start at the same central point (signifying a higher score) and then move outwards.
Here is the report’s Sonar diagram:
The orange squares are the suppliers’ starting positions, and the arrows show their direction of movement. At this early state of DPU development they are all Leaders – although Marvell is on the Leader/Challenger boundary – and all located in the technology-centric area of the diagram.
Nvidia and Pensando are the overall best-positioned, with Fungible, Intel and Pliops next, followed by Kalray and Marvell.
The report contains descriptive sections for each supplier, describing their product’s characteristics, its strengths and its challenges. It advises: “The DPU is a component of the server. Look to purchase it with the server so the vendor provides warranty and support for both as part of the server maintenance plan.” That seems too be an excellent point for all but hyperscaler customers.
Dell’s quarterly revenues rose 21 per cent year-on-year to a record $28.4 billion, but storage revenues were anaemic with a mere 0.9 per cent rise to $3.89 billion.
This was the best third quarter in Dell Technologies history — helped, Dell said, by robust demand, durable competitive advantages, and strong execution. The company generated record revenue of $28.4 billion in the quarter ended October 29, up 21 per cent, with growth in all business units, customer segments and geographies, and strength across commercial PCs, servers and (much less so) storage. VMware revenue was $3.2 billion — up 10 per cent with general strength across its product portfolio. Dell has now separated itself from VMware.
Chuck Whitten, co-COO Dell Technologies, said in a statement: “We’re three quarters into what will prove to be a historic year for Dell, and we are just beginning to write the next chapter of the Dell Technologies story. We are uniquely positioned in the data era, with durable advantages and market-leading positions. Our strategy is focused on growing our core business and in adjacent multi-billion-dollar markets including multi-cloud, edge, telecom and as-a-Service.”
Jeff Clarke, vice chairman and co-COO Dell Technologies, said in his statement: “Our product, global operations and sales teams did an outstanding job this quarter as we shipped a record number of products and delivered record revenue.”
Yes, but storage rather let the side down. We’ll focus on that while acknowledging that Dell has done an outstanding job elsewhere in its results for the quarter.
Storage
Storage revenues in the Infrastructure Solutions Group (revenues $8.4 billion; up 5 per cent) were almost flat while server-plus-networking revenues rose 9 per cent year-on-year to $4.5 billion.
A chart shows a widening gap between the two product categories over the past three quarters:
The seemingly stubborn resistance Dell’s storage shows to product sales growth is holding ISG revenues back. We should say that Dell is the leading storage product shipper, being number one in external enterprise storage, storage software, all-flash arrays, hyperconverged systems, converged systems, and purpose-built backup appliances. Yet it has barely grown storage revenues over the past three quarters while competitors — such as Pure Storage and NetApp — have grown their revenues.
What is the problem? A results presentation slide said there was “Strong storage demand, orders growth in Hyperconverged Infrastructure up 47 per cent, Data Protection up 26 per cent and Midrange up 18 per cent.” There must have been significant weakness elsewhere in the product range to have overall storage growth be just 0.9 per cent
This suggests disappointing high-end array sales could be a factor and possibly also low or no growth in Dell’s PowerScale file and ECS object storage products. There could be increased competition from suppliers such as Infinidat, Pure Storage, VAST Data, Qumulo and others to account for this.
Earnings call
We looked for any clues in Dell’s earnings call. Co-COO Jeff Clarke declared: “We are pleased with our storage performance in Q3, where we saw storage return to growth.”
He declared: “Momentum in our mid-range storage business continues to be led by PowerStore, where 23 per cent of PowerStore customers renewed to Dell storage and 28 per cent were repeat buyers. PowerStore is the fastest-ramping storage product in our history.”
That’s all very well but storage must have been a disappointment to Clarke with its lacklustre growth.
Answering a question, CFO Tom Sweet said: “Storage demand tends to be more back-end loaded in the quarter, and we clearly saw that again this quarter, perhaps a bit more than in prior quarters. And as a result of that, we were not able to convert that backlog to revenue … and we also deferred a fair amount of revenue to the balance sheet, just given the service attach rate as well as the software content within the storage.”
Whitten jumped in to strengthen Dell’s answer here: “We’re encouraged by storage orders growth because in the most strategic category, software defined storage, we grew 47 per cent, mid-range orders grew 18 per cent, which is now the fourth consecutive quarter our mid-range business has grown. … Data protection grew, unstructured data grew and our entry-level orders grew as well in the storage business, so … we remain encouraged by the orders growth.”
Will this backlog translate into growth next quarter? Possibly not, as Sweet said: “The reality is, as we highlighted in the talk track that we’re continuing to face supply chain challenges. And so how much of server demand gets converted or backlog gets converted into shippable revenue is something that the teams are working every day.”
He also said though: “that Q4 tends to be a higher storage quarter for us.”
It seems fair to assume that Dell is facing headwinds in the high-end array, filer and, possibly, object storage space. There must have been quite severe revenue declines in its storage portfolio outside the highlighted growth areas, such as the mid-range. This suggests that there are product weaknesses needing to be fixed and that could be a multi-quarter exercise.
Nutanix reached a negative landmark, with losses exceeding revenues in its latest quarter, but financial analysts are happy with its progress.
Revenues were $378.52 million in its first fiscal 2022 quarter, which ended October 31, up 21 per cent annually, for a loss of $419.8 million. For every dollar it earned Nutanix spent $1.11 — it bought its growth. The year-ago quarter’s loss was $265 million. In the previous quarter it dropped 92 cents in costs for every dollar of revenue. Things have got worse.
