Home Blog Page 97

Dell revenues droop but hint at server and storage recovery ahead

By
Chris Mellor
-

Dell Q3 revenues declined overall but servers showed a quarter-on-quarter upturn as customers bought more AI-optimized compute. The business is now hoping for an AI-boosted recovery next year.

Revenues in the quarter ended November 3 were $22.3 billion, 10 percent lower year-on-year (Y/Y) with a 317 percent increase in net income to $1 billion.

Dell CFO Yvonne McGill stated: “We have proven our ability to generate strong cash flow through profitability and working capital efficiency, including $9.9 billion of cash flow from operations over the last 12 months.”

The company ended the quarter with remaining performance obligations of $39 billion, recurring revenue of $5.6 billion, up 4 percent Y/Y, and deferred revenue of $29.1 billion, up 7 percent Y/Y, primarily due to increases in software and hardware maintenance agreements.

Financial summary

  • Gross margin: 23.1 percent flat annually
  • Operating cash flow: $2.2 billion
  • Diluted EPS: $1.36
  • RPO: $39 billion
  • Cash and investments: $9.9 billion
Jeff Clarke, Dell
Jeff Clarke

The Client Solutions Group (CSG) brought in $12.3 billion, down 11 percent Y/Y. Commercial client revenue was $9.8 billion, and consumer revenue was $2.4 billion. Demand momentum slowed in the quarter. Vice chairman and COO Jeff Clarke said in the earnings call: “The result was CSG revenue was down sequentially and short of our expectations … The recovering ramp in PC demand we were expecting in Q3 has pushed out, with large enterprises and corporate customers remaining cautious with their spending.”

The Infrastructure Solutions group (ISG) recorded $8.5 billion in revenues, an 11.7 percent annual decline. But within that, servers and networking were responsible for $4.7 billion, down 9.4 percent annually but up 9 percent on the prior quarter, with the surge of interest in generative AI netting high-price server sales. 

Clarke added: “We shipped over $0.5 billion of AI-optimized servers, including our XE9680, XE9640, XE8640 and the R750 and R760xa servers. Customer demand for these AI servers nearly doubled sequentially, and demand remains well ahead of supply … The XE9680 is the fastest ramping solution in Dell history. ”

Dell revenue

Dell shipped more than $0.5 billion of AI-enhanced servers in the quarter, and Clarke said: “The pipeline for AI-optimized servers tripled quarter-over-quarter during Q3.” So elevated AI server sales should continue.

Storage didn’t benefit from the AI boost, bringing in $3.8 billion, down 14.2 percent annually and 8 percent sequentially. 

Dell revenue

The chart above shows the abrupt server and networking revenue upturn in Dell’s Q3 while storage revenues fell back sequentially.

Yvonne McGill, Dell
Yvonne McGill

McGill said there was demand growth in data protection and PowerScale, Dell’s scale-out filer product line. 

In Clarke’s view: “AI continues to dominate the technology in business conversation. Customers across the globe are turning their operations upside down to see how they can use generative AI to advance their businesses in meaningful ways.” We would expect this to drive up storage sales as more data will need to be stored for AI analysis, but it is not happening yet.

Clarke added: “We are … seeing the beginning of a traditional server rebound, and historically, storage follows a couple of quarters later.

“This eight-quarter digestion of what was built or bought … through the course of COVID has now worked its way through the system … Datacenters need to be updated, upgraded [with] additional capacity … More data is being created.” 

In the current storage supplier reporting season, Dell, HPE, and NetApp have reported declining storage revenues while Pure Storage and Nutanix have grown their businesses. If Clarke is right, all five should see strengthening demand over the next few quarters. He didn’t use the word “recovery” but did say: “We have early positive signs that there’s a changing in the demand profile of traditional servers.”

Looking forwards to the next quarter, McGill said: “Sequentially, we expect ISG revenue to be up mid-single digits, driven by sequential growth in traditional servers and seasonal growth in storage. We expect CSG revenue to be down low-single digits sequentially. We’re seeing pockets of stability in PC demand, but have yet to see a broader recovery in the PC market.”

The guidance for Dell’s next quarter is for revenues of $22 billion +/-$0.5 billion, a 12 percent annual decline at the mid-point, and producing $88.1 billion revenue for the full 2024 year, a 14 percent annual decline. Looking further ahead, McGill said: “We’re seeing signs of stability and inflection in parts of the portfolio, including traditional and AI-optimized servers. We expect revenue to return to growth next year.”

Clarke said the AI server market represents an expansion of Dell’s total addressable market (TAM): “It’s not coming at the cost of our traditional servers.” That implies that it will provide a TAM-enhancing tailwind for storage as well.

Storage and AI

We would think that would be more for file and object storage than for block (SANs). Clarke agrees, saying in answer to a question: “Is there a relationship between storage and networking with AI? Absolutely.

“These are typically clusters – small clusters, large clusters, [with] high bandwidth needed. We see it ultimately deployed out where the data is being created. Much of the data that will be created in the future is outside of the data center. Much of that data is unstructured. Much of that opportunity is really a nice tie to what we do with our own structured and object assets. And then networking is the high-speed interconnect fabric. They matter, they matter more, whether it’s Infiniband or Ultra Ethernet, those are all exciting new technologies for us.”

