Home Blog Page 255

HBF

By
Chris Mellor
-

HBF – High Bandwidth Flash. This is the stacking of NAND dies one above the other with them all above a logic layer. The whole stack connecs to a GPU. CPU, TPU or other processor server via an interposer unit, just like high-bandwidth memory (HBM). Western Digital’s Sandisk business unit publicized the idea in February 2025.

The aim is to place higher-capacity NAND in the stack layers to replace lower capacity and more expensive HBM. The whole stack has interposer-class connectivity to the server host.

Ocient tames data munching adtech monster

By
Chris Mellor
-

Customer Story. Real-time bidding to place online ads on websites like the one you are looking at now is fuelling enormous growth in stored bid records for analysis by ad spot sellers and buyers. With billions of potential page viewers, cut myriad ways by gender, age, income, location, occupation, interests, etc. and thousands, if not tens of thousands of online outlets, then analysing bid data to find the best way to price ad spots and bid for them is mind-bendingly complex.

Ad spot buyers want to maximise the effectiveness of ad campaign spending while ad spot providers want to maximise their income by optimally pricing their ad spots, such as banners, side bars, etc. according to time of day, week, month, year, etc.

Both sides need to analyse as much data as they can as quickly as they can to get the most accurate decision-making information – but they haven’t been able to look at the gold standard source material – the raw bidder log data – because there is simply too much of it.

A graphic showing building blocks in the adtech market.

This has meant either analysing an extracted subset of the source data (estimation), meaning it’s incomplete, or using a background batch-type job analysis (pre-aggregation), which is slow and incompatible with the kind of near-instant analytics needed by adtech professionals researching both the supply and demand sides of the adtech business. The third option is to develop a custom system with hardware, database software and access software components – which is expensive, time-consuming and complicated.

Ocient has built a massively scalable parallel-access relational database storage system, using NVMe SSDs, for such data, and it can be accessed using SQL. It’s sold a hyperscale campaign forecasting system to MediaMath, an adtech demand side platform (DSP – see graphic above) provider for online ad spot buyers. Its staff now have near-instantaneous (sub-5 second)access to SQL analytic runs on massive raw bid log datasets.

The stats are, well, prodigious sounds the right word, as MediaMath handles more than six million bid opportunities per second with 10–12 petabytes of new records per day for more than 3,500 advertisers. By avoiding having to write a custom system that would have otherwise taken multiple engineers 9–12 months, MediaMath with Ocient was able to save about 50 per cent in time and cost while bringing on and supporting about $100 million of new and existing business. 

Ocient MediaMath video.

MediaMath’s Ocient system features”

  • Low-latency analytics optimised for highly parallelised processing leveraging NVMe solid state hard drives with compute adjacent to storage;
  • Continuous loading and transformation of raw, semi-structured bidder log data with streaming data available for query within seconds;
  • Support for geospatial data types and complex data formats, including arrays, tuples and matrices, to handle deep arrays of audience data and bid opportunity contextual data sets;
  • ANSI SQL queries with powerful windowing, aggregation and machine learning models that process thousands of unique campaign targeting criteria;
  • Zero Copy Reliability leveraging parity coding to ensure data availability and reliability with 60–80 per cent less storage footprint than copy-based data storage and analytics; 
  • GDPR and CCPA compliance;
  • Deployed as a fully managed service in OcientCloud for ultimate flexibility and low operational overhead. 

The loading and transformation of data could reach Tbits/sec speed but that’s not guaranteed with MediaMath.

Anudit Vikram, chief product officer at MediaMath, said “Working with Ocient enabled us to get to market months faster. We left the customisation of our solution to Ocient, meaning we didn’t need ten or more of our own engineers dedicated to making it work. We can now grow our market share and pursue massive new business opportunities while reducing costs and eliminating the resource-intensive process of customising our own solution internally.”

Check out a 22-page Ocient technical backgrounder (Next Generation Campaign Forecasting) on real-time bid analysis to get a better picture of what is going on. But, if you are not an adtech professional, put aside a good hour for this. The terminology is complex and the database details difficult to understand if you are a lay person – like me. I’m more sadtech than adtech.

Storage news ticker – February 11

By
Chris Mellor
-

…

The CXL (Computer Express Link) Consortium announced the acceptance of Gen-Z Consortium’s assets and the Gen-Z specification, and will post the Gen-Z specification as an archive on CXL Consortium’s public website for five years. Upon completion of the asset transfer, the Gen-Z Consortium will begin finalizing operations. (CXL is a high-speed interconnect, based on the PCIe 5 bus, offering coherency and memory semantics using high-bandwidth, low-latency connectivity between the host processor and devices such as accelerators, memory buffers, and smart I/O devices.)

… 

More info (thanks TrendForce) about the Kioxia/WD NAND chip outage due to chemical contamination. Contamination occurred during a wet etching process. Flash production halted at the end of January to enable pipeline cleaning and production restart will be gradual from the second half of February through to mid-March. More than three weeks’ worth of output has been lost. All contaminant-affected wafers must be scrapped. In our view there will be a horrendously large bill at the end of the cleanup and the luckless chemical supply company faces having to pay it.

…

Lightbits Labs, an NVMe overTCP pioneer, has appointed Thomas DeLorenzo as its VP of sales east. His 30+ year CV includes stints at Turing Video, CloudSimple, Infinidat, Brocade Communications, and Veritas Software. He’s been involved in developing sales strategies and go-to-market modelling for Vexata, FusionIO, Nimble, Data Domain, and CLARiiON.

…

In an AWS environment Teradata proved that it can successfully operationalise analytics at scale on a single system of more than 1,000 nodes with 1,023 active users submitting thousands of concurrent queries, using a diverse set of mixed workloads, and with no system downtime or outages. This single system test was executed over an extended period of several weeks on a distributed system consisting of over 1,000 servers with zero system downtime. It ran mixed workloads. Teradata’s analytic data platform featured multi-compute clusters, automated elasticity, low-cost object storage and push button provisioning.