President and CEO Rajiv Ramaswami’s results statement read: “Our first quarter was a good start to our fiscal year, demonstrating strong year-over-year top and bottom line improvement.” The “bottom line” term certainly did not refer to GAAP net income.
CFO Duston Williams said: “We achieved record ACV billings, which grew 33 per cent year-over-year, and saw 21 per cent year-over-year revenue growth, our highest growth in over three years.”
Financial summary
Free cash flow — improved to -$1.9M from -$16.3M a year ago;
Gross margin — 78.5 per cent compared to 78.3 per cent a year ago;
Annual Contract Value (ACV) — $183.3M vs last year’s $137.8M, up 33 per cent;
Annual Recurring Revenue (ARR) — $952.6M vs $569.5M a year ago;
Cash and cash equivalents at end of period — $350.99M compared to $504.5M last year.
The quarter was the third in a row showing annual growth and one with a steep rise:
Nutanix grew its customer count by 570 in the quarter to a total of 20,700 but the increase was the lowest for five years or more. The customer acquisition rate trend is downwards. We might say Nutanix is spending more, as shown by deepening losses, to gain fewer new customers.
But customer spend increased and analysts were not concerned, as we shall see.
There was a minimal impact of Nutanix’s business from the COVID resurgence, as customer businesses have learnt how to do remote working. There was little to no supply chain impact as Nutanix has multiple hardware partners.
Nutanix saw more customers buying more products — 42 per cent of customer deals involved at least one so-called emerging product, which includes all add-ons beyond the basic HCI offering. This was up slightly, 7 points Nutanix said, year-on-year. A basis point is equal to 1/100th of one per cent, which is why we said it was up slightly.
Analyst views
Wells Fargo’s Aaron Rakers told subscribers: “[Nutanix] delivered positive F1Q22 results (and forward guide) driven by a seasonally strong federal business, significant upselling, and continued execution on subscription renewals.” He said Nutanix “provided investors with increased confidence in the company’s path to profitability,” and “Nutanix expects significant growth in emerging products and new ACV bookings in F2Q22.”
Rakers commented on the customer acquisition rate: “While new logo additions decelerated, Nutanix’s ASP per new logo was up year over year and quarter over quarter as it focuses on quality/efficiency of new logos. [Nutanix] now has 1,580 customers that have purchased >$1 million, up 68 quarter over quarter.”
Williams said in the earnings call: “We are generating more new logo ACV bookings with less new logos,” which explain’s Rakers’ view.
William Blair’s Jason Ader said: “Nutanix reported another solid beat-and-raise in its fiscal first-quarter print as the company benefited from improving hybrid cloud infrastructure demand, returns on its solution selling investments (including higher win rates, healthy renewals, and a strong attach rate for add-on products), and enhanced partner leverage.”
Ader also sees a “significant renewal opportunity ahead”. He pointed out that: “Management continues to view its rapidly approaching renewal opportunity as the key to unlocking operating leverage and achieving its target of free cash flow break-even in the next 12–18 months (as well as operating profit in calendar 2023).”
This means analysts were not concerned over the big loss, with Ader saying: “The company gained operating leverage from the higher revenue and spent less than expected.” But he did point out: “Risks to the Nutanix story include competition from Dell/VMware and cloud titans, a high cash burn and deep level of operating losses.”
Guidance
The guidance for the next quarter is for revenues between $400 million and $410 million — an annual increase of 16.9 per cent at the $405 million mid-point. Growth is slowing. Revenue guidance for the whole fiscal 2022 year is $1.615 to $1.630 billion — a 16.7 per cent increase on FY2021 at the mid-point.
Pure has grown third quarter revenues a whopping 37 per cent year-on-year, surging back after last year’s third quarter 4 per cent revenue fall. That was its first negative growth quarter since we started reporting its results more than five years ago.
Revenues in the quarter ended October 31, 2021, were $562.7 million. They were $410.6 million a year ago, with a loss of $28.7 million — less than half the year-ago $74 million loss. But look at the quarterly trends so far this fiscal year. The chart below shows a declining loss trend throughout this fiscal year and if the fourth quarter comes in at the predicted $630 million then Pure could turn a profit — a GAAP profit — its first ever. That would be a landmark event in its history.
See the profit/loss trend in FY2022 at the right end of the chart.
Chairman and CEO Charlie Giancarlo said in his results statement: “With Q3 revenue up 37 per cent year-over-year and with increasing profitability, it’s clear that Pure continues to set the pace for the industry.”
CFO Kevan Krysler said: “Our strong Q3 performance was fueled by increased customer demand and execution across the entire business. We are in a great innovation cycle with our portfolio.”
Giancarlo’s prepared remarks reflected this, as he predicted: “Our next announcement, on December 8th, will … extend the breadth of our FlashArray platform.”
Financial summary:
Subscription services revenue — up 38 per cent year-on-year;
Subscription Annual Recurring Revenue (ARR) — $788.3M, up 30 per cent year-over-year;
Gross margin — 66.6 per cent;
Operating cash flow — $127.0M;
Free cash flow — $101.3M;
Total cash and investments — $1.4B.
Pure gained 345 new customers in the quarter — 12 per cent year-over-year growth — taking its total to, we calculate, just shy of 10,000 (actually 9,992 give or take).
The Q3 revenue bounce back is seen clearly in a chart of revenues by quarter by fiscal year — see the yellow line’s dip and steep ascent.