But analyst Simon Leopold asked: “Is AI pulling money away from storage?” Clarke reiterated his points about AI server processing pulling along unstructured data growth in its wake, and concluded: “I don’t think AI pulls away storage dollars.

“I think what we see is the effect of eight quarters of server decline has impacted storage. Customers are cautious historically as the server business rebounds and recovers. We think our experience tells us that storage lags it by about a couple of quarters. That’s what we’re expecting. We see nothing that suggests that’s different. And in the meantime, the opportunity around unstructured is immense, and we’ll continue to focus on that.”

Comment

Storage suppliers need to get high-capacity file and object stores linking at high speed to AI-munching servers. Any supplier not doing this is heading toward irrelevancy.

NetApp and AWS boost FSx for ONTAP performance

By
Chris Mellor
-

NetApp and AWS say they have given the FSx for ONTAP file system a ninefold performance boost.

FSx for ONTAP was introduced two years ago as a first-party managed file system service in AWS. It included ONTAP’s APIs, thin provisioning, FlexClone, SnapMirror, and SnapVault services. There was a primary storage tier based on EBS SSD-backed volumes, with up to 192 TiB of capacity and a tier that could expand to pebibytes. Now we have scaleout FSx for ONTAP with much more performance and primary tier capacity.

Edward Naim, GM File Storage at AWS, said: “With scale-out FSx for ONTAP file systems, customers can now bring and run their most compute-intensive ONTAP workloads to AWS without needing to change how they manage their data.”

Ronen Schwartz, SVP and GM Cloud Storage at NetApp, added: “With these new capabilities, more customers can leverage our market-leading data management and security capabilities as they look to the cloud to scale operations or quickly adapt to changing market conditions.”

The original scale-up FSx for ONTAP, using a single pair of servers in an active/passive high-availability zone, provided up to 4 GBps throughput and 160,000 IOPS from the SSD-backed primary tier. 

Scaleout FSx for ONTAP supports a single clustered system composed from two to six high-availability pairs of ONTAP file servers, each delivering up to 6 GBps throughput and 200,000 SSD-backed IOPS. A six-pair system can provide up to 36 GBps of bandwidth and 1.2 million IOPS, with sub-millisecond latencies from primary storage. Capacity (disk-based) storage latencies are in the tens of milliseconds area. Capacity in the primary (SSD) tier has risen from 192 TiB to 1 PiB.

NetApp says compute applications get a boost from the higher aggregate performance. The single petabyte-scale namespace and single storage system simplifies large dataset management. Customers can lift and shift, extend or deploy new high-performance workloads on AWS, using ONTAP’s data management features along along with AWS’s agility, scalability, and security.

AWS scaleout FSx for ONTAP will help NetApp compete more effectively with VMware’s Cloud Flex Storage in AWS. This also has primary and capacity tiers, and provides up to 400 TiB of storage capacity with 300,000 IOPS per file system.

Amazon Chief Evangelist Jeff Barr states in a blog post: “This new level of performance will allow you to bring even more of your most demanding electronic design automation (EDA), visual effects (VFX), and statistical computing workloads (to name just a few) to the cloud.” Barr’s blog has details on creating a scale-out FSx for ONTAP file system.

This use of six paired nodes, 12 in total, makes AWS scaleout FSx for ONTAP clusters smaller than on-premises ONTAP file-based clusters, which can have up to 24 nodes in total. That suggests the AWS scale-out FSx could scale out even more. We have asked NetApp why this disparity exists, and have also asked if the scale-out AWS FSx approach will be extended to NetApp’s first-party ONTAP services in the Azure and GCP clouds.

Amazon FSx for NetApp ONTAP scale-out capabilities are now available in the following AWS regions: US East (Ohio, North Virginia); US West (Oregon); Europe (Ireland); and Asia Pacific (Sydney).

VAST Data claims vast cost-saving over Elastic File Service in Amazon’s cloud

By
Chris Mellor
-

VAST Data has released v5.0 of its VAST Data platform software and claims customers can save up to 80 percent of the cost of Amazon’s EFS by using VAST’s software in AWS instead.

This software powers VAST’s disaggregated storage filers on-premises with separate scale-out controller nodes accessing scale-out storage nodes, built with QLC flash drives across an NVMe fabric. File metadata is stored in storage-class memory class drives to speed access. The vendor has added software to this single-tier storage, including a data catalog and a combined transactional and analytical database. Data Engine software is being developed so that VAST can provide an AI-capable, so-called thinking machine, capable of generating insights on its own from data.

Jeff Denworth, co-founder of VAST Data, issued a statement: “Today we’re enabling customers to quickly deploy VAST across AWS regions and establish inter-cluster relationships for data availability – making it that much easier for cloud-native organizations to take advantage of the VAST Data Platform.”

In his view, “The VAST Data Platform is the simplest, fastest, most efficient and most secure way for customers to achieve AI-powered insights from all of their data, across private and public clouds.” Well, AWS for starters, but Azure and GCP are sure to follow.

