HYCU has launched SaaS-based Backup and Recovery for Office 365. “We see SaaS as a natural extension of the customer’s data estate,” CEO Simon Taylor said today in a press briefing. He noted that many HYCU customers are adopting multiple public clouds at the same time.
Simon Taylor
Office 365 is now called Microsoft Office – we’ll refer to it as O365 to stay in tune with HYCU’s announcement. The new fully-managed O365 service is integrated into HYCU’s Protégé multi-cloud management facility.
Subbiah Sundaram, HYCU VP Products, said there are four advantages to HYCU’s Office 365 backup that make it a unique product.
It’s a service and not based on an on-premises machine,
It’s comprehensive and covers more than SharePoint and Outlook,
It can meet data residency requirements,
It has the most comprehensive eDiscovery features.
Multiple competitors offer O365 backup, including Acronis, Clumio, Cohesity, Commvault, Druva, IBM, Rubrik, Veeam and Veritas. They all say they offer comprehensive Office 365 backup with many – Cohesity, Veeam and Veritas, for example – calling out their eDiscovery capabilities. Customers will need a supplier-feature matrix to compare and contrast the various offerings.
In the meantime, here’s a quick feature list for HYCU’s O365 protection.
Coverage of Outlook, Contacts, OneDrive (1x/day), OneNote, SharePoint (3x/day), Teams, Outlook email (12x/day). There is also email journalling.
Granular and full recovery for OneDrive (files), SharePoint (sites), Groups and Teams. Search and recover functionality for Email, OneDrive and SharePoint.
Detailed audit trail of email backups, searches, downloads, deletions. Suspend email expiration on-demand for legal audit, etc.
Many security standards, encryption, key management and compliance standards met.
No sizing involved; it’s all dynamic.
Not only but also
HYCU offers agent-less, purpose-built backup and recovery services for Nutanix, Azure, Google Cloud, VMware, AWS and – now – O365. Services are native to each cloud and can be used for disaster recovery and data migration. The company has a Net Promotor Score of 91 and 2,000-plus customers, which includes Nutanix and GigaOm. The company said revenues grew 450 per cent in 2020.
HYCU has a SAP HANA Disaster Recovery offering for Google Cloud Platform. Nutanix Mine v3.0 object storage is supported by HYCU and has built-in replication to multiple sites.
HYCU is expanding ransomware protection with immutable backups to WORM-enabled S3-compliant targets like Nutanix Objects, and source-to-cloud target encryption.
The company has added agentless backup of physical Linux servers for Red Hat Enterprise Linux and Oracle Enterprise Linux. HYCU has also added network throttles to optimise backup traffic.
A ruling in a court case in the UK concerning intercepts of encrypted EncroChat mobile phone messages has decided that data in the phone’s RAM was being stored: nonsense in an IT sense, but nonetheless allowing law enforcement to use the information.
The judgement was outlined in a Register story – “EncroChat hack case: RAM, bam… what? Data in transit is data at rest, rules UK Court of Appeal” – and we delved into the actual judgement to see what was going on.
The UK Court of Appeal ruled that RAM counted as data storage. How can this be the case?
Two meanings
There are two meanings of storage that appear relevant to this: IT and common parlance. In everyday speech storage means the retention of something for future use and memory the faculty by which the mind stores and remembers information. Storage and memory overlap. Both, we might say, are persistent.
But we might also say that, in humans, when the power is switched off at the end of our lives, then the contents of our memory disappear, being to this extent “volatile”. (But we are playing with words here.)
In IT memory contents are volatile, disappearing when power to the DRAM is switched off, and storage is persistent, with the contents unaffected by on-off power cycles.
Storage vs transmission
In the court case a distinction was drawn between storage of information and transmission of data in telephone devices. Different warrants have to be obtained by the police for intercepting and copying transmission data and for copying “stored” data.
Information recovered via a stored data warrant cannot be used in a case involving information transmission. Neither can recovered transmission data be used in a case involving stored data.
The information recovery from the EncroChat phones was achieved with a storage warrant and not a transmission warrant. Defence counsel made the case that the warrants were inappropriate as the recovered data was being transmitted, as it had been copied from the phones’ DRAM and not their storage, or ‘Realm’ as it was called.
This argument relied upon the transmission process including the preparation of the data being included in a message. Thus that data would be fetched from storage and copied into DRAM. There it would be used to construct a message and be formatted before being sent to the phone’s transmission hardware: radio chip and antenna.
The justices ruled that “what was intercepted, was not the same as what had been transmitted because what had been transmitted was encrypted. It cannot therefore have been ‘being transmitted’ when it was intercepted: it can only have been ‘being stored’.” The ruling likened the extracted data to a “draft”.
