Home Blog Page 333

It’s that time of year again: Gartner flings distributed files systems and objects across the MQ

It’s all change at Gartner’s distributed files systems and objects Magic Quadrant, with three suppliers leaving, one arriving, and major and minor movements across the board.

The leaders have all done well as have niche players Huawei and DDN. Gartner said end users have been reporting unstructured data as growing between 30 per cent and 60 per cent year over year.

Its MQ report states: “The steep growth of unstructured data for emerging and established workloads is now requiring new types of products and cost-efficiencies. Most products in this market are driven by infrastructure software-defined storage (iSDS), capable of delivering tens of petabytes of storage.

“iSDS can also potentially leverage hybrid cloud workflows with public cloud IaaS to lower total cost of ownership (TCO) and improve data mobility. New and established storage vendors are continuing to develop scalable storage clustered file systems and object storage products to address cost, agility and scalability limitations in traditional, scale-up storage environments.”

A Gartner MQ is like a doubles tennis court with niche players and visionaries one side of the net, and challengers and leaders the other. The objective, rather than blast balls across the net and out of court, is to move into the leader’s section of the court and as far from the net and to the edge of the court as possible. 

Here is the October 2020 MQ: 

Pretty straightforward with 14 suppliers but there is an oddity. IBM has bought Red Hat so there should really be a single IBM entry with the Red Hat one removed. A look at last year’s MQ will show who’s left the court, who has arrived, and how players have moved:

The leavers include Western Digital, with Quantum acquiring its ActiveScale product and appearing for the first time. SUSE has departed as has SwiftStack, which was bought by Nvidia.

By doing some image magic we can overlay the two charts and see how players have moved:

The 2019 entries are in grey while the 2020 entries are in blue.

In the leaders’ box leading leader Dell has moved a good way to the right, improving its completeness of vision, ditto second placed IBM. Both Scality and Qumulo have also moved significantly to the right.

Huawei has migrated from the niche players’ box to become a challenger. Cloudian and Hitachi Vantara have both moved slightly, while Quantum’s position shows less ability to execute compared to Western Digital, whose position it inherited/acquired.

The niche players box show DDN moving up and to the right, improving on both the vision and execution axes. Caringo has moved in the opposite direction and Inspur has gained some so-called “vision completeness”.

What do the vendors say?

A Qumulo spokesperson positioned the company in David and Goliath terms: “Companies like us … are rocket ships of growth and taking customers from all over the world out of the hands of all kinds of competitors, but most often, they are coming from the dinosaur vendors.”

IBM said it had been in the leader’s box for five years. 

NetApp adds S3 to ONTAP, SW-only SolidFire and pay-as-you-grow pricing

NetApp has added a dose of cloudification to its product lines with a series of announcements covering its core ONTAP software, SolidFire all-flash array software, and Keystone subscription offering.

The company is now describing itself as cloud-led, data-centric and operating in a hybrid on-premises and multi-public cloud environment.

César Cernuda.

NetApp’s new president, César Cernuda, hired in May, emitted this prepared quote: “With its rich data-centric software innovation, NetApp is uniquely positioned to help organisations quickly adapt and sustainably transform in today’s hybrid cloud world. Now, we make it easier for them to develop applications in the cloud, move applications to the cloud, or create cloud-like experiences on premises.”

The latest version of ONTAP, v9.8, adds:

  • FlexCache support for SMB,
  • SnapMirror Business Continuity (SNBC) provides continuous availability with instant failover for workloads like Oracle and SAP,
  • S3 protocol support for small-scale object workloads complementing separate StorageGRID object storage offering,
  • IPsec inflight data encryption,
  • FabricPool cloud tiering gets FAS hard disk drive systems as an added hot tier source,
  • Sending snapshot copies to S3 object storage on-premises or in the public cloud,
  • One-click firmware upgrades,
  • Improved file system analytics.

NetApp said ONTAP storage efficiency had increased by “up to” 33 per cent, without providing details. It claimed (access) temperature-sensitive storage efficiency further optimised cold data to reduce a customer’s storage footprint.