The v5.0 release feature set includes:

  • Multi-Protocol Namespace on AWS – Combines multiple protocols (SMB, NFS, S3, and DataBase tables) into a single namespace, allowing users to read and write data from any protocol and eliminating the data silos and unnecessary copies caused by independent data systems.
  • Multi-Cluster Manager for AWS Instances – UI for managing and deploying VAST clusters across private clouds and AWS, streamlining cloud bursting workflows and multi-region availability. Customers can transparently access data for AI workflows across the globe. 
  • Intelligent Anomaly Detection – Using machine learning, every VAST cluster is now able to report on anomalous user events and data behaviors to administrators via the VAST Data Uplink call home service. This allows customers to detect ransomware attacks, anomalous API calls, as well as other unusual data access patterns in VAST clusters both on-premises and running in the cloud.
  • Real-Time Audit Log Analytics – Audit logs can be directly streamed into, and queried from, the VAST DataBase, providing security teams with real-time insight into incidents without requiring external Security Event and Incident Management (SEIM) systems.
  • New Acceleration for the VAST DataBase – New VAST DataBase functionality such as projections and additional predicate pushdown, allowing the system to reduce query times.

An IT Market Strategy report by Merv Adrian envisages an AWS customer with 300TB of EFS data which costs $90,000/month. Using VAST v5 software instead, with 3:1 data reduction, the monthly AWS cost ($13,089) and VAST software cost ($6,965) is $20,054. 

The file services area in AWS is getting crowded, with EFS competing with NetApp, Qumulo, Dell, Weka and now VAST as well – all pursuing a file fabric concept covering the on-premises and public cloud environments.

Check out VAST’s AWS microsite for more information.

Pure Storage revenues grow amid competitor declines

By
Chris Mellor
-

Unlike HPE and NetApp’s results, Pure Storage revenues grew in its latest quarter, but a massive telco order deliverable in 2025 and subscription sales growth sent its next quarter guidance down.

Charlie Giancarlo, Pure Storage
Charlie Giancarlo

Revenues in its third fiscal 2024 quarter, ended November 5, were $762.8 million, up 13 percent year-on-year (Y/Y) with a profit of $70.4 million, the third profitable quarter in its history, and contrasting with the year-ago $787,000 loss. Product sales of $453.2 million were 5.1 percent up Y/Y, affected by the higher 26 percent growth in subscription services to $309.6 million. There was a consistent macroeconomic climate during the quarter and this is expected to continue.

CEO Charlie Giancarlo said in prepared remarks: “We are pleased with Pure’s Q3 financial results. We saw healthy demand for our portfolio throughout the quarter … Evergreen//One, our storage-as-a-service consumption offering, saw continued, extraordinary growth, more than doubling year-over-year.”

There’s downside to this high subscription sales growth: lower product purchases in the next few quarters. Giancarlo pointed out: “The outperformance of Evergreen//One this year has been significantly above our prior expectations, and we now expect this strong level of demand to continue through Q4. While this success is a long anticipated and welcome expansion of our business model, its over-performance will have an effect on near-term revenue.”

Pure revenues

Financial summary

  • Gross margin: 74 percent
  • Operating cash flow: $158 million
  • Free cash flow: $113 million
  • Subscription ARR: $1.3 billion, up 26 percent Y/Y
  • Remaining Performance Obligations: $2.1 billion, up 30 percent Y/Y
  • Total cash and marketable securities: $1.35 billion

Pure reported a number of AI wins in the quarter across its portfolio, which included Portworx (container software) and FlashBlade sales. FlashBlade experienced record sales, helped by the  QLC flash capacity-enhanced FlashBlade//E product shipping. Giancarlo said: “The early success of our //E family reinforces our belief that Flash is replacing disk, enlarging Pure’s opportunity in enterprise and eventually cloud storage.”

He elaborated on this in the earnings call: “We intend to be very aggressive in pricing and especially as we penetrate the lower price performance tiers of the disk market. It’s a new market for flash and the opportunity there is huge, more than half of the overall enterprise storage market, and we believe we have a special opportunity that ahead of every other player in the market. So our intention is the E family grows is to be very aggressive in that segment.”

CFO Kevan Krysler added: “We will be aggressive in our disk takeout strategy with our E family price performance solutions.”

Pure’s all-flash array (AFA) product line grows revenues by taking business away from competitors’ hybrid disk drive arrays. Both HPE and NetApp have large installed bases of such arrays and much of their AFA business lies in converting them to all-flash arrays and not so much in gaining new customers. Pure, with no installed disk array base, is hungrier for new customers to fuel its growth, and gained 353 in the quarter, taking the total to nearly 13,000.

Pure only sells flash storage and its AFA run rate, based in this quarter, is $3.1 billion. NetApp reported a $3.2 billion AFA run rate yesterday.

 

Having overtaken NetApp’s AFA run rate a couple of quarters ago but then fallen back, can Pure overtake NetApp in AFA sales and stay ahead? In the short term, it does not look like it as Pure’s guidance for its next quarter is $782 million, 3.5 percent lower than a year ago. Its guidance for the full fiscal 2023 is $2.82 billion, a 2.5 percent Y/Y rise. 

This low guidance was explained by Krysler: “During Q3 we closed a large non-cancelable product order with a large telco customer totaling $41 million that is included in the RPO balance at the end of Q3. Based on the timing of the shipment schedule for this order, product revenue is not expected until next year which impacts our revenue outlook this year.”

The order is for a 5G system and Pure’s product will ship in calendar 2024.

This, together with product sales being replaced by subscription sales, “creates a short-term impact on revenue growth,” Krysler said. He reckons annual revenue growth would be 7 percent, and not 2.5 percent, if the higher-than-expected shift to subscription sales and telco order are taken out of the equation.

The stock market reflected some short-term investor disquiet about this, with Pure’s stock price falling 14 percent in after hours trading.