No ‘technical terms’ in the 2016 Act
The Appeal Court justices disagreed with “expert witnesses” who explained that making a copy was part and parcel of the act of transmission, stating that while this might be true from an expert’s point of view, it was not the intent of the government when drafting the relevant Act of Parliament.
The judgement stated: “The experts have an important role in explaining how a system works, but no role whatever in construing an Act of Parliament. They appear to have assumed that because a communication appears in the RAM as an essential part of the process which results in the transmission it did so while ‘being transmitted’.
“That is an obvious error of language and analysis. It can be illustrated by considering the posting of a letter. The process involves the letter being written, put in an envelope, a stamp being attached and then the letter being placed in the post box. Only the last act involves the letter being transmitted by a system, but all the acts are essential to that transmission.”
As the relevant Act defined only two states for data: being stored or being transmitted, then, if it was not being transmitted it was being stored.
The reasoning was as follows: Data in RAM is not being transmitted. It is necessary for it to be in DRAM for a message to be prepared for transmission, but the act of preparation is not transmission. Therefore the data must be in a stored state and the warrant used to recover it was valid.
In other words, the common parlance meaning of storage takes precedence over the technical definition, despite the very technical issues at play.
Lance Cpl. Brandon Allomong, Marine Aviation Logistics Squadron, (right) and fellow competior wolf down their pizza during one of the four Pizza Eating Contests at this years BayFest celebration. Allomong was the winner of this contest by managing to take down one whole pizza. He now has a year of free pizza from Papa John's Pizza.
IBM has launched the pizza box-sized FlashSystem 5200, its most compact storage system to date. The company has further updated the 5000 line with new 5015 and 5035 systems and added Spectrum Virtualize for Public Cloud on Azure.
The FlashSystem 5200 is faster and stores more data than its predecessor, the 5100, but base price averages out at 20 per cent lower.
Denis Kennelly, GM for IBM Storage, said in a statement: “Systems that provide global data availability, data resilience, automation, and enterprise-class data services are more critical than ever. Today’s announcement is designed to bring these capabilities to organisations of any size.”
IBM FlashSystem family. The 9200R is a rack mount 9200.
The current FlashSystem 5000, 5100, and 7200 and 9200 models are 2RU enclosures with up to 24 x 2.5-inch format FlashCore Module Drives (FMD) mounted vertically across the front.
FlashSystem 5200
The 5200 is half the height at 1RU – so rack densities are doubled – and has up to 12 FMDs mounted horizontally in two rows across the front. The NVMe FMDs use 96-layer NAND which is organised into a combination of SLC flash and QLC flash and come with a seven-year warranty in 4.8, 9.6, 19.2 or 38.4TB capacities. Industry-standard storage-class memory drives fit in the same slots.
In-drive FMD hardware compression yields a general 2:1 data reduction ratio. Software deduplication in the 5200 controller ramps this up to 5:1. Thin provisioning can yield a further 2:1 effective data reduction.
A clustered system supports up to four 5200 enclosures for a maximum of 48 drives, meaning a theoretical maximum raw capacity of 1.84PB using 48 x 38.4TB drives. SAS SSDs are supported in the expansion enclosures.
FlashSystem product range characteristics.
After data reduction, the 5200 scales from 38TB effective capacity to 460.8TB in the chassis and out to 1.7PB.
The FlashSystem 5200 has one 8-core 2.3GHz Skylake-D Intel Xeon D-2146NT processor per controller with two dual-active controllers per control enclosure. The 5200 is equipped with a PCIe gen 3 bus and has eight x 32Gbit/s Fibre Channel or 25Gbit/s iSCSI ports.
The system incorporates all the software features of the 5100, including high-availability and external storage virtualization. But it offers 66 per cent greater maximum I/Os than the 5100 and 40 per cent more data throughput at 21GB/sec.
The operating system is Spectrum Virtualize, which manages up to 300 external storage arrays and adds their capacity to the FlashSystem pool. A Smart Data Placement algorithm puts the hottest data in the SLC flash for faster access.
The IBM FlashSystem 5200 supports Red Hat OpenShift, Container Storage Interface (CSI) for Kubernetes, Ansible automation, and VMWare and bare metal environments. The system comes with IBM Storage Insights, which provides visibility across complex storage environment.
Future plans
IBM Spectrum Virtualize for Public Cloud is software that enables users to replicate or migrate data from heterogeneous storage systems between on-premises environments and IBM Cloud or Amazon Web Services. IBM will extend the same capabilities to Microsoft Azure, starting with a beta program in the third quarter of 2021.
The company is developing IBM Cloud Satellite software to enable customers to build, deploy and manage cloud services from any vendor anywhere.