The trends are good and Pure has uplifted its full year forecast to $2.1 billion — a 25 per cent increase on FY2021. This, with its Q4 $630 million contribution, rebuts any ideas that the Pure would be hit hard by the departure of sales chief Dominic Delfino earlier this month. He was replaced by Dan FitzSimons, who got a “special shout-out” from Giancarlo.
Earnings call
Giancarlo said in the earnings call that Pure had: “double-digit quarter-over-quarter growth across all product lines and across both US and the international markets” and “Pure continues to take share.”
He made a remark that strengthens our belief that Pure could make a profit in the next quarter: “We are also pleased with our strong profitability trend continuing through this fiscal year.” Krysler said he saw no sign of demand relaxing in his initial look at the next fiscal year.
Giancarlo acknowledged supply chain problems but said they had been largely overcome. “This past quarter, global semiconductor availability was more challenging than last quarter, and we expect this environment to continue into next year. However, our operations team and the strong partnerships we’ve built with our suppliers have continued to work well, minimising impacts to our customers and our business.”
Pure will publish its first environmental, social and governance report early next calendar year, with Giancarlo saying: “Pure’s products use dramatically less energy and create far less waste than competitive offerings.”
Hyperscaler FlashArray//C sale
Krysler dropped this gem about a product sale: “Our sales growth this quarter also includes sales of FlashArray//C to one of the top 10 hyperscalers.” A $10 million-plus FlashArray//C sale to a hyperscaler was predicted back in August. Without that sale the quarter’s growth “would more likely be in the high 20s,” rather than the reported 37 per cent. This particular sale could be repeatable with Krysler saying: “No reason for us not to believe that it’s repeatable and conversations continue.”
Giancarlo said that: “It’s also worth noting that the overall footprint savings was a key part of winning the initial deal,” referring to a disk-based alternative.
He talked about a disk-to-flash crossover, saying: “We believe very strongly that as flash continues to decline relatively to the declines in magnetic disk, that there’s inevitably going to be a crossover point where every player everywhere including the hyperscalers will start switching from disk to flash. And it’s just a matter of time and their particular use case or instance before that happens. … before flash is used in a more … mainstream way in the hyperscale environment.”
All the action is centred down on the bottom-left in Gartner’s latest hyperconverged infrastructure (HCI) software magic quadrant (MQ).
This annual MQ rates HCI suppliers on low-to-high “completeness of vision“ and “ability to execute” axes, defining a rectangular space divided into four squares: Niche Players with low vision and executive ability, Visionaries with more vision but low executive ability, Challengers with higher executive ability but low vision, and Leaders with high vision and high ability to execute. It’s a quick guide to supplier choices for Gartner clients, backed up by a separate and more thorough critical capabilities report.
This year’s HCI MQ has no change in the Challengers and Leaders quadrants, the former being empty and the latter the domain of just two players: Nutanix and VMware, and their respective positions are effectively unchanged. Nutanix is maybe closer to the ideal balanced execution ability/vision line.
Down in the bottom left, we wave goodbye to DataCore, which has been ejected this year. We see Quantum inheriting the acquired Pivot3 slot and moving to the left and downwards. StorMagic, with its vSAN product, moves into the Visionaries quadrant, which pleases it mightily. An announcement said: “This is the fourth consecutive year that StorMagic has been included in the report, and we are incredibly proud to be recognised, for the first time ever, as a ‘Visionary.’”
Andrew File System developer AuriStor updated attendees at an IT Press Tour briefing about its work on the file system with an HPC and large enterprise customer base dating back 16 or more years.
AuriStorFS (a modern, licensed version of AFS) is a networked file system providing local access to files in a global namespace that has claimed higher performance, security and data integrity than public cloud-based file-sharing offerings such as Nasuni and Panzura.
AuriStor is a small and distributed organisation dedicated to expanding the popularity and cross-platform use of AuriStorFS.
AFS background
The Andrew File System (AFS) began life as the Andrew Project by Carnegie Mellon University, which was founded in a merger of the Carnegie Institute of Technology and the Mellon Institute of Industrial Research in 1900. The founders of these two institutes were Andrew Carnegie (steel industry magnate) and Andrew Mellon (banking magnate) — hence the eponymous Andrew Project.
AFS is a scalable client-server distributed file system, like NFS and SMB, with a global namespace, location independence, client-side caching, callback notification to clients of file system content changes, and replicated access to read-only data. It looks like a local file system on an AFS client. An AFS cell entity is one or more AFS servers and their clients forming an administrative domain.
OpenAFS is an open source implementation of AFS based on code made available by Pittsburgh-based, IBM-owned Transarc in 2000. Transarc was started up in 1989 by several Andrew Project members and IBM was an initial investor.
AuriStor
Jeffrey Altman.
AuriStor was founded in 2007 as Your File System, Inc., by CEO Jeffrey Altman. It is a small — you might even say minute — company, with just five employees listed on LinkedIn. However it says it has a distributed team with members in its offices in New York City, Cambridge MA, Edinburgh, Scotland, and Nova Scotia, Canada.
A briefing presentation slide showed just eight contributing developers since January 2019.
AuriStor’s aim was to accelerate AFS development by selling a licensed version of AFS, and thus fund its own engineering and support effort. Its AuriStorFS is claimed to be a better cross-platform offering than Microsoft’s Windows DFS, more reliable than Gluster, as performant as GPFS, and more cost effective than Panasas for general storage needs. AuriStor wants to sell its AuriStoFS to large, medium and small enterprises, and even individuals with smartphone client software.