Nutanix hits record revenues amid evolving market

By
Chris Mellor
-

Hyperconverged infrastructure software vendor Nutanix announced record revenues in its latest quarter, its eighth of successive growth, and hit a $2 billion annual revenue run rate for the first time.

Revenues in the quarter ending October 31 were $511.1 million, up 18 percent on a year ago, with a loss of $15.9 million, significantly lower than the year-ago $99.1 million loss. It gained 380 new customers, taking its cumulative total to 24,930. Nutanix also began a share repurchase program to bolster its stock price.

Rajiv Ramaswami
Rajiv Ramaswami

President and CEO Rajiv Ramaswami said: “We delivered a solid first quarter with results that came in ahead of our guidance. The uncertain macro backdrop that we saw in our first quarter was largely unchanged.”

CFO Rukmini Sivaraman added: “Q1 ’24 was a good quarter in which we exceeded the high end of the range on all guided metrics. ACV billings in Q1 was $287 million, above the guided range of $260 million to $270 million and a year-over-year growth rate of 24 percent.”

Performance in its US Federal business was stronger than expected, with a sale of its August-launched GPT-in-a-box system. The partnership with Cisco, which recently canned its HyperFlex HCI product in favor of selling Nutanix, is also off to a good start. Nutanix “secured a few wins for this new offering, which were conversions of customers, who had previously been planning to purchase Cisco’s HyperFlex,” Ramaswami said.

But he tempered future prospects by saying in the earnings call: “Our pipeline takes six to nine months to go close deals for us, right? So we don’t expect a huge amount of Cisco this year.”

Although the macro environment didn’t change from the last quarter, Sivaraman said: “We saw a modest elongation of average sales cycles, relative to the year-ago quarter.”

Nutanix revenues

Financial summary

  • Gross margin: 84 percent vs 81 percent a year ago
  • Free cash flow: $132.5 million vs $45.8 million a year ago
  • Operating cash flow: $145.5 million vs $65.5 million a year ago
  • Cash and short-term investments: $1.57 billion vs $1.39 billion a year ago

Ramaswami is hoping for “an ARR compound annual growth rate of approximately 20 percent through fiscal year ’27 and generation of $700 million to $900 million of free cash flow in fiscal year ’27.”

The outlook for Nutanix’s second 2024 quarter is for $550 million in revenues +/- $5 million, 13.1 percent growth annually. Guidance for the full year is now $2.11 billion +/-$15 million, up 13.3 percent on last year at the mid-point. Sivaraman said: “We are seeing continued new and expansion opportunities for our solutions, despite the uncertain macro environment.” The fiscal 2024 guidance has been raised for all guided metrics, including revenue, ACV billings, non-GAAP operating margin, non-GAAP gross margin, and free cash flow.

One of these opportunities is for Red Hat OpenShift to run on top of Nutanix. Ramaswami said Red Hat competes with VMware on the application side and Nutanix infrastructure also competes with VMware so it is a natural partnership. Several Nutanix customers are running OpenShift, including Global 2000 banks. One G2K bank win in the Asia-Pacific region chose Nutanix because of uncertainties about VMware’s future as part of Broadcom. Other Nutanix deals this quarter also benefited from that uncertainty.

But the opportunity is complicated as Nutanix can run on top of the VMware hypervisor. Also, some VMware customers have signed long-term deals and cannot shift across easily to Nutanix, even if they wanted to. 

Ramaswami mentioned this in the earnings call, saying: “We have a significant pipeline of opportunities, and it’s growing and a good degree of engagement with prospects, driven by these concerns. It’s just difficult to predict timing and magnitude of wins.”

In general, Nutanix is benefiting from weakened competition from VMware and Cisco’s withdrawal from HCI. These two opportunities will provide a positive growth environment for Nutanix in the next few quarters if not years.

HPE fails to crack $8 billion quarter barrier

By
Chris Mellor
-

HPE managed to grind out a small annual revenue increase in fiscal 2023 despite disappointing final quarter results.

Revenues in the quarter ended October 31 were $7.35 billion, down 6.6 percent annually, with a profit of $642 million, compared to the year-ago $304 million loss. Full fiscal 2023 revenues were $29.14 billion, 2 percent up on 2022’s $28.5 billion, with profits 133 percent higher at $2.03 billion. HPE raised its annual dividend in consequence.

President and CEO Antonio Neri stated: “In fiscal year 2023, HPE clearly demonstrated that our strategic investments and extraordinary innovation across the growth areas of Edge, Hybrid Cloud, and AI are resonating with customers.””

The Intelligent Edge (Aruba) and HPC &AI (Cray) business segments delivered the goods as the Compute and Storage segments fell short, as numbers and a chart indicate:

HPE revenues
HPE revenues

If we chart the same segment results by fiscal year, the results of Neri’s bets on the Edge and HPC/AI markets become clearer: 

HPE revenues

The Intelligent Edge and HPC & AI segments are the only growth business units over the fiscal 2019-2023 period. Compute revenues are so large, though, that the rise in Edge and HPC/AI revenues have not yet compensated for the decrease in compute, leaving HPE quarterly revenues coasting along in the $6.5-8 billion area for 24 quarters in a row – level pegging almost, as a third chart makes clear:

HPE revenues

Unless HPE stops its compute revenue decline, or Edge and HPC/AI revenues surge higher, it will stay at this quarterly revenue level, unable to break out. A halt to its storage revenue decline would be a good thing as well but storage revenues are dwarfed by compute, where HPE needs to take share from Dell and Lenovo if it is to make progress and bust the $8 billion/quarter barrier.