Cloud Satellite standardises a set of Kubernetes, data, AI and security services. It will be delivered as-a-service from a single console, managed through the IBM public cloud and is currently in beta test.
The software is expected to become generally available in March and IBM will add support for it to the FlashSystem portfolio, SAN Volume Controller, Elastic Storage System and Spectrum Scale.
The University of Pisa relies on multiple Dell EMC Power-something arrays for its storage needs. AWS has given its on-premises public cloud presence local backups, Microchip has brought out the fastest PCIe switch to date while data protector and file manager Quantum has sidestepped a potentially nasty lawsuit.
University of Pisa is a Dell EMC Power user
Italy’s University of Pisa uses a raft of Dell EMC storage technologies.
Leaning tower of Pisa.
A PowerStore system stores scientific computing applications for genomics and biology, plus chemistry, physics and engineering. It delivered a 6x performance improvement on a previous unnamed storage system.
The University uses the system to support remote learning. CTO Maurizio Davini, said: “As we transitioned to remote learning, we needed reliable, scalable technology to provide our 53,000 students and faculty with quick, easy access to critical data and applications at all times, from any location. Dell EMC PowerStore is … a game-changer.”
The university supports VDI and remote workstations and database workloads with a PowerMax storage array. A PowerScale all-flash system handles artificial intelligence and bare metal high performance computing (HPC) workloads, The university expects unstructured data volumes to double within a year.
AWS Outposts gets local backups
Amazon’s on-premises appliance AWS Outposts now supports local snapshots for Amazon Elastic Block Store (EBS) volumes. This makes it easier to comply with data residency and local backup requirements.
Until now, Amazon EBS snapshots on Outposts were stored by default on Amazon S3 in the AWS Region. EBS Local Snapshots on Outposts is a new capability that enables snapshots and Amazon Machine Image (AMI) data to be stored locally. The feature is handled through the AWS Management Console, AWS Command Line Interface (CLI), and AWS SDKs. You can also continue to take snapshots of EBS volumes on Outposts, which are stored in S3 in the associated parent region.
AWS said customers can easily migrate, replicate, and recover workloads from any sources directly into AWS Outposts, or between AWS Outposts devices, without requiring the EBS snapshot to data to go through an AWS Region. This also allows CloudEndure Migration and Disaster Recovery services to copy data locally, improving recovery times and assisting customers with strict data residency requirements.
Microchip’s PCIe Gen 5 switch
PCIe gen 4 is just arriving in products, giving the twice gen 3 PCIe bus speed and now Microchip brings out a PCIe Gen 5 switch, doubling speed again.
PCIe Gen 5 runs at 128GB/sec across 16 lanes. PCIe Gen 4 does 64GB/sec and PCIe Gen 3 operates at 32GB/sec.
Microchip PCIe Gen 5 switch
Microchip’s PFX Switchtec supports 28 to 100 lanes and up to 48 non-transparent bridges (NTBs). It comes with XpressConnect retimers, which extend the physical distance PCIe Gen 5 supports using copper wires. The switch comes with a suite of debug and diagnostic features.
Dr. Debendra Das Sharma, Intel fellow and director of I/O technology and standards, said in a statement: “Intel’s upcoming Sapphire Rapids Xeon processors will implement PCI Express 5.0 and Compute Express Link running up to 32.0 GT/s to deliver the low-latency and high-bandwidth I/O solutions our customers need to deploy.”
Starboard Value sues former Quantum execs
Starboard Value has dropped its law suit against Quantum. However, in an amended filing, the activist investment fund, is still suing Jon Gacek former Quantum CEO, and former CFO Paul Auvil in the California Superior Court in Santa Clara County, alleging mis-representation and fraud.
In a 10-Q Filing, dated 27 January, Quantum said it “expects to continue to incur expenses related to this litigation, subject to potential offset from insurance. At this time, the Company is unable to estimate the range of possible outcomes with respect to this matter.”
Starboard bought Quantum shares between 2012 and 2014 and gained board seats. It agreed not to seek more control and to support Quantum’s slate of directors if company performance objectives were met.
Jon Gacek
Starboard alleges Gacek and Auvil artificially inflated Quantum’s earnings in its fiscal 2015 year to meet these objectives. Gacek and Auvil resigned in November 2017. Initially, Starboard included Quantum as a defendant, but Quantum rebutted this and Starboard has filed an amended complaint that mentions Gacek and Auvil onlyk
Shorter news items
Amazon S3 now supports AWS PrivateLink, providing direct access to S3 via a private endpoint within the customer’s virtual private network. This eliminates the need to use public IPs, configure firewall rules, or configure an Internet Gateway to access S3 from on-premises.