The SolidFire Element all-flash array’s noisy neighbour-killing software is now available in SW-only form to run on commodity server hardware. This new Enterprise SDS (eSDS) offering is packaged in containers and contains selectable components. Qualified server hardware is listed in NetApp’s Interoperability Matrix Tool.

The latest Element SW version adds:

  • Drive lock + software encryption,
  • Software encryption at rest,
  • Multilayer volume security with access group + CHAP + VLANs,
  • Multifactor authentication,
  • Centralised key management through industry-standard protocols.

An updated Keystone Flex subscription provides so-called “pay-as-you-grow” pricing and public cloud integration through the NetApp partner ecosystem. NetApp claimed: “Rapid service adoption with field-proven deployments enables us to bring the cloud to you in as little as 2 weeks.” 

DDN hauls NVMe storage chassis to Tintri products

DDN has injected its all-NVMe flash storage chassis into its Tintri product line to start the cost-cutting process of unifying product hardware across its two divisions.

Tintri was acquired by DDN, along with Western Digital’s IntelliFlash and Nexenta’s software-defined storage, in a series of acquisitions in 2018 and 2019. DDN formed an At-Scale division to produce and sell its original DDN array products, and an Enterprise division to sell the Tintri, IntelliFlash and Nexenta products under an Intelligent Infrastructure theme. It has just announced new Tintri T7000 and IntelliFlash H-series arrays.

IDC Research VP Eric Burgener provided a quote for DDN: “Moving to a common hardware platform for both their At-Scale and Enterprise business units leverages economies of scale in engineering, and will offer reliability, manageability and time-to-market advantages with new releases for customers.”

The Tintri VMstore array line has a T1000 all-flash system for remote and branch offices along with an EC6000 enterprise-class line for data centres. It’s heart is a 2 rack unit (RU) chassis containing SAS SSDs.

These two are joined by a new T7000 which uses DDN hardware, with 10 drive bays for NVMe SSDs and 30 per cent faster performance than the EC6000. The drives can be self-encrypting and Tintri offers non-disruptive drive-by-drive expansion. The OS allows FIPS-compliant software encryption with KMIP. 

Tintri VMstore T7000.

No datasheet has been made available so the speeds and feeds details are not yet known.

Where the Tintri VMstore arrays are targeted at specific virtual server and SQL Server application workloads, the IntelliFlash arrays are general in nature. They offer both all-flash and hybrid flash/disk storage for database, virtualisation and data protection workloads needing both block (FC, iSCSI) and file (NFS, SMb3) data access.

The product range consists of the hybrid T-Series, mixing SSDs and disk drives for fast performance with bulk capacity, the N-Series all-flash NVMe higher performance systems, the HD-Series for all-SAS SSD storage, and the H-Series mixing NVMe SSDS and disk drives for an even higher performance hybrid system than the T-Series.

Tintri has announced new H6100 and H6200 systems and a data sheet says the H6100 has 23 to 92TB  raw NVMe SSD capacity and 96 to 1344TB raw disk capacity. The larger H6200 has 46 to 368TB  raw NVMe SSD capacity and 720 to 6,480TB raw disk capacity.

Tintri IntelliFlash H-Series.

An image of the H-Series shows two cabinets and we understand the upper one is the same basic chassis as is used for the T7000, albeit with a different bezel.

DDN says the H-Series autonomously optimises NVMe SD-to-HDD ratios to meet performance and capacity needs. Hybrid modular expansion shelves provide a route to 25PB of effective capacity in 18RU.

Data services include snapshots, cloning, replication, encryption, deduplication, compression, thin provisioning, and backup integration. Concurrent SAN and NAS access is supported as is concurrent NFS and SMB access.

Tintri supplies IntelliFlash analytics and system monitoring capabilities which apply AI, rules engines, predictive decision making, and data-driven workflows to improve overall system health, reliability, and predictability. 

H-Series general availability is planned for the fourth 2020 quarter while the T7000 should arrive in the first 2021 quarter.