Neri added: “We delivered record performance against key financial metrics this year. Our steady execution resulted in higher revenue, further margin expansion, larger operating profit, and record-breaking non-GAAP diluted net earnings per share and free cash flow.”

Quarterly Financial Summary

  • Gross margin: 34.8 percent up 1.95 percent annually
  • Free cash flow: $2.2 billion
  • Operating cash flow: $2.8 billion, up $326 million annually
  • Greenlake ARR: $1.3 billion

Wells Fargo analyst Aaron Rakers told subscribers: “We think HPE delivered solid results/guide, and more importantly highlighted continued momentum in AI orders, which we think should overshadow incremental Intelligent Edge weakness and uncertainty over a Compute recovery.”

Regarding the storage business segment, Rakers said HPE “noted that storage demand has been flat to up for three straight quarters, but highlighted Alletra growth of over 50 percent year-over-year as a pocket of strength.”

The outlook for the next quarter is for revenues of $6.9 billion to $7.3 billion, representing a 10.4 percent year-on-year decline at the $7 billion mid-point.

AWS radically speeds up small S3 object access with new tier

By
Chris Mellor
-

AWS has introduced an S3 Express One Zone tier (S3 1XZ) to speed up small object access latency by up to 10x.

The announcement was made at AWS re:Invent 2023, and S3 1XZ co-locates AWS EC2, ECS, and ELS compute in the same availability zone as the storage, reducing the latency of compute to object storage access. AWS Chief Evangelist Jeff Barr blogs that latency to the first byte is in the single-digit milliseconds area compared to S3 standard, whose latency is described as milliseconds, and which we can now deduce to be double-digit milliseconds.

Jeff Barr

Barr makes the point that time to first byte is pretty close to time for last byte for small objects, unlike with large objects where data streaming time lengthens with object size. So S3 1XZ is, he says, ideal for data-intensive applications with hundreds or thousands of parallel compute nodes accessing small objects, such as AI/ML training, financial modeling, media processing, real-time ad placement, and high performance computing. It can deliver a significant reduction in runtime compared to S3 standard class.

No actual performance benchmarks have been supplied by Amazon.

The pricing is the usual Amazon maze of numbers, and based on capacity and activity. S3 1XZ is almost seven times more expensive per GB per month at $0.16 than S3 Standard’s $0.023 for the first 50 TB/month, and even higher than the next 450 TB’s $0.022/month and the $0.0021/month AWS charges for more than 500 TB.

S3 standard charges $0.005 per 1,000 PUT, COPY, POST, and LIST requests, and $0.0004 per 1,000 for GET, SELECT, and all other requests. S3 1XZ halves these charges to $0.0025 and $0.0002 respectively.

Barr makes the point that S3 1XZ’s lower latency and request costing “means that your Spot and On-Demand compute resources are used more efficiently and can be shut down earlier, leading to an overall reduction in processing costs.”

The new S3 class is implemented by having a new directory bucket type, a new bucket naming method, and new authentication model. Barr’s blog has coded implementation examples.

Amazon S3 Express One Zone is available in the US East (N Virginia), US West (Oregon), Asia Pacific (Tokyo), and Europe (Stockholm) Regions, with plans to expand to others over time.

NetApp’s Q2: Down we go again but things are getting better

By
Chris Mellor
-

Revenues in NetApp’s second quarter of fiscal 2024, which ended on October 27, were $1.56 billion, at the mid-point of its guidance and 6.1 percent down year-on-year. The $233 million profit was 69 percent lower than a year ago.

CEO George Kurian’s statement said: “Q2 improved on our solid start to FY’24 in what continues to be a challenging macroeconomic environment. We delivered another strong quarter, with revenue above the midpoint of our guidance and all-time highs for gross margins, operating margins, and EPS.”

Revenues were down year-on-year for the fourth quarter in a row, however. CFO Mike Berry said: “Q2 was a very solid quarter in what continues to be a challenging macro environment with soft IT spend.” 

Chart showing the four most recent quarters have been declines

Financial summary

  • Billings: $1.5 billion, down 9 percent y/y
  • Gross margin: 72 percent, up 0.58 percent y/y and a record
  • Operating cash flow: $135 million compared to $214 million a year ago
  • Cash, cash equivalents & investments: $2.62 billion
George Kurian

NetApp reported Hybrid cloud revenues of $1.41 billion, down 7.8 percent year-on-year, while public cloud revenues rose 8.4 percent to $154 million. Product sales of $706 million were 15.7 percent down annually but up 19.7 percent quarter on quarter as all-flash array sales increased, helped by lower cost models. Kurian said: “Our all-flash array business benefited from growth of the AFF C-series.”

The all-flash array ARR was up 3.2 percent to $3.2 billion.

Kurian added: ”We remain relentlessly focused on managing the elements within our control while driving better performance in our storage business and building a more focused approach to our Public Cloud business. We are seeing positive results from these actions, with increased profitability and a stronger position for delivering long-term growth.”

For the public cloud that meant a strategic review in the quarter. Kurian said: “As a result, we will continue to prioritize cloud storage offerings delivered through the hyperscalers, while refocusing some services, such as Cloud Insights and Instaclustr, to complement and extend our hybrid cloud storage offerings, creating greater differentiation and additional value for customers.”