Videocams are watching you – and Scale Computing has software to store the images. It has announced HC3 hyperconverged infrastructure options for video surveillance, security and IoT edge applications.
Veeam has launched Veeam Backup for Google Cloud Platform, claiming ultra-low RPOs and RTOs. The software automates Google-native snapshots to protect VMs across projects and regions. Backups are stored in Google Object Storage for long-term retention.
Database virtualizer Delphix says a Covid-19-caused surge in demand accelerated its annual growth rate by over 85 per cent for the fiscal year ending January 2021, pushing it into non-GAAP profitability. Delphix also achieved a Net Promoter Score (NPS) of 89 during the year. The company now presents itself as a supplier of programmable data infrastructure.
NVMe-over-TCP supplier Lightbits Labs said it increased sales in 2020 by more than 500 per cent through a significant uptick in IaaS, SaaS, financial services, and video gaming customers.
We’re hiring!
VAST Data has hired Helen Protopapas as VP of finance, Tom Whaley as VP of sales, and Rick Franke as VP of global customer success, services and support. VAST said it is experiencing hyper-growth with accelerated global expansion and customer adoption. Whaley comes from NetApp and Franke from VMware.
Cloud storage service supplier Backblaze has hired Frank Patchel as Chief Financial Officer. Patchel has worked for multiple software as a service (SaaS) technology companies and overseen the successful sale of two businesses to public companies while serving as their president.
Isabelle Guis
SoftIron has appointed Phil Crocker as VP business development and channel with a focus on HyperDrive, the company’s software-defined storage system, which is based on Ceph. He joins from HPC storage supplier Panasas.
Commvault has appointed Isabelle Guis as CMO, to replace the departing Chris Powell. She was previously Salesforce’s VP for product marketing at Sales Cloud.
Database supplier SingleStore has appointed Oliver Schabenberger, former COO and CTO of SAS, as Chief Innovation Officer. SingleStore recently announced an $80m Series E investment round and a strategic partnership with SAS.
An Intel staffer who left the company to join Microsoft walked out the door with a USB stick holding 3,900 confidential files about the Xeon processor, his former employer alleges.
Intel accuses Dr. Varun Gupta of trade secrets file theft. According to documents filed in the US District Court in Portland, Oregon, Gupta worked at an Intel facility in Portland for almost ten years, in product marketing and strategic planning and business development.
He left in January 2020 to join Microsoft as a Principal for Strategic Planning in Cloud and AI. At Intel he had access to Xeon processor documents about pricing structure and strategies, parameter definition and manufacturing capabilities.
On his last day at the company, Gupta is alleged to have copied about 3,900 documents onto a Seagate FreeAgent GoFlex USB drive bearing an identified serial number. According to Intel he also copied information onto a Western Digital My Passport USB drive, also with an identified serial number. Some of the files were marked ‘Intel Top Secret and ‘Intel Confidential’.
In its court filing, Intel accuses Gupta of “deploying that information [in his role at Microsoft] in head-to-head negotiations with Intel concerning customised product design and pricing for significant volumes of Xeon processors”.
Specifically, he “used that confidential information and trade secrets to gain an unfair advantage over Intel in the negotiations concerning product specifications and pricing. Gupta had no way of knowing this information but for his access to it during his employment at Intel.”
An Intel security team began an investigation to determine the nature and scope of Gupta’s knowledge. With the assistance of Microsoft, “forensic analysis ultimately showed that Gupta had taken thousands of Intel documents, placed them on one or more of at least two USB drives (including the Seagate Drive), and accessed them on multiple dates throughout his employment by Microsoft.”
For example, “forensic analysis revealed that between February 3, 2020 and July 23, 2020, Gupta plugged the Western Digital Drive into his MS Surface at least 114 times.” Documents accessed included ”a slide deck relating to Intel’s confidential engagement strategy and product offerings for Xeon customised processors.”
After first denying he had the Seagate drive, “Gupta admitted to Microsoft that he did in fact have the Seagate Drive in his possession and only then that he turned it over to Microsoft for analysis.” Microsoft then commissioned a forensic analysis of the Seagate Drive. The Western Digital drive has not been found.
Intel seeks a jury trial and wants damages of at least $75,000, payment of its legal fees, and a restraining order preventing Gupta from using any confidential Intel information.
Earlier this week, Blocks & Files reported the sudden departure of Rubrik’s Chief Revenue Officer Brett Shirk and several senior execs from the company. Since then we learnt that head of engineering Vinod Marur has also left the company. We caught up with Rubrik CEO and co-founder Bipul Sinha to ask him what was going on.
On our phone call yesterday Sinha painted the picture of a company growing at such breakneck pace that it needs to transition to new exec leadership as it enters new phases of growth.