AuriStorFS is backwards compatible with AFS and OpenAFS. It has been certified for Red Hat Enterprise Linux 8.4, making it the only AFS-family software certified for use on any Enterprise Linux distribution. It is validated for Debian, Ubuntu, CentOS, and Oracle and AWS public cloud use. Client support includes Linux, macOS and Windows.
The AuriStorFS source code is available to licensed organisations wanting to participate in its development as part of a private community.
Competition
Here is a competitive matrix from the AuriStor website:
It appears to be somewhat dated. It does not include, for example, Panasas, Spectrum Scale, DAOS or WekaIO.
Scalability and performance
An AuriStor cell can store up to 2259 file streams of up to 16 exabytes — the maximum allowable distributed database size. OpenAFS tops out at 2GB — a tiny fraction of that.
The number of Volume IDs per cell is 264 — much more than the 231 limit of OpenAFS. There can be 290 objects (directories or files) per volume compared to OpenAFS’s 226.
Auristor’s timestamp granularity is 100ns which compares to OpenAFS’s 1 second.
There’s no need to go on. AuriStorFS is ridiculously more scalable than OpenAFS.
AuriStor says AuriStorFS is faster than OpenAFS and performance-competitive with Lustre, GPFS (Spectrum Scale), Panasas and NFS v4.
Customers
AFS, through IBM’s support, gained HPC and large enterprise users, such as Morgan Stanley, which in 2004 had more than 25,000 hosts in 50+ sites on six continents. Morgan Stanley said AFS could provide WAN file sharing better than NFS as it had a better client:server ratio of hundreds to one compared to NFS’s then 25:1.
At that time AFS was seen as dated, but Morgan Stanley’s investment in it made it impossible for the firm to migrate away.
Morgan Stanley view of AFS in 2004.
Goldman Sachs is another AFS user, and replaced its OpenAFS client with the AuriStorFS client in 2017, deploying it to 25,000 hosts — a number which has since grown. Its experience is discussed in a 2020 YouTube video by VP Core Engineering Tracy Di Marco.
Slide from Goldman Sachs’ VP Tracy Di Marco’s presentation.
Di Marco said AuriStorFS “allows us to provide hundreds of thousands of hosts that perform business functions to the firm, at close to local disk speed, without repeated, possibly expensive, unnecessary network usage, and with a standard small disk allocation, rather than more expensive larger disks.”
Slide from Goldman Sachs’s VP Tracy Di Marco’s presentation.
Other AuriStor customers include CERN, Intel, the IRS, Qualcomm, United and KLM airlines. We couldn’t find any HPC customers and think that the number of such is low. WekaIO, for example, has never encountered AFS, OpenAFS or AuriStor in the HPC market.
Comment
This is a niche business. AuriStor is a small business exclusively focussed on its own products and supporting a relatively small number of extremely large customers. It is funded by license sales and not by venture capitalists, making it markedly different from competing file system suppliers such as WekaIO.
The main competition facing AuriStor in our view comes from the public clouds — AWS, Azure and Google — and applications such as CTERA, Egnyte, Nasuni and Panzura. Hammerspace, with its global filesystem namespace, is another competitor.
A main focus of AuriStor marketing is to encourage its adoption by OpenAFS users. A secondary focus, in our view, is to prevent defection by its user base to the public cloud.
If it succeeds in moving down market from enterprise users with 20,000-plus servers then it will increasingly come into competition with CTERA, Egnyte, Nasuni and Panzura and may suffer from appearing more complex to install and manage than these suppliers’ products. At the same time it has more bells, whistles, levers and hooks for admin staff to use, secure and optimise it.
Pricing
The starting price for an AuriStor installation is $21,000/year for a perpetual use license with four database service and four file service instances. Additional file or database service instances within a cell start at $2,500 each and decrease to $1,000 each based upon quantity. Single server cells are licensed at a base price of $6,500.
Fact Sheet
Get a 24-page downloadable AuriStor fact sheet as a PDF from here.
See Tracy Di Marco’s “Leveraging AFS Storage Systems to Ease Global Software Deployment” presentation here and the slides here.
CDS, which supplies datacentre component monitoring services, announced its Raytrix MVS Insight offering with monitoring of a broad array of server, network, and storage environments and integration with leading ITSM providers. Raytrix MVS Insight supports thousands of devices and systems from all the major OEMs including HPE, Dell EMC, Hitachi, IBM, Fujitsu, Juniper, CISCO and Brocade. The product provides comprehensive monitoring at enterprise scale, from single devices to thousands of systems. It features a holistic overview of monitored environments across systems and locations, a broad array of notification methods from SNMP to Slack, alert correlation and predictive utilisation normalisation, and flexible alert filtering and scheduling to avoid network congestion.
…
Startup HighTouch, which is developing software to enable companies to get their data out of data warehouses and into an application — a reverse to the normal ETL process — has raised $40 million in a B-round of funding. HighTouch software copies data from the data warehouse tables into a SaaS application’s tables, such as Salesforce. Data analysis is done inside the SaaS app and not the data warehouse. HighTouch integrates with BigQuery, Snowflake, and Databricks, and enterprise applications such as Anaplan, HubSpot, Mixpanel, Salesforce, Stripe, and Jira. Policies can be set up to define sync frequencies to keep the destination app up to date with data changes in the source warehouses.