“We will integrate other services that are sold as standalone subscriptions today, such as data protection, into the core functionality of Cloud Volumes. We will also carefully manage the transition of cloud storage subscription services to align to customer preference for consumption offerings.”

“And, we have decided to exit the SaaS backup and virtual desktop services.”

A lot of attention is being paid to growing public cloud revenues but they will have to grow at a high rate to offset declining hybrid cloud revenues, as the chart above shows.

The outlook for the third quarter is for revenues of $1.59 billion +/- $0.08 billion, representing a 3.9 percent increase at the mid-point and – potentially – growth at last. Kurian said: ”We expect the momentum we saw in Q2 to continue through FY’24, despite continued softness in the demand environment due to the challenging macro.” 

Berry said: “We expect our continued focus and discipline to deliver year-over-year growth in the second half of the year. As a result, we are raising all our guidance measures for fy’24.”

Full fy2024 revenues are now forecast to be 2 percent down on fy2023 at $6.23 billion.

Nvidia launches RAG-enhanced service to reduce AI errors

By
Chris Mellor
-

Nvidia has unveiled a RAG-based Retriever service to stop generative AI hallucinating inaccurate responses.

The NeMo Retriever is a generative AI microservice that enables enterprises to connect custom large language models to enterprise data, thereby delivering accurate responses to an AI application user’s requests and questions. It essentially involves semantic retrieval, with retrieval-augmented generation (RAG) capabilities that utilize Nvidia algorithms. It is part of Nvidia’s AI Enterprise software platform and is available in the AWS Marketplace.

Jensen Huang, Nvidia co-founder and CEO, stated: “Generative AI applications with RAG capabilities are the next killer app of the enterprise. With Nvidia NeMo Retriever, developers can create customized generative AI chatbots, copilots and summarization tools that can access their business data to transform productivity with accurate and valuable generative AI intelligence.”

Nvidia graphic

NeMo Retriever supports production-ready generative AI with commercially viable models, API stability, security patches, and enterprise support, unlike open source RAG toolkits. It contains embedding models that capture relationships between words, enabling LLMs (large language models, the technology underlying generative AI) to process and analyze textual data.

Nvidia claims that enterprises can connect their LLMs to multiple data sources – such as text, PDFs, images, videos, and knowledge bases – so that users can interact with data and receive accurate, up-to-date answers using simple, conversational prompts.

It asserts that enterprises can use NeMo Retriever to achieve more accurate results with less training, speeding up time to market and supporting energy efficiency in the development of generative AI applications.

Cadence (industrial electronics design), Dropbox, SAP, and ServiceNow are collaborating with Nvidia to integrate production-ready RAG capabilities into their custom generative AI applications and services.

Preventing LLM hallucinations

Retrieval-augmented generation is a way of minimizing hallucinatory generative AI chatbot LLM responses from underlying LLMs trained on a general trove of information rather than domain-specific data. It works by turning a user’s request into a set of coded parameters called vector embeddings. These are then used to query a domain-specific database of information local or specific to the user’s organization. This database contains vector embeddings of data relevant to the organization and the RAG query looks for similarities between the input request vector embeddings and the database’s vector embedded content, choosing the content with the closest similarity.

Minimising hallucination isn’t the same as eliminating it, though. The RAG process requires verification to ensure the output is contextually relevant, grounded, and accurate. Vectara has developed an open source Hallucination Evaluation Model to assist in this verification.

Storage news ticker – November 28

By
Chris Mellor
-

Cribl, which supplies a data engine for IT and security, announced that its Cribl Edge scalable edge-based data collection system is available through AWS Marketplace as an Amazon Elastic Kubernetes Service (EKS) add-on.

…

Couchbase announced a Capella columnar service on AWS at AWS re:Invent 2023, enabling organizations to harness real-time analytics. Capella columnar introduces a columnar store and data integration into the Capella Database-as-a-Service (DBaaS), allowing for real-time data analysis on the same platform as operational workloads.

…

Druva has been named AWS Global storage partner of the year at AWS re:Invent 2023.

…

Google admits there is a data loss issue with Google Drive for desktop via a statement on its support website: “Drive for desktop (v84.0.0.0 – 84.0.4.0) Sync Issue.”

According to the statement, “We’re investigating reports of an issue impacting a limited subset of Drive for desktop users and will follow up with more updates.”

Meanwhile, the search giant advises users: “Do not click ‘Disconnect account’ within Drive for desktop; Do not delete or move the app data folder.”

For Windows users, that’s %USERPROFILE%\AppData\Local\Google\DriveFS or for macOS users ~/Library/Application Support/Google/DriveFS

A bit of optional advice: “If you have room on your hard drive, we recommend making a copy of the app data folder.” 

More information can be found on the Drive for Desktop site.

…. 

Hitachi Vantara announced that Turkish information and communications technology provider Türk Telekom has deployed four NVNE drive-based Hitachi Virtual Storage Platform (VSP) 5000 systems. It is more environmentally sustainable than the previous hybrid disk/flash storage, reducing Türk Telekom’s total footprint from 23 cabinets to nine and decreasing power and cooling requirements from the replaced cabinets by approximately 60 percent. The upgrade has also led to a 30 percent reduction in total cost of ownership. Türk Telekom serves more than 15 million broadband and 25.6 million mobile subscribers. Türk Telekom’s storage infrastructure includes hardware from a variety of vendors but it chose long-term partner Hitachi Vantara for most of its critical storage environments. 