Bipul Sinha
“We are on an exponential growth curve,” Sinha told us. “Obviously if you think about Rubrik as a high growth startup [then] what high growth does is that an average company takes 10 years to get to a point where a high growth startup can do it in two, three, four years. And then you have to constantly re platform the company for the next phase of growth.“
He thinks Rubrik is in the half a billion to two billion dollars transition phase right now. The company goes from zero to $100m, $100m to $300m, $300m to $500m, and $500m to $2bn is the next phase of growth.”
As a consequence, “when you enter that next phase of growth, you have to really think about things like talent, leadership, how are we approaching the market, and things like that, and that always leads to to new talent coming in to really accelerate the strategic direction and lead the next phase of growth.”
He told us that this new growth phase required different sales leadership, but emphasised: “Brett is an incredible leader, and he’s really built a powerful sales engine that we have today. And that really sets the platform for our next phase.”
So how well is Rubrik doing right now? The full 2020 year and fourth 2020 quarter, ended January 29, set revenue records, according to Sinha, who noted that Rubrik had grown against the background of a data management market that is changing to a service and subscription orientation.
Bullet points from his comments include;
“We added 300-plus new employees just last quarter.”
“Our customer base is also increasing, very very rapidly. We now have over 3200 customers worldwide [and] our product is installed in over 55 countries around the world.”
“Customer spend on Rubrik is also growing very rapidly with over 200 customers with more than a million dollar spend.”
“We are going through the subscription transition of Rubrik. And quarter over quarter, our subscription product revenues are growing rapidly… we were nearly 70 per cent subscription last quarter.“
“So again, very very high growth and this is very important data for us because, as we mature our SaaS model, it is important to have subscription as the as the majority of our business.”
Enterprise SaaS supplier
“So what we’re seeing,” Sinha said, ‘is the thing is that we sell into the large enterprise segment, you’re not an SMB player and large enterprise segment when you’re selling like hundreds of thousands of dollars deals or millions of dollars deal it always has to be high touch.
“Even as a SaaS platform, you will have high touch sales, just like ServiceNow’s SasS platform. When you sell… millions of dollar deals to people [they do] not buy on phone or cell service budget because they want to understand the alignment vision, they are trusting you with their most important asset. It’s going to be a big deal.”
This “demands a different kind of sales process, a different kind of of sales leadership to definitely orient us in a different direction.”
New CRO and engineering head
Brian McCarthy.
Sinha said Rubrik is appointing a new CRO, Brian McCarthy, who currently occupies the same role at ThoughtSpot. McCarthy is a “leader who has long experience in software and SaaS, to really lead the company and double down on the direction that we are going.”
Arvind Nithrakashyap (known as “Nitro”), Rubrik co-founder and CTO, has replaced SVP Vinod Marur as the head of engineering. Sinha said: “So, Nitro as you know, has been my strategic partner from day one… He is the spiritual kind of guru of Rubrik engineering and… we brought him back to really strategically direct again the next round of Rubrik’s [growth].”
Regarding Marur, Sinha said: “We’re very thrilled about that our VP of Engineering, the role Marur had, has moved on to his next opportunity. And he has done a fabulous job of really building the engineering team and the management team and the structure and brought in like lot of great discipline and a structure from Google. And he he wanted to pursue his passion in the next company.”
So, like Shirk, Marur built a great team that, now needs new leadership. It’s a ruthless world in accelerated growth companies like Rubrik.
HPE has announced the Very Read-Optimised SSD replacement option for SATA-connected disk drives in Apollo, ProLiant and Synergy servers. The 2.5-inch and 3.5-inch VRO SSDs are plug-in replacements for the drives and deliver better TCO than 10,000rpm HDDs, the company claims.
“When solid-state drive (SSD) performance meets 10K hard-disk drive (HDD) price points, you get the best of both worlds,” HPE said in a Community Experts blog post this week.
“And for years, that’s exactly what HPE has been working toward: Enabling you to experience the breakthrough performance, reliability, and energy efficiency of SSDs on HPW ProLiant, Apollo and Synergy platforms – at the closest possible price to the HDDs. Now available to replace HDDs in popular workloads, that’s exactly what we’re excited to deliver.”
HPE VRO SSD
VRO SSDs are optimised for a typical mix of >80 per cent random reads and <20 per cent sequential writes (large block size). Example storage workloads include vSAN capacity tiers, NoSQL databases, business intelligence, Hadoop, analytics, object stores, content delivery, and AI and machine learning data lakes.
The VRO drives have a 6GBit/s SATA interface, and are available in 1.92TB, 3.84TB, and 7.68TB capacities in the 2.5in form factor and 3.84TB and 7.68TB versions in the 3.5-inch form factor. An HPE QuickSpecs data sheet also lists a 960GB 2.5-inch drive.