…

DNA storage silicon supplier Twist Bioscience announced $250 million of common stock in an upsized, underwritten public offering of 4,545,454 shares at a price to the public of $55.00 per share. The offering is expected to close on or about February 15, 2022. Twist intends to use the net proceeds of the offering, along with its existing cash, cash equivalents and short-term investments to scale its investment in its R&D organization, which includes investing in pharmaceutical biologics drug discovery and in DNA data storage, to increase its investment in its commercial organization to support the growth of its NGS, synbio, pharmaceutical biologics drug discovery programs and its global expansion, to scale its NGS operations and to expand its capacity, and for the remainder to fund working capital and general corporate purposes. 

…

Pure’s Portworx gets AWS cloud-native blessing

By
Chris Mellor
-
070403-N-6674H-069 PEARL HARBOR, Hawaii (April 03, 2007) - Kuhu Kaleo Patterson performs a traditional Hawaiian blessing as Commander, Navy Region Hawaii (CNRH) Rear Adm. T. G. Alexander, center, joined on his left by Rep. Neil Abercrombie, and Sen. Daniel K. Akaka, along with representatives of Forest City Military Communities and CNRH use Hawaiian o-o sticks during a military housing groundbreaking ceremony on Ford Island. The groundbreaking represents phase-two of a joint venture between CNRH and Forest City Military Communities Hawaii to construct more than 800 new single-family homes on Ford Island. U.S. Navy photo by Mass Communication Specialist Seaman Paul D. Honnick (RELEASED)

Pure Storage and AWS have a 3-year deal to use Portworx container management software with Amazon’s Elastic Kubernetes Service (EKS) and so provide enterprise scale containerised apps in Amazon’s cloudy IT empire.

EKS provides a basic capability to run and scale Kubernetes on AWS with no need to install, operate, and maintain a Kubernetes control plane or nodes. Pure’s Portworx software is layered on top of that to add services such as persistent storage, data protection, disaster recovery, data security, cross-region and hybrid data migrations, and automated capacity management.

Deepak Singh, AWS VP of Compute Services, said of the deal: “We are excited to work with Portworx to provide customers another option for backup and data management on Amazon EKS.”

Reed Glauser, director of engineering at CHG Healthcare, provided a customer perspective: “As demand for healthcare workers skyrocketed, we needed to rapidly deliver new digital tools that better serve our customers, who count on us to provide more than 30 per cent of the temporary medical employees in the United States. We started by migrating most of our technology infrastructure on AWS and adopting Kubernetes to develop, test, and deploy new applications, but we needed a storage layer that was as flexible as the other elements of our cloud-native stack.”

So: “Portworx added the capabilities to Amazon EKS we needed, enabling us to optimize and automate our storage management, spin up new clusters and migrate seamlessly, and significantly accelerate and streamline our cloud development life cycle.”

There are two parts of the Portwork offering involved here. Portworx Enterprise enables users to request storage based on their requirements for capacity, performance level, resiliency level, security level and access, protection level, and more. Portworx PX-Backup adds point-and-click backup and recovery for all applications running on Kubernetes, even stateless ones.

The Pure-AWS deal includes Pure announcing an Early Access Program for Portworx Backup as-a-Service (BaaS) on AWS. Pure says Portworx BaaS is one of the many as-a-Service offerings the company will deliver to its customers in the future.

The pair’s deal, which they called a strategic engagement, means Pure gets AWS’ blessing for its Portworx software, which should help sales (shorter term win), and AWS gets a Portworx on-ramp to its cloud for Pure customers looking to go cloud-native (longer term win).

Will Pure look for similar arrangements with Azure and the Google cloud? Will AWS look for similar arrangements with HPE and NetApp and others? Watch this space. The (cloudy) sky’s the limit.

How Pure Storage helps customers guard against ransomware

By
Chris Mellor
-
ransomware_criminal

Paid Feature Ransomware is one of the great scourges of our Time. And it’s getting worse. We talked to Shawn Rosemarin, global vice president for Emerging Technology Solution Sales at Pure Storage, about how the company can help customers protect themselves

Blocks and Files: What happens if a Pure customer finds to their chagrin that they’ve been hit by a ransomware attack? What’s the best outcome they can hope for?

Shawn Rosemarin: In the ideal scenario, the customer has prepared and secured their infrastructure to sniff out and discover ransomware attacks as early as possible. It’s the dwell time that affects the difficulty of restore. The sooner you can know that someone’s doing something that they shouldn’t be doing, the quicker and easier it is to restore. It’s the amount of time and the amount of data that needs to be restored that is a direct driver of how long and complex it will be for you to get to that point of restoration.

The other assumption is that there’s no panic because the organisation has reliably prepped for the event. Everybody knows what their roles and tasks are. They also know which applications are essential to business operation, and how long they can be out of service without affecting operation. That would be the recovery time objectives, how long can they be down? How much data can they afford to lose? Can they afford to lose an hour, a minute, a second, and depending on their industry, that would come down to that RPO.

Can Pure Storage software or services detect ransomware and alert customers to it?

Across Pure storage arrays, we do look at what is happening. We work closely with partners in the industry in the log analytics space, like Splunk or Elastic, to power the security incident and event management architectures. 

So we look at what’s happening, for things that are deemed to be different or odd. Like someone suddenly spending a lot of time in shares that they have access to, but they wouldn’t typically go there.

The thing to think about with dwell time is that the longer I stay in the system, the more I can get access to other credentials. If I happen to snag a set of admin credentials, I could go and create problems, but I’m not going to do that yet, I’m going to use the credentials to get into other systems, look at where the backups are. What would be my attack vector that would give me the maximum leverage or cause the maximum pain so that that organisation would just pay the ransom?

How can Pure help customers recover if their data starts getting compromised?

So the first thing that happens is we know there’s a breach, and we have a plan or we put together a plan very quickly. Then the organisation is typically going to go to their insurance company and declare a ransomware attack, the insurance company would then involve a third party security forensics team like Mandiant to come in and discover what actually happened. What did they gain access to? What sort of data has been compromised? Are there some backdoors that have been set up?