…
SaaS data protector HYCU introduced a Global PACE (Partners Accelerating Cloud Environments) Program. It eliminates tiers of engagement making it easier for partners to sell HYCU services. There are separate tracks for Managed Service Providers (MSPs), Cloud Service Providers (CSPs) and Managed Security Service Providers (MSSPs) plus enhancements for reseller and distributor partners. The program features a zero conflict promise, margin assurance, minimum advertised price, and seed and premier tiers.
…
JetStreamSoftware has announced general availability of JetStream disaster recovery for Azure VMware Solution (AVS), offered through the Microsoft Azure Marketplace and sold by Microsoft sales teams. Microsoft and Jetstream say it’s the first cloud-native, disaster recovery as a service (DRaaS) for VMware-based clouds and on-premises VMware environments. An integration with NetApp enables Azure NetApp Files (ANF) to support storage expansion of larger data sets independent of compute, enabling a much faster recovery time and near-zero RTO — at a lower cost. The offering combines Azure Blob Storage, JetStream DR (VMware-certified), Azure VMware Solution, and ANF storage expansion. Microsoft and JetStream say it will radically transform the economics of disaster recovery for enterprises.
…
Andy Langsam.
Cloud storage provider Wasabi has opened a London, UK, region to expand the availability and speed of services throughout the UK. It is located in an Equinix datacentre and is Wasabi’s second European region. The first opened in Amsterdam in 2019. The London centre has high-speed network access from multiple carriers, making it easy for UK customers to connect, plus power and space expansion capabilities. Wasabi has plans to open additional storage regions in 2022.
…
Veeam appointed Andy Langsam as GM Kasten Kubernetes Business Unit and SVP of Public Cloud for Veeam in August. He was Veeam’s SVP Public Cloud & Enterprise. He is an ex-SVP Sales at SolarWinds and COO of N2SW, acquired by Veeam in January 2018. As GM of the Kasten business unit he took over from prior GM, Niraj Tolia, the CEO and co-founder of Kasten, which was acquired by Veeam in September 2020. Tolia is now SVP for strategy at the Kasten Kubernetes BU.
Datto management believes it is currently the largest pure-play backup software supplier to the MSP market and growing faster than competitors ConnectWise, Kaseya, and N-able. (Thanks to Jason Ader of William Blair.) It think it is well positioned in the long term to capitalise on secular trends toward outsourced IT via the roughly $130 billion managed service provider (MSP) channel. Datto counts more than 18,200 MSP customers (out of an estimated 125,000 MSP providers worldwide), which creates abundant opportunity for deeper penetration. Management is looking to substantially expand its security capabilities over time to fulfil its stated mission of securing all digital assets for its MSP/SMB customers.
…
Quest Software announced GA of SharePlex v0.1.2 which can replicate Oracle databases in real time to MySQL and PostgreSQL. This can be useful when creating mobile or API-based applications, supported by PostgreSQL or MySQL databases, that require data from a legacy Oracle system. SharePlex for PostgreSQL and MySQL supports AWS (RDS and Aurora) and Azure (Azure Database for PostgreSQL and Azure Database for MySQL). It also supports replication to Oracle data warehouses, SQL Server data warehouses, PostgreSQL data warehouses, Kafka, which can then feed other systems, and Event Hubs, which can then feed other systems in the Azure ecosystem.
…
A new release of Raidix’s Era software RAID, version 3.4, provides improved automatic error correcting in case of write hole (inconsistent checksum or data), file drive error monitoring, and support for multi-path NVMe drives.
…
Semiconductor technology developer Rambus has posted a blog about CXL v2.0. It states:”Server architecture — which has remained largely unaltered for decades — is now taking a revolutionary step forward to address the yottabytes of data generated by AI/ML applications. Specifically, the datacentre is shifting from a model where each server has dedicated processing and memory — as well as networking devices and accelerators — to a disaggregated ‘pooling’ paradigm that intelligently matches resources and workloads.” CXL 2.0 supports memory pooling using persistent memory and internal-to-the-host DRAM.
“By moving to a CXL 2.0 direct-connect architecture, datacentres can achieve the performance benefits of main memory expansion — and the efficiency and total cost of ownership (TCO) benefits of pooled memory.”
…
SK hynix’s desire to upgrade its DRAM manufacturing operation in Wuxi, China, by using Extreme Ultraviolet (EUV) lithography equipment to draw finer lines on wafers and thus make denser chips, has had doubt raised over it, according to Reuters. The export of EUV gear to China could fall foul of US State Department rules preventing high tech exports to China.
…
The Taipei Times reports that China’s Alibaba group will lead a consortium offering ¥50 billion ($7.8 billion) to take over the bankrupt Tsinghua Unigroup which makes semiconductor products. There are several Chinese state-backed bids to take over Tsinghua Unigroup, which is seen as important to China’s desire to be self-sufficient in making semiconductor products. It owns, for example, Yangtze Memory Technology Corp.
…
Yangtze Memory Technology Corp. 128-layer 3D NAND Chips.
DigiTimesreported Chinese NAND fabber YMTC has improved the yield rate for its 128-layer 3D NAND. Output could reach 100,000 wafers per month in the first half of 2022. That could/would mean lower NAND chip sales in China for Kioxia, Micron, Samsung, SK hynix and Western Digital.
Peter Oan - Wikipedia public domain - https://upload.wikimedia.org/wikipedia/commons/2/21/Peter_pan_1911_pipes.jpg
What is a startup? What do you call a startup that refuses to grow up and have an IPO — a Peter Pan?