The drives come with a three-year warranty and a lifetime of 700 drive writes (which is represented by HPE as 0.2 drive writes per day). The random read IOPS performance is 51,000, with a peak maximum of 63,000, and random write IOPS are 12,600, peaking at 13,000.
The drives use 96-layer QLC flash and prices start at $739.99. For comparison, on Amazon a 2TB Barracuda Pro 7,200rpm disk drives costs $55.49 while a 1.2TB 10,000rpm Seagate Enterprise Performance disk drive retails at $190. The VRO SSDS have quite the price difference.
Benchmarks
The VRO SSDs deliver 7x faster object reads and 6x faster object writes at a lower TCO, according to a joint HPE Micron product brief. The claims are based on testing three Apollo 4200 Gen 10 servers, each fitted with either eight x 8TB HPE 7,200rpm 3.5-inch SATA disk drives or eight x 7.68TB VRO SSDs. The test environment was Ceph with object reads and writes tested.
The disk drive based Apollo system cost $152,760 while the VRO equivalents cost $234,384 – 53.4 per cent more.
The disk-drive Apollos delivered 3.0GB/sec read and 1.5GB/sec write bandwidth. The VRO SSD Apollos went much faster, hitting 23.1GB/sec read and 9.0GB/sec write bandwidth.
Divide the system cost by the read GB/sec number and this works out at $50,920 per GB/sec for the disk drive Apollos and $10,146.5 per GB/sec for the VRO version. This is the justification for HPE’s claim that VRO SSD Apollos delivered 5x lower cost per read GB/s at lower TCO.
VMware has added persistent container storage in a VMware Cloud Foundation update, via vSAN Data Persistence platform plug-ins with Cloudian and MinIO.
VMware Cloud Foundation (VCF) is a software bundle for deploying operating and managing hybrid clouds. The stack includes ivSphere, vSAN, SDDC manager, NSX, and the vRealize suite.
VCF 4.2 adds:
vSAN Data Persistence platform to manage S3-compatible object storage via support for Cloudian and MinIO.
vSAN HCI Mesh disaggregates compute and storage resources in a hyperconverged system, helping to share capacity across vSAN clusters and scale resources more efficiently,
NSX-T 3.1 Federation provides a cloud-like operating model for network admin staff, with centralised management, networking and policy configuration, and synchronised operational state across large-scale federated NSX-T deployments.
SDDC Manager Security Hardening means links between SDDC Manager and other components are more secure.
With NSX-T Federation, customers can use stretched networks and unified security policies across multi-region VCF deployments. This simplifies disaster recovery and moving workloads, VMware said.
The vSAN Data Persistence platform enables MinIo and Cloudian object storage to provision storage to Kubernetes-orchestrated containers. Cloudian HyperStore and MinIO Object Storage are available for VMware Cloud Foundation with Tanzu through this vSAN Data Persistence platform integration.
VCF 4.2 should be available between February and May. Customers can purchase the object storage offerings directly from Cloudian and MinIO.
Fujifilm is working on a petabyte tape cartridge. This behemoth could go into production from 2035, according to a slide deck in a company press briefing today.
A 1PB tape far exceeds the highest capacity on the LTO roadmap, which extends out from the-about-to-be-delivered LTO-9, with 18TB raw capacity, to LTO-12 and 144TB. Assuming a doubling every LTO generation, it would be LTO-15 when a 1PB+ tape was reached.
A 10PB raw capacity tape library today would need 830 LTO-8 tapes compared with 10 x 1PB tapes. This would increase the tape library’s capacity 83X over its capacity using LTO-8 cartridges.
In real life, tape cartridges hold data compressed at a 2.5:1 ratio, so a 1PB raw tape cartridge would have a 2.5PB effective capacity.
Today’s Fujifilm tape uses a Barium Ferrite magnetic layer. The company is likely to follow this with Strontium Ferrite media, which have smaller nano-particles that are less than 60 per cent of BaFe particle size. This increases areal density and hence capacity. The company has already demonstrated a 580TB raw capacity tape using Strontium Ferrite media.
For the 1PB tape cartridge, Fujifilm is developing technology based on Epsilon Ferrite, which has even smaller nano particles.
The company is also working on a F-MIMR (Focused Millimetre Wave-Assisted Magnetic Recording) technique. This excites the particles in the media so that data can be written and leaves them with stable bit values at room temperatures and in millimetre wave-free conditions.
Yesterday we reported VAST Data’s Howard Marks’s argument that hard drives are IO disks are IO sluggards, caching doesn’t work, and overall all-QLC flash array costs are the same as disk drive arrays.