This is a big part of it, because I have to decide what point can I restore to that is clean, assuming I have infrastructure to restore to. If the dwell time was 30 days, if I go back 31 days ago, I may be clean, but can I actually patch myself against them coming right back in?

But the most important piece is, do I have a non-encrypted backup to actually recover from? Or do I have non-encrypted snapshots to recover from? Snapshots are easy to restore but, in many cases, organisations cannot afford to keep 30, 45, 60 days of snapshots, so now you’re looking at coming back from backup. And that’s where you’re getting into large volumes of data that need to be restored quickly. And our experience is that with traditional tape or disk-based backup, we’re looking at roughly one terabytes, two terabytes an hour.

With our Pure Storage FlashArray//C, we see recovery speeds of eight terabytes an hour, and in FlashBlade recovery speeds of 40 to 270 terabytes an hour. So there are orders of magnitude of restoration time here significantly different than what has traditionally been viewed as kind of a backup.

Because the paradigm changed, right? We think about backup speeds, and how quickly it could be backed up, so how quickly could we get the SQL database done so that the impact on performance wasn’t there? However, these days, it’s restore time, how quickly can we restore the data of those critical applications?

Would people typically restore from snaps, say on FlashArray, and backups on FlashBlade?

Here’s the way I would say we look at it: a basic protection would be to use snaps to restore. I’m going to turn on SafeMode on my primary array so that the snapshots are being written to a volume that cannot be encrypted, it’s immutable.

If we get to a little bit better, we’re not only running snapshots, we’re actually taking backup volumes. We’re putting backup volumes now into those safe mode repositories. And now in the event that we can’t restore with snaps, we can actually go get an immutable backup.

In the best scenario, we would have a FlashArray//X connected to a FlashBlade so that we have all of our most recent backup volumes ready to be restored incredibly quickly. So in an event, we can mount from the FlashBlade and then put the FlashBlade volumes back onto our primary array.

Is there any difference between the way SafeMode operates on FlashArray//X, FlashArray//C and FlashBlade?

SafeMode operates the same in both scenarios. It’s an immutable volume. And it’s interesting because we’ve seen a lot of vendors say unless you’re the system administrator you cannot change or modify the array. SafeMode is different, we actually assume that anybody could be causing the damage, and your credentials could be compromised or stolen.

Because once a ransomware attack has happened, you can’t trust anybody’s credentials in the attacked organisation?

That’s right. It’s called zero trust, so I should give every employee only access to what they absolutely require. But we always have to assume any set of credentials can be compromised.

What happens from a SafeMode perspective, I have something called ‘Eradication Timer’. So let’s assume I have administrator credentials and go and delete a backup. With SafeMode there’s an eradication timer, and no matter who deletes it, it actually doesn’t delete for 30 days, 45 days or 60 days. So if it is deleted, no problem, it’s actually not deleted, it’s sitting on a timer that’s been preset.

Now, if you want to go and change that or reconfigure it, you have to call support and you have to be one of a few named individuals. So you have to identify yourself by name, and you have to state your PIN that only you know. And only then would support on our side go in and actually reset or change the eradication timer and the settings on SafeMode.

Security protection companies imply that recovery from ransomware is getting easier and easier. What is your response to that?

If you were to ask me whether you could restore from snapshots easily, I would tell you, yes, snapshots are incredibly easy. And I think that in the event that you were restoring data, or even a user file that they had accidentally deleted, that would be easy. The hard part is, as I said earlier, figuring out how did we get here. We have seen first-hand what is required to restore working with various organisations. This is not an easy task. I would love to put a marketing veneer and say, you know it’s one click. That’s not the way this works.

Sponsored by Pure Storage.

Quantum’s revenues fall as savage supply chain issues ravage costs

By
Chris Mellor
-

Quantum saw a 2.8 per cent decline in revenues in its latest quarter as severe supply chain issues – some suppliers “holding it to ransom” – cost it $62 million in unfilled orders and sent it into cost-cutting mode to get through the coming quarters of constrained supply.

Revenues in the quarter ended Dec 31, 2021, were $95.3 million compared to $98 million a year ago. There was a loss of $11.1 million, much worse than the year-ago loss of $2.7 million. But bookings rose and were higher than revenues for the fifth quarter in a row as demand grew strongly, particularly with tape use by hyperscalers and in conjunction with Pivot3 video surveillance orders.

Chairman and CEO Jamie Lerner said in the earnings call: “While our bookings were at the highest level they’ve been at in years, we have to recognise that the global supply situation put us in a position where the results have fallen beneath all of our expectations. But I’m confident in the underlying strength of this business. Our bookings levels, our hyperscaler wins, the strides we’re making in video surveillance, tell me that we have a growing business.”

Demand up

The CEO added that “backlog increas[ed] sequentially to $62 million, another quarterly record for the company, with strong demand.”

Demand for tape was notably good. “As of today, four of our hyperscale customers now utilise more than an exabyte of storage capacity, and we continue to view this business as a growth driver in future years.”

Three consecutive quarters of growth come to a stop with the latest quarter

Orders grew. “One of our largest hyperscale customers … converted a large order for LTO-8 drives to LTO-9 drives during the quarter and increased their order by $10 million.”

But it could not be satisfied – there weren’t enough LTO9 tape drives. 

Video surveillance was affected too. “We closed a multimillion dollar video surveillance deal at a government agency, which is currently in backlog.”

Object storage grew. “We closed a multimillion dollar object storage deal with a genomics research institution and closed our first six-figure software-only object storage win at a large semiconductor manufacturer with a third party providing the hardware component.”

The recurring revenue transition accelerated. “With more than 255 customers utilizing Quantum’s subscription solutions, up 30 per cent sequentially and 98 per cent year-over-year, and with more products transitioning to a subscription contract, our year-over-year subscription revenues increased more than 190 per cent.”

Lerner said “We are … closing a higher number of deals, both in number and dollar value across a larger customer base.” 

But it was to no avail.