…

Informatica is doing more with AWS and has announced its AI-powered Intelligent Data Management Cloud (IDMC) – catalog, integration/engineering, API/App integration, quality/observability, master data management/360 applications, governance/privacy and marketplace – the platform supports Amazon’s Bedrock generative AI service. Informatica has earned AWS certification for IDMC integrations with AWS HealthLake. Informatica is an Launch Partner for Amazon S3 Access Grants – an S3 access control feature that helps customers manage Amazon S3 permissions for their data lakes at scale, with detailed audit history in AWS CloudTrail for end-user access to Amazon S3.

…

Micron is guiding F1Q24 revenue approaching $4.7 billion vs the prior guide at $4.4 billion +/-$200 million. This, Wells Fargo analyst Aaron Rakers believes, “reflects a better pricing environment.” Micron expects the pricing outlook to remain positive for remainder of the year. It thinks, “2025 will be a record year for the memory industry.” It is optimistic on its HBM3e prospects with its product having ten to 15 percent better performance and 25 percent better power consumption than competing Samsung and SK hynix product.

…

The Keystone Group, the UK’s largest steel lintel manufacturer and Europe’s fastest-growing roof window manufacturer, has reduced costs, increased efficiency, and secured its data with Peer Software’s distributed file system solution, PeerGFS. The Keystone Group decided to implement PeerGFS to provide local file access of global data to engineers in remote sites such as Swadlincote and Cwmbran. The ability to efficiently replicate data and prevent version conflicts through integrated file locking across mirrored sites was a key selling point. It ensured seamless collaboration among users working on the same projects, simultaneously. The firm was able to retire its Citrix desktop infrastructure, saving budget on licensing and infrastructure support. Additionally, PeerGFS facilitated high availability and disaster recovery, minimizing downtime and reducing the risk of data loss for the entire workforce.

…

Qumulo announced Global Namespace – software that scales to exabytes anywhere unstructured data is needed, with seamless integration with AWS-based Qumulo file storage. It provides customers access to files in any location and intelligently caches the data where the performance is needed by local users. Global Namespace allows customers to choose their file infrastructure based on the location-specific business requirements – from any AWS region, core data centers, or at the edge.”

Qumulo plans to onboard select customers for a private preview of its next-gen managed file storage service on AWS. With this, customers can experience claimed unparalleled storage capabilities designed to lower the cost of cloud file storage, increase performance, and scale elastically in response to business demand.

…

Rubrik said it can further protect Amazon S3 data, to allow customers further visibility into where sensitive S3 data lives and who has access to it, while also creating recovery plans. These cyber posture features draw upon capabilities from Rubrik’s recent acquisition of Laminar.

Benefits:

  1. Autonomously discover, classify, and provide context on all known and shadow S3 data, without that data leaving the customer’s environment;
  2. Assess the security posture of sensitive data against security policies and data compliance requirements;
  3. Continuously monitor sensitive data within S3 for risky user activity or leakage and provide early warning of emerging threats;
  4. Identify and remediate redundant S3 data to help reduce cloud costs;
  5. Rapidly recover the most recent clean copy using a range of recovery patterns, including object-level and whole bucket.

Rubrik’s NAS Cloud Direct is now available in the AWS Marketplace and Rubrik also announced new but unspecified data protection features for Amazon EKS.

Rumor central – strengthened by a Bloomberg report – suggests Rubrik may be looking to run an IPO as soon as Q1 2024.

…

Seagate’s chief commercial officer,  BS Teh, predicts that in 2024:

  • Data-hungry AI will compel datacenters and enterprises toward high-density hard drive storage to future-proof their data value by saving raw data sets as well as insights produced by AI and LLM processing;
  • Significant gains in HAMR technology will enable datacenters to refresh their fleets of lower-capacity drives with higher-capacity alternatives;
  • Hard drives will continue to play a pivotal role in cloud data storage, as the value gap between hard drives and flash storage continues to hold strong in the hard disk’s favor.

Regarding the third point, he elaborated: “Latest analyses from IDC, TrendFocus, and Forward Insights confirm that hard drives will remain the most cost-effective option for most capacity-centric storage tasks. The demand for storage capacity in the cloud – where the vast majority of the world’s data resides – is only expected to increase, and hard drives will be the primary beneficiary of this exabyte growth. Hard drive storage will offer mass data storage at less than one-fifth the cost of comparable all-flash solutions on a per bit basis. Relative to datacenter architectures, the value gap will not come close to closing next year – or over the next decade.”

…

Decentralized Web3 storage provider Storj announced Storj Select – a feature that delivers customizable distributed storage solutions to meet specific security and compliance requirements of organizations with sensitive data, such as healthcare and financial institutions. With Storj Select, customer data is only stored on points of presence which meet customer specified qualifications – such as SOC2, GDPR, and HIPAA. The firm is also announcing CloudWave as a customer, a provider of healthcare data security offerings, which has chosen Storj Select to provide compliant data storage.

TrendForce anticipates diversification in HBM market

By
Chris Mellor
-
Head torch MIlky Way at night searcher

Research house TrendForce detects signs of the standardized HBM (High Bandwidth Memory) market splintering with HBM4 as suppliers and customers seek extra performance.