VAST develops high end, high performance all-flash arrays. Infinidat also develops high-end, high performance arrays – but they are largely disk-based. Who better to ask to respond to Mark’s arguments?
Ken Steinhardt.
Field CTO Ken Steinhardt told us: “We’re a bit surprised that the blog mentioned in that article is so heavily focused on component device specifications to the exclusion of full storage systems. I suppose that’s the kind of discussion that some vendors prefer when discussing their product, but we believe that it misses the mark.”
“At Infinidat we believe that customer value is derived from much higher macro level system characteristics – specifically high performance, 100 per cent availability, replication capabilities, ease-of-use, ecosystem support, post-sales support effectiveness, environmental issues, and financial considerations – but not media.”
According to Steinhardt, disk vs SSD comparisons are mis-guided because it’s like trying to “compare two different cars to determine which one will win a race based upon the horsepower of the engines. The greatest determining factor will be the skill and intelligence of the driver behind the wheel, and not the car itself. Just look at Lewis Hamilton.”&
Infinidat software is the reason that its arrays are faster than all-flash competitors, he says. The software makes its disk drive system fast. For example, in February 2019 an InfiniBox F6000 array pumped out 1.43 million IOPS and 15.2GB/sec throughput. Since then, the company has added support for NVMe over Fabrics – and the arrays go even faster. NVMe-oF cuts the Ethernet link transfer time from about 100 microseconds to around 30 microseconds.
According to Steinhardt, Infinidat’s software, using a Trie (digital tree) algorithm, can “provide ultra-high performance at true hyperscale – as opposed to the more limited approaches of hashing algorithms or the more primitive use of binary trees that are found in all other storage products.”
Cache hits and misses
In his article, Marks wrote that caching cannot make up for disk drive slowness in PB-scale arrays: “You would think that with big hard drives delivering so few IOPS/TB caching the hot data and metadata is a good idea… and it is, but a cache can’t completely make up for a lack of IOPS in it’s backing store, especially when facing demand peaks.”
Marks points out: ”the latency of each cache miss is the 5-10ms it takes to read from disk; 20-40 times the 500μs SSD read latency.” Increasing the cache size is a dead-end . “You would think that you could just make the cache bigger, to keep the number of cache misses down, but that ignores the fact that cache effectiveness falls off dramatically with cache size.”
In response, Steinhardt boasts: “Infinidat is unique among enterprise storage offerings in its use of patented true Deep Learning with our Neural Cache software to deliver radically higher cache-hit ratios than any other storage product in the market.”
Disk drives cost less than SSDs
Steinhardt argues Marks is wrong on the future of storage costs, and has “either misleading or outright inaccurate opinions… Here’s an independent take on the future of storage costs [see chart below]. This is from Gartner, which is hopefully a more credible source, and one that doesn’t have an axe to grind in this debate, unlike some vendors.”
Disk or SSD – Infinidat is agnostic
According to Steinhardt the technologies that” offer the best combinations of performance, reliability, capacity, and cost today for each of those levels respectively are DRAM, Flash, and HDDs. But that’s just today. Any or all of them can (and likely will) change for those dynamics in the future as technologies change and newer technologies like SCM, QLC, PLC, HAMR, MAMR, etc. continue to evolve.
“Infinidat can potentially use any of them, indeed we’ve already demonstrated such and we’re not going to arbitrarily praise or slam any of them – that would be silly. The order of magnitude differences in cost and performance give each of them a role to play in multi-PB data centres”
To conclude, Infinidat thinks that the basic storage medium is much less important than the software driving an array. The company says its proprietary software ensures that Infinidat disk array is faster than all-flash rival, inviting us to draw the conclusion of what it could accomplish using all-flash media.
WekaIO, the very fast file access startup, this week released the Limitless Data Platform. As far as we can see this packages old news under a new label.
Weka claims Limitless Data Platform is storage without equal, helping accelerate outcomes and improving productivity in the hybrid cloud. So-called key enhancements include Snap2Object now supporting snapshots being stored on-premises or in the public cloud or both.
However this snapshot feature was previously announced in March 2020, when Weka stated: “Weka’s unique snap-to-object feature allows users to easily create a replica of the production data and instantly push it to any S3 object store — on-premises or in the cloud — enabling snapshot-based replication.”
Other new features mentioned in this announcement are also pre-existing. For example;
Enterprise-Grade Security using XTS-AES 512 bit keys. This was mentioned in a blog by then CTO Andy Watson in October 2019.
Any Hardware: Weka can be deployed on industry-standard AMD or Intel x86-based servers supporting PCIe v4. AMD and PCIe gen 4 support was mentioned in a Weka and Supermicro document in November 2020.