Supply down

Quantum’s business is seeing a terrific growth in demand coupled with a severe downturn in component supply, as Lerner said: “We’re just having very good traction with our orders, but we are having a hell of a time fulfilling those orders.”

There were “unprecedented headwinds … within our supply chain.”

Tape supply was affected. “During the third quarter, LTO-9 tape drives started shipping, but overall supply of LTO-9 drive has been constrained due to initial manufacturing challenges.”

There were “broad-based shortages of components for servers, network cards and circuit boards.” An example: “we’re just getting much longer lead times sometimes as much as five months for just a server.”

According to Lerner, “We’ve just had so many de-commitments or delivery schedules that our suppliers did not meet or pricing levels that they did not meet.”

He expressed his frustration with the some component suppliers, saying “We can’t be in a position where we’re held for ransom quarter after quarter.” Some suppliers of low-margin, low-value components – $2 or $3 socket connectors for example – simply stopped making them and switched to high value, higher margin products. Others instituted unacceptable plus price rises. 

As a result, “We’re designing our products to use more popular components that have more supply available and are available at lower prices.”

Cash getting tight

Quantum’s results announcement contained this surprise: “Cash and cash equivalents including restricted cash was $4.3 million as of December 31, 2021, compared to $23.2 million as of September 30, 2021.”

That was a reduction of 81.5 per cent.

CFO Mike Dodson said that, in the quarter, “We experienced much higher manufacturing costs, combined with higher freight, warehouse and other logistics costs.”

Although Quantum’s cash went down a lot, “We remained in compliance with all debt covenants. But given our current expectations that the supply chain disruptions we have experienced in the last four quarters will continue in the foreseeable future, we have begun to work with our lenders to address any potential future covenant compliance issues, as well as any potential need for additional liquidity.”

That could mean another two to three quarters of constrained revenues for Quantum. Will other storage suppliers in the same boat? Tape library-shipping SpectraLogic for example?

What to do

Lerner said Quantum’s response was this: “We are immediately implementing a series of cost reduction measures. We are instituting pricing increases across our product categories, and we are focusing on supply chain and operational excellence.”

Its prices have already risen five per cent and will probably go higher. Dodson talked about “doing things like shutdowns, right. And then it’s looking at your typical items that you look at, the temporary employees, the contractors. You look at all your discretionary spending, you look at all of your contracts and where can you renegotiate or postpone or push-out.”

The outlook for next quarter, Quantum’s Q4, is for revenues of $92 million, plus/minus $5 million. The mid-point would be a slight year-on-year decline from the prior Q4’s $92.4 million, and imply full year revenues of $369.6 million – a 5.8 per cent rise year-on-year. A slight rise, but a rise nonetheless. With supply constraints looking to continue for a few quarters yet, Quantum will be thankful for what it can get.

Storage news ticker – February 10

By
Chris Mellor
-

…

Alluxio, developer of an open source data orchestration platform for data driven workloads, achieved 3x year-on-year revenue growth in its FY2022. It said seven out of the ten largest companies by market cap run Alluxio.

…

In-cloud data protector Clumio has announced strong growth metrics for 2021 due to increased demand for cloud-native apps and ransomware protection. It grew its AWS customer base by over 400 per cent and cloud ARR by more than 450 per cent. AWS accounts protected by Clumio grew by over 700 per cent year-over-year.

…

FalconStor Software announced the expansion of its go-to-market relationship with Hitachi Vantara, combining Hitachi’s flash storage arrays with FalconStor‘s continuous data protection, synchronous metro-area replication, data migration, snapshot, and disaster recovery software. Under the terms of the agreement, Hitachi Vantara will resell FalconStor StorGuard, running in x86 servers, with its VSP flash arrays to provide flash performance and protection for applications and data. The combined system is available through Hitachi Vantara to end users and approved channel partners and is supported by Hitachi and FalconStor worldwide.

…

Data protector NAKIVO announced year-on-year revenue growth of 30 per cent worldwide in its fourth 2021 quarter. NAKIVO has over 20,000 paid customers in 166 countries. The customer base grew by 21 per cent in Q4 2021 vs Q4 2020. The number of new customers grew by 23 per cent in the EMEA region, 21 per cent in the Asia-Pacific region, and 18 per cent in the Americas.

…

An IDC paper explores the business value of Scality RING object storage. Its interviews with Scality customers revealed that 68 per cent of their capacity on Scality’s RING directly supports revenue-generating business operations. Customers reported:

  • The total cost of their operations is 59 per cent lower with Scality;
  • They see payback on their Scality investment within eight months;
  • Their data analytics teams are 19 per cent more productive with Scality RING;
  • They see 366 per cent ROI over five years with Scality RING;
  • They are 53 per cent faster deploying new storage.

…

Vultr, an independent provider of cloud infrastructure, announced general availability of its partner program. Partners can earn industry-best margins on cloud compute and cloud storage, while building profitable value-added services and managed services on top of Vultr’s platform. Vultr offers cloud compute, cloud storage, and bare metal, through a web-based control panel or API. Launched in 2014, Vultr has served more than 1.3 million global customers, 22 cloud datacentres around the world – more than any other independent cloud provider – and has surpassed $125 million in annual recurring revenue (ARR).

…

Replicator WANdisco has completed a significant data migration for a major client, thought to be CN, the Canadian National Railway, to Microsoft’s Azure cloud. This client, using WANdisco’s LiveData Migrator, successfully migrated data off its operational Hadoop cluster, to the Azure cloud. There is 70TB of such data with an extension possible to move another 30TB/year to Azure and a possible future IoT use case.

…

HPE’s Zerto business unit is offering a free edition with the ability to protect ten VMs with Zerto for one year. Once you fill out the form to request the Zerto Free Edition, you’ll receive an email with your unique key and download instructions. The background is set out in a Zerto blog about fighting ransomware.

…

Nasuni bids for a seat at the table with cloud titans and NetApp

By
Chris Mellor
-

Independent cloud file services supplier Nasuni is seeking a spot at the top cloud file services table, citing an improved technology that doesn’t require lock-in, and claimed lower costs than rivals NetApp, AWS, Azure and Google Cloud Platform.