Is Nebulon, a four-year-old company that has raised $14.8 million in a single VC round a startup? Obviously, yes.
Is VAST Data, a six-year-old company that has raised $263 million in four funding events, a startup? We’ll say yes.
Is Fungible, a seven-year-old company that has raised $310.9 million in three rounds a startup? We’ll say yes, again.
Is Databricks, a nine-year-old VC-funded company that has raised $3.6 billion in eight funding events, a startup? The VCs clearly expect a great exit and don’t see this company as a never-to-grow-up Peter Pan.
What of Cloudian, an eleven year-old company that has raised $173 million in six funding events? Still a startup? Well, it’s not so clear, and we’d probably say not.
And how about Delphix, a fourteen-year-old company that has raised $124 million in four funding events. Do we still call it a startup? Fourteen years is a long time to be in startup mode.
Are fifteen-year-old Egnyte and Scale Computing, with $137.5 million and $104 million raised respectively, startups? The founders are still in control. Both say they are growing. Both are VC-funded. And yet to call them startups just seems wrong — their starting-up phase is surely over.
The classic startup in the storage space is a venture capital-funded enterprise growing at a faster-than-market rate through technology development and demo, prototype product build, first product ship, to general availability and market acceptance, followed by an IPO or an acquisition.
Startup-to-IPO journey
Let’s look at three classic startup-to-IPO stories.
All-flash array supplier Pure Storage raised $470 million over six rounds and took seven years from founding to IPO. Six years later it is still going for growth over profitability and making losses.
Hyperconverged Infrastructure (HCI) vendor Nutanix took eight years and five funding events, raising $312.6 million, to reach its IPO. It’s still loss-making five years after IPO as it keeps on prioritising growth over profitability.
It took Snowflake nine years from founding to run its IPO. That effort needed seven funding rounds and $1.4 billion in raised capital.
The overall picture is eight years from startup to IPO. This is startup success. This, or being acquired — especially being acquired in a bidding war, as Data Domain. Acquisitions like this are successful but other acquisitions can be made at a discount to raised capital because the acquiree had not reached escape velocity and was a semi-successful (Actifio) or semi-failing (Datrium) company.
Escape Velocity
We can view venture capital as the rocket fuel needed to enable a startup to reach escape velocity and become a self-sustaining — meaning ultimately profitable — business.
Some startups clearly fail to reach escape velocity, such as Coho Data, which closed down six years after being founded and raising $65 million. Ditto Data Gravity, which closed down in 2018 six years after startup with $92 million raised.
A semi-success or semi-failure was Datrium, which was bought by VMware nine years after its founding with $165 million raised. Tegile was another similar company, bought by Western Digital with $178 million raised nine years after being founded
Other startups fail or semi-fail and get bought by private equity, which aims to extract value from the ruins and then sell it off. Thus PE-owned StorCentric bought seven-year-old Vexata, into which VC backers had ploughed $54 million.
Veeam was bought for $5 billion by private equity, but this was not a failure — it was a signal of its success, and an outlier.
And what of the rest? Older, VC-funded startups, such as Cloudian, Databricks and Delphix, that have not reached escape velocity?
The in-betweeners
Such companies are no longer operating in startup mode, surely? You can’t sustain that sense of developing completely new technology to change the world for nine, eleven or fourteen years. The backing VCs might still hope for a profitable exit — they surely do with Databricks. Yet these post-puberty companies, if you will forgive the expression, these near-teen and teenage companies, are in-between the startup phase and whatever their exit phase from VC funding will be.
HPEer Neil Fleming, tweeting personally, wrote: “Can we talk about what defines a startup? Databricks is eight years old, has 2K employees and turn over of $425Mm (according to Wikipedia). What’s the threshold of when something stops being a startup and starts being a business?”
Storage consultant Chris Evans identifies Peter Pan companies as the ones that never grow up. We listed pre-IPO, pre-acquisition, post-startup storage companies that could be classed as “in-betweeners” in a table, showing their age, number of funding events, and total raised:
This is far from an exhaustive list.
They are all nine or more years old, with Everspin topping this characteristic at 19 years of age. The in-between stage can clearly be a long march.
Pivots
Some of them are marked as having pivoted — meaning that they changed the marketing direction of their technology. Thus Kaminario started out as an all-flash array hardware vendor, pivoted to being an all-flash array software vendor and then pivoted again, with a rebrand to Silk, and focussed its software technology on accelerating databases in the public cloud. It is fourteen years old and still has its founding CEO in charge. He’s managed to raise $313 million in nine rounds and the VCs are still, it appears, hoping for a successful exit. Similarly, GridStore became HyperGrid, which became CloudSphere. Reduxio has become Ionir, and Primary Data evolved into Hammerspace. These companies may prefer to say that their new enterprises are fresh startups, but we can see a technology trail and a founding team link between the old and new companies.
The oldest inbetweener we have in the table is Everspin. ExaGrid is 18 years old and still hoping for an IPO. HCI vendor Pivot3 was bought by Quantum 19 years after it started up, passing through 11 eleven funding events which raised $247 million. That was not a successful exit.
Nantero, developing carbon nanotube memory, is 20 years old, with $120 million funding accumulated through eight funding events.
One of the longest-lived in-betweeners is HPC storage vendor Panasas, at 21 years, and $155 million raised in, we think, eight rounds. It may well be profitable. We should note that DataCore, although older, at 23 years old, is an outlier. It had not been a VC-backed startup for many years until it took in $30 million in 2008. The company has been profitable for ten years and very recently bought Maya Data.