HBM is a set of memory die bonded to an interposer device which connects them to GPUs faster and with greater capacities than the x86 socket-connected DIMM approach can provide.

There have been three main HBM generations and a fourth is being defined. A table lists their characteristics and suppliers:

B&F table derived from TrendForce table. Grey bars equal production periods

TrendForce says Nvidia, which has used SK hynix HBM products until now, is set to diversify its supply and is evaluating HBM3e chips from Samsung and Micron. It says Micron provided 24 GB samples using its 8-hi (layers) chip to Nvidia by the end of July, SK hynix in mid-August, and Samsung in early October.

Nvidia can take up to two quarters to verify a supplier’s HBM product so preliminary results could be expected by the end of this year for the early submissions with full results by the end of 2024’s first quarter. Nvidia can then decide from which supplier to purchase products.

As AMD and Intel are also shipping GPUs, and GPU demand is being pushed up by generative AI workloads, the total HBM market is set to increase in size quite rapidly.

TrendForce believes that the fourth generation, HBM4, will launch in 2026 “with enhanced specifications and performance tailored to future products from Nvidia and other CSPs.” HBM4 will feature a 12 nm process bottommost logic die supplied by semiconductor foundries to the memory fabricators. “This advancement signifies a collaborative effort between foundries and memory suppliers for each HBM product,” the analysts say.

TrendForce thinks HBM4 is set to expand from the current 12-hi to 16-hi stacks, spurring demand for new hybrid bonding techniques. The 12-hi HBM products will launch in 2026 with 16-hi products following in 2027.

It sees an HBM customization trend emerging, moving away from an interposer connecting the memory to GPUs, and exploring options like stacking HBM directly on top of the logic SoC (GPU) as a way of gaining extra performance and perhaps lowering costs.

They say this “is expected to bring about unique design and pricing strategies, marking a departure from traditional frameworks and heralding an era of specialized production in HBM technology.”

This means that HBM customers, the GPU and other logic SoC suppliers, could be locked in to an HBM vendor because of specialized design and connectivity options. Alternatively, HBM suppliers could start selling their own HBM + logic SoC (processor) products.

Save your tiers: Wasabi supplies single-tier S3 cloud storage only

By
Chris Mellor
-

Cloud storage supplier Wasabi is a stick-to-your-core-focus business and its core focus is offering disk drive-based S3 object storage in its cloud, with the pitch being that it’s cheaper than Amazon with no egress fees.

Update: Note on Backblaze reserve capacity deals added. 29 November 2023.

Jonathan Howes

Although Wasabi is not a technology trailblazer it must be doing something right because, as EMEA VP Jonathan Howes pointed out, Wasabi is growing at twice the cloud storage market’s average growth rate, according to IDC numbers. In late 2022 it had more than 40,000 customers and Howes says it now has 60,000.

Howes and Wasabi’s VP for Cloud Strategy, David Boland, briefed us earlier this month and extended what we had learnt from an interview with Wasabi CEO and co-founder David Friend in February. Boland said Wasabi basically replicates its cloud datacenter pattern in each of its 13 regions. A set of servers and disk-based storage chassis plus software will be installed in a co-location datacenter, often one run by Equinix. It’s expanding by growing the customer base in a region and also by adding regions.

There is no wish to add additional tiers of storage, such as, for example, a tape archive tier, or faster-access SSD tier. It does supply a file access service, Wasabi Cloud NAS, but this is basically a NAS gateway running on-premises and providing file-access to files stored as objects in Wasabi’s S3 cloud repository.

It also supplies multi-tenancy and accounting software to its MSP customers so they can sell white label or Wasabi-branded cloud storage to their customers and get up and running in a short time. 

David Boland

Howes said Wasabi has qualified its software as a backup target for data protection suppliers, such as Veeam. He claimed that it’s not seeing competition from on-premises, object-based backup targets, and suggested these serve a separate market. Nor is Wasabi seeing competition from Web3, decentralized storage suppliers, it told us.

Howes agreed that Wasabi is a tier 2 cloud storage provider, in general revenue terms, with AWS, Microsoft and GCP in tier 1, and CSPs such as Backblaze and OVH Cloud in a tier3.

Wasabi charges $6.99 per TB per month pay-as-you-go storage while Backblaze charges $6.00 per TB per month, with neither charging egress fees. However Backblaze does levy an egress fee of $0.01/GB once a customer downloads more than three times the average monthly data stored amount.

Backblaze pricing.

Even though Backblaze has an upfront pay-as-you-go pricing advantage, Wasabi says it can undercut it by selling reserved capacity storage with 1, 3 and 5-year terms, giving customers substantial pricing discounts for capacity purchase commitments on this basis. Backblaze also offers reserve capacity pricing deals. Wasabi says it covers more geographic territories than Backblaze.

Howes says customers find egress charges objectionable. While allowing that Wasabi operates under an AWS/Azure and GCP price umbrella, he doesn’t think think that the big three will lower their egress prices – because they make too much money from it, and also because Wasabi and its competitors are not hurting the big three enough.

S3 cloud storage is fertile ground. Howes says Wasabi can stick tightly to its S3 knitting and continue to grow for the foreseeable future. 

1...969798...438Page 97 of 438

ABOUT US

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

FOLLOW US

© Situation Publishing, 2018-2024