Update. Nasuni’s mistaken 192TB filesystem limit for AWS FSx ONTAP corrected. 18 February 2022.

Its ambition is to grab $5 billion share of the cloud file market, which is quite audacious seeing as it would be equal to the share of cloud filer giant NetApp – or the combined share of the top public cloud trio of hyperscalers AWS, Azure, and Google Cloud Platform.

Nasuni, unsurprisingly, sees NetApp and its data fabric offerings as its biggest competition. Nasuni believes it can grow and prosper by similarly working with the titan cloud trio – it has co-sell deals with them – and, in theory, offering simpler technology, lower costs and better services.

It’s a competitive sphere, with not much new growth, it seems. Nasuni president and CEO Paul Flanagan said “40 per cent of our business is business we’ve taken” from competitors.

Co-selling

COO David Grant said “We believe that the cloud companies, Azure, and AWS, will win their fair share of this market with their own first-party offerings. We also believe that NetApp will win a fair share of the portion of this market, because they have a lot of happy customers that have been using them for 20 years. 

“But we also believe that there is room for an independent vendor being us in this case. We aspire to be that independent. … We think it’s essentially a race right now that when you look at pure cloud file storage sales, Azure AWS, NetApp and Nasuni are roughly the same size, maybe Azure’s a little bit bigger.”

He said “We have co-sell relationships with all three cloud providers. Last year, they brought us about 30 per cent of our business.”

He said Microsoft brings Nasuni into accounts when scale is needed. “Azure Files is a good product, but it’s a good product for small and medium sized businesses. It’s not a product suitable for scale.”

It’s different with AWS. “They have a lot of high-end HPC type offerings [like] FSX for NetApp, but in the middle, they don’t cover a lot of the general NAS with that product, because … it’s extremely costly [and] they have a lot of limits on how big it can scale.”

Google has “been a great relationship for us. They brought us a lot of good accounts in 2021. And they continue to sell us as an option, as … they do not have a file storage product except for on the high end side of things with file store.”

Nasuni’s growth

Flanagan said “We’re building a file services platform that is servicing some of the largest companies in the world. … our revenues continue to grow,” with a 30 per cent year-on-year annual recurring revenue rate in 2020 and in 2021. The new customer growth rate was 82 per cent in 2021 and the year ended with around 600 enterprise customers. Between 98 and 99 per cent of its customers renew their yearly subscriptions.

He claimed the company has a “120 per cent revenue expansion rate.”

The Nasuni boss added “So that means if a customer is spending $1, this year, next year, they’re going to spend $1.20. In the year after that, they’re going to spend $1.50, and the year after that they’re going to spend $1.85.”

How the technology works

Nasuni presents a file system to its users who access it via so-called edge appliances. These are typically virtual machines or actual x86 appliances running Nasuni’s software on-premises or in the public clouds and offered in a SaaS business model. They provide local caching and access to Nasuni’s global UniFS filesystem data and metadata stored in object form in the public clouds. 

Flanagan said “We have fundamentally changed the entire architecture by writing to object storage.”

UniFS has built-in global locking, file backup, disaster recovery and file synchronisation – reflecting Nasuni’s origin as a file sync and sharer – and helps with ransomware recovery through its versioning system. Because it is cloud-native, Nasuni says it is limitlessly scalable – unlike its main competitor, NetApp.

This point was hammered home by Nick Burling, Nasuni’s product management VP, who said “The edge caching is fundamental” and backed up by limitless cloud object storage and limitless numbers of immutable snapshots with single pane of glass management. This then “leads into being able to displace other services that customers are having to leverage, like file backup [and] disaster recovery.”

The edge VMs or appliances have Nasuni software connectors that enable file migration off existing NAS systems into Nasuni’s UniFS system. Nasuni’s edge software also deduplicates, compresses and encrypts locally written data before sending it up to the customer’s chosen cloud object store.

There are separate control and data planes and “Nasuni is never in the data path for our customers. They own and manage their own object storage account.”

Nasuni’s software is built to be generic in the way it provides gets, puts and versioning to the cloud object stores. It does its own versioning rather than relying on each cloud vendor’s scheme. Burling said “We made a decision early on to build out our own versioning system independent of any object store – it took a lot of time. It’s a very, very big piece of IP for us. But it gives us the flexibility we need.”

He told us “If you use the versioning of the [CSP’s] object store, we’ll never get the performance that you can get from our custom versioning. … They built versioning for lots of different use cases. We built our versioning specifically for a file system. And that really changes the game in terms of how you how you construct, where you convert from, how you construct, where you convert back to, and the relationships between objects.”

Object Lambda

An aspect of this cross-CSP approach affects added data services, as Burling explained. “When we see the opportunity to implement new features of the object store, we look for opportunities where we see a large scale change across different object stores. And one that we see on the horizon is this idea of every protocol being enhanced by custom code.”

An example of this is “The S3 protocol being able to be extended with custom code inside of AWS, which is called Object Lambda.” The Object Lambda idea is to apply filtering functions to a data set before it is passed on destination applications for e-commerce, analytics, marketing, etc. so that the same source dataset can be used for each application with no dataset copying involved.  

Burling said “There are other things they’re developing inside of Google and Azure, as far as we understand, that have a lot of potential to really open up new possibilities for us and for other customers.” Although the actual services in each CSP’s environment will be specific to that CSP, “Underneath all of that will be the same Python code. And the implementation of the function is different per different platform. But having the flexibility to have your own get or your own put that does a transformation is the part where we see independent value across the different providers.”

Further, “We see Python as a function … just think of it as an algorithm that we can easily transport to different platforms. And it just happens to get wrapped up in either Lambda, or it’s in Cloud Functions or it’s an Azure function on the other platforms.”

Nasuni vs NetApp

Even in a single cloud use case, Burling claimed, “Because of our architecture, we’re still better than the first-party services that our customers are looking at, because of those same things: scale, simplicity, and cost.”