It is ridiculous, we would assert, to view DataCore, Everspin, Nantero and Panasas as startups.
We think a better term is in-betweener, until a better one comes along at any rate. We won’t refer to any company more than nine years old – the Snowflake startup-to-IPO period – as a startup any more.
The second generation of Panzura’ CloudFS SaaS filesystem software, also known as Data Services, introduces two-tier licensing with complementary basic and per-user tiers, functional licensing, high-availability nodes and search across multi-vendor file storage systems.
Panzura says its global cloud file system acts like an enterprise NAS, and looks like a standard Windows file share to applications and users. The software has automated, distributed file locking and real-time data consistency for every location in a global cloud network. It is not using capacity-based licensing for its CloudFS software and has made it more enterprise-capable by adding support for highly available nodes, bringing in file scanning on non-Panzura systems, audit data retention improvements, plus a preview of automated CloudFS node configuration and management.
Edward Peters.
Panzura’s chief innovation officer, Edward Peters, said: “Now [users] can go beyond an exclusively storage-based approach and separately license access for value-added functions such as enhanced audit and search capabilities.”
The configuration and management services include faster connection status detection of CLoudFS nodes, new inventory management tables, a Dataset View of a customer’s data sources for a CloudFS deployment, and a Node View feature to access an inventory of all CloudFS nodes.
Basic tier
The Basic tier allows file access and provides:
Inventory of CloudFS components;
Summary of component operational status;
Monitoring and reports on metrics for the CloudFS network;
Configurable alerts if storage, system and cloud thresholds are exceeded — quotas can be set — and may require attention;
List of plug-ins for nodes registered on Data Services.
Disaster recovery capabilities enable the instantaneous rollback of files to a previous, granular state prior to accidental deletion or file corruption by ransomware or other malware exploits, with no data loss.
Licensed tier and functions
This is a tier of licensable services with separate licensing for Search, Audit and Audit Retention.
The Search license is provides up-to-the-second results and insight into the entire global file system including recovery services, quota services and data analytics.
This global function enables search and data analysis across all connected file systems, including Panzura and other compatible SMB and NFS file shares, for files, directories and snapshots. Data Services ingests the metadata of files and directories from connected file shares, including CloudFS, NetApp, and Dell EMC Isilon according to a periodic schedule, and indexes them into a searchable database.
The Audit license allows users to track and audit file opens, changes and access, find renamed or deleted files and recover them with a single click.
Audit functionality is provided by ingesting file logs in near real-time, providing a one-click view into user actions taken on files such as read, write, delete, move or update permissions.
The Audit Retention license provides for 180 days of data-log retention — double the prior amount. Multiple licenses can be purchased to cover longer data-retention periods.
Comment
In September, Panzura said it had been officially refounded, almost a year and a half since it was bought by private equity and given a new CEO. Refounding means, we think, that a founders-type mindset has been reimposed on the company to get it growing again and to get fresh employee commitment. The website was given a new look and the company gave itself a fresh new logo as well to mark a break from the past. Behold:
New Panzura logo.
Let’s see if Panzura gets growing again and becomes a file collaboration and sharing company to be reckoned with. Download Panzura white papers from here to get more details of its software’s capabilities.
AT SC’21 Dell EMC announced new and updated Dell Technologies Validated Designs available for Siemens Simcenter Star CCM+. They provide multi-physics CFD simulation on the latest Dell EMC PowerEdge servers with PowerSwitch and InfiniBand networking, and options for HPC storage. A new Dell EMC PowerSwitch S5448F-ON top-of-rack switch has higher density with 100/400Gbit/sec speeds, and simplifies complex network design, deployment and maintenance when running the Dell EMC SmartFabric OS10 with integrated SmartFabric Services, or CloudIQ for cloud-based monitoring, machine learning and predictive analytics.
…
MemVerge (in-memory computing with virtualised memory) has a SpotOn Early Access Program. It asks: “Do you have non-fault-tolerant and/or long-running workloads, and would like to run them on much lower cost AWS Spot Instances? Then join our SpotOn early access program to mitigate the risk of terminations which can save you a ton of heartache and money.”
…
Micron has shipped the first batch of samples of LPDDR5X memory built on its first-to-market 1α (1-alpha) node. It’s designed for high-end and flagship smartphones, powered by artificial intelligence (AI) and for 5G. Micron has validated samples supporting data rates up to 7.5Gbit/sec, with samples supporting data rates up to 8.533Gbit/sec to follow. Peak LPDDR5X speeds of 8.533Gbit/sec deliver up to 33 per cent faster performance than previous-generation LPDDR5.
…
AKQUINET, MT-C and the Fenix consortium have teamed-up to enhance the multifunctional Data Mover NODEUM to provide a data mover service within the Fenix infrastructure. This service will allow users to seamlessly manage data stored in high-performance parallel file systems as well as on-premise cloud data repositories. There are different Fenix locations, including the European supercomputing centres JSC (Germany), BSC (Spain), CINECA (Italy), CEA (France) and CSCS (Switzerland). Federated cloud object stores are used at these locations, which enable researchers to exchange their data. With the Data Mover, researchers can copy their data locally to the existing parallel file systems in order to process them on the fast HPC systems. They can also copy data to cloud object stores to make them accessible to other researchers globally.