Your Cloud. Your Way. Weka’s Cloud-Native architecture enables customers to dynamically scale up or scale down resources in the Cloud-based on workload requirements. This was mentioned in a 2020 ESG review which said: “The storage capacity can scale by adding NVMe SSDs to a server, by adding more servers with NVMe devices, or by allocating additional S3 capacity, and the performance can scale by allocating additional CPU cores or adding server instances to the global namespace.”
Weka’s Limitless Data Platform announcement did not mention a new version of its software. We’ve asked the company what is actually new in this announcement. Ken Grohe, President and Chief Revenue Officer, told us: “No new software. We are taking this occasion to showcase the 16+ features that we have delivered to the market over the last 8 months that have been hidden until 2/2/21.”
He added: “The brutal truth is that people still find us from speed, and buy us the first time for speed reasons. But their next 3-8 purchase over the next 2 years post initial deployment now are a repeatable and measurable 7-9 times the initial $$ and PB size deployments across NVMe tier, Object store, and Cloud as validated by the leader in autonomous driving, Life Sciences, and the US Federal Government etc and are expanding their Weka footprint for Simplicity and Scale reasons.”
Disk drive IO limitations will cripple the use of larger disk drives in petabyte-scale data stores, according to VAST Data’s Howard Marks, who says Flash is the only way to go as it can keep up with IO demands.
Marks is a well-regarded data storage analyst who these days earns his crust as “Technologist Extraordinary and Plenipotentiary” at VAST Data, the all-flash storage array vendor. Happily for the industry at large, he continues to publish his insights via the VAST company blog.
Howard Marks
In his article The Diminishing Performance of Disk Based Storage, Marks notes that disk drives have a single IO channel which is fed by a read/write head skittering about on the disk drive platters’ surface. This takes destination track seek and disk rotation time, resulting in 5ms to 10ms latency.
This is “20-40 times the 500μs SSD read latency” and limits the number of IOs per second (IOPS) that a disk drive can sustain and also the bandwidth in MB/sec that it can deliver.
“As disk drives get bigger, the number of drives needed to build a storage system of any given size decreases and fewer drives simply deliver less performance,” Marks writes. He illustrates this point about overall IOPS and bandwidth with a table based on a 1PB array with 225MB/s of sustained bandwidth and 150 IOPS per drive:
Disk drive capacity numbers and performances in a 1PB array
This table shows 1,000 x 1TB drives are needed for a 1PB array, dropping to 50 x 20TB drives. He concludes: “As disk drives grow over 10 TB, the I/O density falls to just a handful of IOPS per TB. At that level, even applications we think of as sequential, like video servers, start to stress the hard drives.”
I-O. I-O. It’s off to work we go (slow)
He says disk drives cannot get faster by spinning faster because the power requirements shoot up. Using two actuators (read/write heads or positioners in his terms) per disk effectively splits a disk drive in half and so can theoretically double the IOPS and bandwidth. “But dual-positioners will add some cost, and only provide a one-time doubling in performance to 384 IOPS/drive; less than 1/1000th what an equal sized QLC SSD can deliver.”
Marks also points to the ancillary or slot costs of a disk drive array – the enclosure, its power and cooling, the software needed and the support.
He sums up: “As hard drives pass 20TB, the I/O density they deliver falls below 10 IOPS/TB which is too little for all but the coldest data.”
It’s no use constraining disk drive capacities either.
“When users are forced to use 4 TB hard drives to meet performance requirements, their costs skyrocket with systems, space and power costs that are several times what an all-flash system would need.”
“The hard drive era is coming to an end. A VAST System delivers much more performance, and doesn’t cost any more than the complex tiered solution you’d need to make big hard drives work.”
Comment
As we can see from the 1PB data store example that Marks provides, 50 x 20TB drives pumps out 14GB/sec total while 500 x 2TB drives pumps out 141GB/sec. Ergo using 50 x 20TB drives is madness.
But today 1PB data stores are becoming common. They will be tomorrow’s 10PB data store. Such large arrays will need more drives. Therefore, as night follows day, and using Marks’ own numbers, the overall IOPS and bandwidth will grow in proportion.
A 10PB store would need 500 x 20TB drives and its total IOPS and bandwidth can be read from the 2TB drive size column in Marks’ table above – i.e. 141GB/sec. Data could be striped across the drives in the arrays that form the data store and so increase bandwidth further.
Our caveats aside, Marks raises some interesting points about disk drive scalability for big data stores. Infinidat, which builds big, fast arrays incorporating high-capacity hard drives, is likely to have a very different take on the matter. We have asked the company for its thoughts and will follow up in a subsequent article.