He singled out a bid against NetApp where a $32 billion global manufacturer that designs and distributes engines, filtration and power generation products wanted to deploy 10PB of file data in AWS and looked at both NetApp (FSx for ONTAP) and Nasuni. 

There had to be 55 instances of FSX ONTAP because of a 192TB filesystem limit.

Update: The 192TB filesystem limit mentioned on the Nasuni slide is incorrect. An AWS spokesperson told us: “Each Amazon FSx for NetApp ONTAP file system can scale to petabytes in size.” 18 February 2022.

Burling said “They’re going to have to tier across SSD and S3, because the metadata has to remain an SSD, if they want to be able to have disaster recovery. So [they’re] not able to benefit from the full cost savings of of object storage with a first-party AWS service.”

The Nasuni alternative looked like this:

It featured a single tier, with no SSDs needed for metadata, and comparatively slimmed down edge access as full Windows server edge systems are not needed. Nasuni hopes to win the bid, on the basis of its simplicity and cost advantage, with Burling explaining “This is very early in the process for us and so we’re not at the point of being able to do a full public case study.”

Nasuni roadmap

Nasuni wants to build out file data services in five areas: ransomware protection, data mobility, data intelligence, anywhere access, and SaaS app integration. We can expect an announcement about enhanced ransomware protection in a few weeks.

Data mobility means the ability to write to clouds simultaneously, not just to one cloud. Data intelligence refers to the custom functions for clouds, such as Object Lambda, which has  already been mentioned. It also refers to better management of chargeback and showback. The latter provides iT resource usage information to departments and lines of business without the cost or charging information associated with chargeback.

On the SaaS integration front, Burling said “We have customers that have workflows with critical third-party SaaS services, and they want them to be better integrated into their overall experience with Nasuni.”

Nasuni is also setting up a Nasuni Labs operation – a GitHub account managed by Nasuni for the supply of open source additions to its base product. Examples are the Nasuni Management Console API, a Nasuni Analytics Connector, Amazon OpenSearch and Azure Cognitive Search.

Comment

Nasuni is fired up on the strength of how its co-sell agreements are working, its place in the marketing strategies of the public cloud providers, and its experience winning bids against NetApp and other filer incumbents such as Dell EMC’s PowerScale/Isilon. 

NetApp users can be ferociously loyal to NetApp, which has built a multi-cloud data fabric strategy that AWS, Azure and GCP view positively – witness its OEM deals. It’s also building a CloudOps set of functions with its Astra and Spot product sets which differentiate it from Nasuni and are winning it new customers.

There’s no quick win here for Nasuni. It faces hard work to progress, but it is confident of progressing. It has loyal customers of its own, and the cross-cloud added functions like Object Lambda look a useful idea. Let’s see how it pans out this year and next.

VAST becomes backend Commvault target and tech alliance partner

By
Chris Mellor
-

Commvault backup data can be stored in VAST Data‘s scale-out file storage systems in a new partnership.

VAST’s all-flash Universal Storage system provides fast backup data ingest and restoration and is an alternative to purpose-built backup appliances (PBBA), such as Dell EMC’s PowerProtect (Data Domain), ExaGrid appliances and others. VAST claims it provides up to 50x more backup restore performance, at a cost that is 40 per cent lower than legacy PBBAs. 

VAST also claims its deduplication combines with Commvault’s software to bring the effective cost of flash in line with HDD-based storage systems, but hasn’t provided example data reduction ratios to illustrate. By contrast, ExaGrid has ESG evidence it can produce 15:1 data reduction when storing Commvault backups.

Commvault has made VAST a Commvault Technology Alliances Partner and says the two have completed validation of VAST’s Universal Storage platform with Commvault’s Complete Data Protection Suite. VAST claims its storage system can run different workloads without impacting backup and restore jobs.

Pure Storage, a VAST all-flash array competitor with its FlashArray (block) and FlashBlade (file + object) systems, has a Commvault-Pure reference design combining its Complete Data Protection Suite software with the FlashBlade array to provide fast data protection and rapid recovery. The design features Pure’s SafeMode snapshots feature that creates read-only snapshots of backup data and associated metadata, as a way of combatting malware.

The VAST-Commvault alliance features VAST’s Indestructible Snapshots feature which prevents backup copies and snapshots from being altered or destroyed and similarly fights malware attacks.

The alliance with Commvault strengthens VAST’s enterprise supplier credentials and we might expect it to be followed by alliances with other enterprise-class data protectors such as, for example, Veritas, Veeam and Rubrik.

Infinidat strengthens InfiniGuard with automated network fencing and more throughput

By
Chris Mellor
-

Infinidat has added automated fenced-in networks to its InfiniGuard data protection array to safeguard forensic analysis of malware attacks, and doubled throughput with new B4300 hardware. The resulting software is branded InfiniSafe, and is claimed to enable a new level of enterprise cyber-resiliency.

The InfiniGuard system is a mid-range F4200 or F4300 InfiniBox array front-ended by three stateless deduplication engines (DDEs) and running InfiniGuard software providing virtual tape library functionality (disk-to-disk backup) as a purpose-built backup appliance (PBBA). It supports Commvault, IBM Spectrum Protect, Networker, Oracle, Veeam, Veritas and other software products backing up heterogeneous storage vendor’s kit.

Stan Wysocki, president at Infinidat channel partner Mark III Systems, provided an announcement quote: “What I’m most excited about is providing our customers a comprehensive storage-based ransomware solution that combines air-gapped immutable snapshots with an automated fenced-in network to determine safe recovery points, and then delivering near-instantaneous recovery.”

InfiniGuard arrays are preinstalled in a 42U rack with four drive enclosures preconfigured. There are InfiniGuard B4212N and B4260N systems, shipping from April 2021, based on the Infinidat mid-range F4200 array. The B4300s are based on the F4300 InfiniBox – a go-faster F4200 upgrade. Some differences between two now discontinued B4200 products and two B4300 products are listed in the table below:

The InfiniGuard software currently offers immutable snapshots, a virtual air-gap and virtually instantaneous recovery, based on all disk drive spindles being used in the restore operation. It can also replicate an InfiniGuard system’s contents to a remote InfiniGuard system.