…
OWC announced the OWC miniStack STX, the world’s first Thunderbolt 4-certified storage and hub expansion box that stacks with the Mac mini. It is also a Plug and Play expansion companion for Thunderbolt or USB-equipped Macs, PCs, iPads, Chromebooks, and Android tablets. It has three Thunderbolt 4 (USB-C compatible) ports, a universal HDD/SSD bay and an NVMe M.2 SSD slot to provide storage capacity expansion and can be combined in a RAID 1 configuration. There is up to 770MB/sec of storage performance, and the OWC miniStack STX is good for bandwidth-intensive video editing, photography, audio, virtual machines, and everyday data backup and access tasks.
OWC miniStack STX.
…
NAND Flash controller maker Phison announced its consolidated revenue in October 2021 ($216M) was the highest single month in the company’s history, and cumulative revenue for the first ten months of 2021 was also a record high for any ten-month period ($1.848B). It has increased its head count to more than 3000 globally, including expanded Enterprise SSD Engineering Lab in Colorado. Phison recently signed a ten-year green power purchase agreement to minimize its environmental impact.
…
PoINT Data Replicator 2.0 has been announced with S3 Bucket Notification. Version 2.0 is based on GUI independent services. Several replication jobs can be run in parallel and independently. Jobs can be triggered manually; alternatively, execution is controlled by a newly integrated scheduler. With the automated S3 bucket notification PoINT Data Replicator can be used to execute backups of object storage systems. With this function, only new objects are identified and replicated, so that the replication process is much more efficient. The regular jobs then perform a full data synchronization for control purposes.
My mental data and storage classification landscape has been getting confused and briefings with suppliers made it obvious a refresh was needed.
Going back to the basic block-file-object data types I had a landscape clear-out and then tidied everything up — notably in the mission-critical-means-more-than-block area. See if you agree with all this.
Update. Suppliers and storage categories diagram updated to remove VAST DAta from block category and add FlashArray//C. 25 Oct 2020.
Block storage is for structured — meaning relational or traditional database — data, and is regarded as tier 1, primary data, and also as mission-critical data. If the ERP database stops then the business stops. Typically such data will be stored in all-flash arrays attached to mainframes — big iron — or powerful x86 servers.
Illustrating this diagrammatically started out being simple, on the left, but then became more complex as we moved rightwards:
File and object data is regarded as unstructured data, also often as secondary data — which it may be, but that is not the whole story. Because file data certainly, and object data also, can be mission-critical. It can be primary data and will be put in all-flash arrays as well.
It’s different from block-based primary data where the applications are all database-centric.
Primary file and object data storage tends to be used in complex application workflows where applications work in sequence to process the data towards a desired end result. Think entertainment and media with a sequence of steps from video capture through special effects stages with layers of effects built up, through editing to production. The same set of files is being used, added to, and extended in a concurrent, real-time fashion needing low latency and high throughput.
The same is true in a hospital medical imaging system, though with fewer stages. There is image capture which can involve many and large scan files, analysis and then comparison with potentially millions of other files to find matches, and follow treatment and outcomes trails to find the best remedial procedure.
This needs to be done in real time, while the consultant doctor is with the patient and reviewing the captured image scans.
These both seem similar to a high-performance computing (HPC) application involving potentially millions of files with low-latency and parallel, high-throughput access by dozens of compute cores running complex software to simulate and calculate results based on the data.
We might say that primary file and object storage is needed by mission-critical applications (at the user level certainly and organisation level possibly) which have enterprise HPC characteristics. A separate use case is for fast restore of backup data.
Secondary and tertiary data
Secondary data is different. It is unstructured, file and object-based, not block, and it is much less performance-centric, both in a latency and throughput sense. The need is more for capacity than performance. This does not, therefore, need the speed of all-flash storage and we see hybrid flash/disk or all-disk arrays being used.
Tertiary data is the realm of tape, of cold storage, of low access rate archives. These can be either be on-premises in tape libraries or in the public cloud, hidden behind service abstractions such as Glacier and Glacier Deep Archive.
Suppliers
Having laid out this data and storage landscape we can match some suppliers’ products to it and see how valid the ideas are.
Let’s start with a straightforward matching exercise: Pure Storage. Its FlashArrays are for storing and accessing primary block data. Its FlashBlade products are for storing and accessing primary unstructured data and for the fast restore of backup data.
Block area extended to overlap File and Object area to reflect Pure’s FlashArray//C product.
Infinidat is another straightforward case. Its arrays store block data, hence primary data and, although predominantly disk-based, have effective DRAM caching and caching software to provide faster-than-all-flash performance. Recently it has launched an all-flash system with the same DRAM caching.
VAST Data supplies a one-trick hardware pony — this is meant in a positive way — an all-flash system with persistent memory used to hold metadata, and compressed and deduplicated data stored in QLC flash. This so-called Universal Storage is presented as a single store for all use-cases except primary block data.
NetApp supplies all-flash FAS for primary data, both file and block, hybrid FAS systems for secondary unstructured file data, and StorageGRID systems for object data, both primary (all-flash StorageGRID) and secondary.
A look at Dell’s products suggests that PowerMax is for primary block data, PowerStore for primary block and file data and secondary file data, and PowerScale for primary and secondary unstructured file data. ECS is for object data.
We can position Qumulo too. It’s for primary and secondary file data, with WekaIO targeting primary file data with its high-performance file software.
The two main take-aways from this exercise are that, first, unstructured data refers to both primary and secondary data storage, and second, that primary unstructured data is being used in enterprise HPC applications across a variety of vertical markets.