Performance boost

The new B4300 hardware increases the DDE CPU power by moving from six-core to 20-core CPUs and doubling DDE memory. The network link between the DDEs and the backend storage is 16Gbit/s FC, and a new 25GbitE Ethernet frontend option has been added.

InfiniGuard software has been updated to use all the cores and the result is a throughput jump from 74TB/hour to 180TB/hour. This can reduce backup windows by 50 per cent or so, as well as shortening remote replication time. This means you get to a fault-tolerant (replication complete) state faster.

Recovery time is also reduced. Infinidat can now demonstrate the recovery of a 1.2PB backup data set in 13 minutes and 30 seconds, at which point it’s available to the backup software to do a restore. Eric Herzog, Infinidat’s CMO, tells us “No one else in the PBBA space can do a full recovery in less than 30 minutes.”

Competitor HPE’s fastest StoreOnce deduping PBBA, the 5660, has a 104TB/hour throughput rating. Dell EMC’s high-end PowerProtect DP8900 runs at up to 94TB/hour. InfiniGuard B4300 is almost double that speed. However, In January 2019, ExaGrid CEO Bill Andrews claimed his firm’s EX6300E PBBA scaled out two a 400TB/hour ingest rate. Its EX84 currently scales out to 32 systems and a 488TB/hour ingest rate in the EX2688G configuration. That beats InfiniGuard’s B4300s in sheer speed but is not comparing apples to oranges in terms of system configurations. InfiniGuard looks to be one of the fastest single node, disk-based, PBBAs out there, if not the fastest.

Pure’s FlashBlade offers up to a claimed 270TB/hour recovery performance from its all-flash datastore. A single node all-flash VAST Data/Commvault system offers up to 144TB/hour restore speed. Four nodes offer up to 576TB/hour and eight nodes 1.15PB/hour; this is what scale-out design is supposed to do.

Network fencing

The latest InfiniGuard release adds support for fenced and isolated networks. Network fencing isolates a device, such an InfiniGuard array from its connecting network. That means backups can be test restored inside a closed and isolated environment to find the latest known good backup and not affect primary systems at all. This means InfiniGuard customers can disregard any compromised backups and only recover good data.

Herzog suggests that several competing PBBA vendors have cyber-resilient systems roughly similar to InfiniGuard but they can require two separate systems. With its network fencing InfiniGuard can provide the cyber resilience in a single system.

Storage news ticker – February 8

By
Chris Mellor
-
W. Thomas Stanley.

…

Cloud file data manager (sync and sharer) Nasuni has a new board member: W. Thomas Stanley, the president and CRO for Chainalysis, CRO at Tanium before that, and a 13.5-year NetApp veteran, leaving as SVP and GM for the Americas in September 2019. He joined Nasuni’s board, taking the second independent seat, in November last year. Paul Flanagan, Nasuni CEO said “He will play a key role in helping guide Nasuni in the next phase of our growth as we continue to disrupt the file storage and data management market with a cloud approach.”

The other independent board member is Joel Reich who was chief product officer at NetApp for 16 years. 

…

Qumulo says it had a good 2021, achieving its best-ever quarterly results in overall billings, software subscription billings, cloud billings, and customer adoption. It increased its billings by 75 per cent from Q3 to Q4 alone. Qumulo added more than 200 customers in the year and its customers have stored over 2 exabytes (EB) of data, more than doubling last year’s record. Its latest Qumulo Core v5.0 release features Qumulo on Azure as a Service, Qumulo Shift for Amazon S3 to export/import data to/from AWS S3 while leaving data in its native object format, Qumulo on HPE and Supermicro all-NVMe, NFSv4.1 Support and NVMe-Hybrid to enable the performance of NVMe at the cost of dense disk.

…

Ghazal Asif.

Data protector/manager Rubrik, appointed Ghazal Asif as VP of global partners and alliances to drive continued growth and engagement with current and future partners. Asif joins Rubrik from Google where she was the head of channel partners in EMEA for Google Customer Solutions, leading channel go-to-market strategy and execution for the region. 

…

Kaseya’s Unitrends business launched ION and ION+ data protection appliances. They are smaller than rackmount products and the ION is a silent all-NVMe SSD device featuring automated disaster recovery (DR) testing, immutable storage, and AI-based ransomware detection. The ION+ has more powerful CPUs and additional memory. It comes in a small tower form factor with the ION’s software plus full-service backup and recovery. Both the ION and ION+ can be used with Unitrends Forever Cloud for offsite immutable storage and Unitrends Disaster Recovery as a Service (DRaaS) for hands-free failover to eliminate downtime.

IBM constructs CyberVault around new FlashSystem family

By
Chris Mellor
-

IBM is fortifying its new PCIe 4-enhanced FlashSystem family with CyberVault features designed to detect, prevent and recover from malware attacks.

Update; gen 3 FlashCore Module information added. More performance data added. 9 February 2022.

FlashSystem data is actively monitored in real-time and ransomware recovery accelerated by using validated restore points. Two new mid-range and high-end FlashSystems get PCIe 4 connectivity along with new controller CPUs, and have both capacity and speed increases.

David Chancellor, director enterprise systems for IBM biz partner Gulf Business Machines, said “Cyber resilience is clearly a top priority for our customers. [IBM Cyber Vault’s] ability to help dramatically reduce recovery times is exactly what cyber resilience teams need to keep the business running.”

IBM mainframe CyberVault graphic.

IBM did not say how CyberVault’s active monitoring and restore point validation work. Its mainframes have a Z CyberVault feature with active monitoring and safeguarded copies and FlashSystem CyberVault is based on this scheme and uses regularly taken safeguarded copy point-in-time snapshots. CyberVault automatically scans the copies created regularly by Safeguarded Copy looking for signs of data corruption introduced by malware or ransomware. It uses standard database tools and automation software.

FlashSystem rack and (inset bottom to top) 5200, 7300 and 9500 models.

New hardware

There are three members in the FlashSystem family, all running the same Spectrum Virtualize code: 

  • Existing 5200 – 1U – 12x NVMe drives in base controller chassis
  • New 7300 – 2U – 24x NVMe drives and replacing the 7200
  • New 9500 – 4U – 48x NVMe drives and replacing the 9200

The 7300 supports dual-port and hot-swappable 4.8, 9.6, 19.2 and 38.4TB gen-3 FlashCore Modules (FCMs – drives). The 19.2 and 38.4B FCMs have a PCIe 4 connection. Commercially available NVMe SSDs with 1.92, 3.84, 7.68, 15.36 and 30.72TB capacities are supported in dual-port, hot-swap form. It also supports SCM drives. The expansion chassis support the use of 2.5-inch and 3.5-inch disk drives as well, caused by convergence of the FlashSystem and Storwize product ranges.

Italicised columns are older systems included for reference.

The gen 3 FlashCore Module has an onboard processor (computational storage) to handle background storage tasks and uses industry standard QLC (4 bits/cell) and faster SLC (1 bit/cell) NAND technology to address performance and capacity requirements. At 38.4TB raw capacity (or 116TB effective), this makes it possible for IBM to deliver 1.1PB effective capacity per single rack unit in the FlashSystem 5200, 7300 and 9500 products.

IBM Master Inventor Barry Whyte said the gen 3 FCM’s “internal computational storage capabilities have been enhanced with a new “hinting” interface that allows the FCMs and the Spectrum Virtualize software running inside all FlashSystem products to pass useful information regarding the use, heat, and purpose of the I/O packets. This allows the FCM to decide how to handle said I/O, for example, if its hot, place it in the SLC read cache on the device. If its a RAID parity scrub operation, de-prioritise ahead of other user I/O. Single FCMs are now capable of more IOPs than an entire storage array from a few years ago!”

Tom Richards, systems and storage practice lead at IT consultancy Northdoor, told us the gen-3 FCMs have improved performance and double the Drive Write Per Day (DWPD) rating over standard SSDs. The in-line compression hardware increases the effective capacity of the larger modules to 3:1. FCM above the entry 4.8TB modules in the previous generation were limited to just over 2:1 compression. The new FCM 4.8, 9.6, 18.9 and 38.4TB modules can achieve hardware-accelerated compression of up to 22, 29, 58 and 116TB effective capacities respectively.

Richards told us the new FCMs will be supported in both of the newly announced arrays, along with the existing 5200 arrays (although mixing of old and new FCM types will not be supported within the same array).

The 9500 supports the same FlashCore Module and commercial SSD capacities as the 7300, plus Storage Class Memory with support for 1.6TB drives specified. The 9500 also supports SAS SSDs but not disk drives.

Performance

The FlashSystem 9500 offers twice the performance, twice the number of drives, and twice the connectivity of the 9200. IBM claims it provides performance gains of 50 per cent, 2x increase in connection options, and a 4x increase in workload support, but neglects to say what product is being used in the comparison. A customer said that the Cyber Vault FlashSystem reduced overall cyber attack recovery time from days to hours with a “comparable DS8000 function.”

An IBM spokesperson said the 7300 delivers 45GB/sec bandwidth, 3.5 million cache hit IOPS, and 580,000 IOPS with a 70 per cent read, 30 per cent write, 50 per cent cache workload. The faster 9500 outputs 100GB/sec bandwidth, 8 million cache hit IOPS, and 1.6 million 70/30/50 IOPS.

Whyte said: “I’ve been calling this box “the beast” and for any ‘The Chase’ fan’s I’m not referring to Mark Labbett, but the new top of the line FlashSystem. With 48 x NVMe slots, 12 x Gen4 PCIe Host Interface slots (48 x 32Gbit/s FC, 12 x1 00GbitE, 24 x 25GbitE), 4 x PSU, 4 x Battery, 4 x boot devices – this is the largest and most powerful FlashSystem we’ve produced, and, based on the performance measurements, it has to be one of, if not the fastest, highest bandwidth single enclosure systems in the world!”

Connectivity

FlashSystem 7300 Model 924 systems include eight 10Gb Ethernet ports as standard for iSCSI connectivity and two 1Gb Ethernet ports for service technician use. These models can be configured with up to three I/O adapter features to provide up to 24x 32Gb FC ports with SCSI and FC-NVMe support, or up to 12x 10/25Gb or 100Gb Ethernet ports with iSCSI and NVMe RDMA support.

FlashSystem 9500 systems can be configured with twelve I/O adapter features to provide up to 48x 32Gb FC ports, up to 20x 10/25Gb Ethernet (iSCSI, NVMe RDMA capable) ports, and up to 12x 100Gb Ethernet (iSCSI, NVMe RDMA capable). Two 1Gb Ethernet ports are provided for management and service actions.

According to Richards, “Both of the new FS7300 and FS9500 include the option for 100GbitE iSCSI host connectivity and, crucially, the ability to drive these host connections at full speed thanks to the new packaging at PCIe 4,” (in the 9500).

The two systems can both have end-to-end NVMe connectivity. Both come with utility model options which delivers a variable capacity system, where billing is based on actually provisioned space above the base. The base subscription is covered by a three-year lease that entitles you to utilise the base capacity at no additional cost. If storage needs increase beyond the base capacity, usage is billed based on the average daily provisioned capacity per terabyte, per month, on a quarterly basis.

IBM says that its customers can be helped to achieve sustainability goals because of the new FlashSystem’s greater performance and capacity in a smaller and more energy efficient footprint than prior models.

The FlashSystem 7300 requires IBM Spectrum Virtualize licensed machine code level 8.5 or later, for operation and the 9500 requires equivalent Spectrum Virtualise software. Get a FlashSystem 7300 datasheet here and a 9500 datasheet here.

Bootnote: IBM has refreshed its SAN Volume Controller (SVC) product, giving it a CPU upgrade with 2 x 24-core Intel Ice Lake processors, making it more powerful. The SV3 node hardware is essentially based on the same node canister used in the 9500.

1...254255256...452Page 255 of 452

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