Watch out Samsung: SK hynix confirms buy of Intel’s NAND fab and SSD business

South Korean memory semiconductor supplier  SK hynix will pay $9bn to buy Intel’s NAND foundry and SSD business, it has been confirmed.

SK hynix is paying 208 trillion won (about $9bn) for the bulk of Intel’s Non-Volatile Solutions Group (NSG), meaning the Dalian fab, the NAND wafer and components operations, NAND IP and staff, and the SSD business, leaving the group with its 3D Point operations.

Bob Swan, Intel’s CEO, offered a canned statement: “I am proud of the NAND memory business we have built and believe this combination with SK hynix will grow the memory ecosystem for the benefit of customers, partners and employees. For Intel, this transaction will allow us to further prioritize our investments in differentiated technology where we can play a bigger role in the success of our customers and deliver attractive returns to our stockholders.”

Those attractive returns have been lacking from Intel’s NAND business. Swan may be proud but Intel is walking away from a business that couldn’t deliver the return on investment it wanted.

SK hynix CEO Seok-Hee Lee also issued a statement: “By taking each other’s strengths and technologies, SK hynix will proactively respond to various needs from customers and optimize our business structure, expanding our innovative portfolio in the NAND flash market segment, which will be comparable with what we achieved in DRAM.”

Lee’s company is getting an SSD product line including NVMe interface SSDS using QLC flash. Intel is retaining its Optane 3D Point non-volatile memory business. It sold its interest in 3D XPoint manufacturing to Micron in late 2018, and its Optane business could have entered profitability after a lot of investment since its July 2015 announcement.

Analyst Mark Webb told us: “Intel Optane Persistent memory is the big strategic item and it is managed by DCG (Data Centre Group), not NSG.”

The two companies will seek government approvals and this could be obtained by the second half of 2021. The deal is structured in two stages. In phase one, and consequent on government approval, SK hynix will pay Intel $7bn and get the NAND SSD business (including NAND SSD-associated IP and employees), as well as the Dalian facility.

A second phase will see SK hynix paying Intel $2bn and acquire IP related to the manufacture and design of NAND flash wafers, R&D employees, and the Dalian fab workforce. This is planned to take place in March 2025. Until then Intel will continue making NAND wafers at Dalian and retain all IP related to the manufacture and design of NAND flash wafers.

SK hynix will use the acquisition to grow its NAND chip and SSD businesses, particularly its enterprise SSD business, with a focus on higher-value SSDs.

It said that, in the six months ended June 27, 2020, Intel’s NSG earned $2.8bn revenues from its NAND business, which provided $600m operating income.

Current NAND market supplier revenue shares.

The deal will see the SK hynix – Intel combination having a 23 per cent NAND market revenue share, second only to Samsung with 31 per cent. As Kioxia (17 per cent) and Western Digital (16 per cent) jointly operate a NAND foundry complex their combined share is 33 per cent. Micron at 13.7 per cent is someway behind.

Industry research outfit TrendForce said that, in terms of product competitiveness: “SK hynix has an advantage in the mobile market, including eMCP and eMMC products, accounting for more than 60 per cent of SK hynix’s total NAND Flash revenue in 2019. On the other hand, Intel has been performing superbly in the enterprise SSD market. Not only is Intel on par with Samsung in enterprise SSD, but it has also captured more than 50 per cent of the Chinese market. Enterprise SSD yields the highest profitability among the entire range of the company’s NAND Flash end-products.”

Webb does not believe that mooted industry consolidation benefits will be very meaningful: “This is not as great for NAND consolidation as one might think. Intel sells very little open market NAND to mobile or card/USB market. Intel sells SSDs and raw memory to enterprises (Google, Amazon, Oracle, etc).

“As such, if this were to move to Hynix, it will not change the pricing competition in the card market and mobile market which along with consumer SSDs are the most competitive. Intel is a small player outside of enterprise SSDs.”

Intel will use the overall $9bn proceeds to build up its artificial intelligence, 5G networking and intelligent, autonomous edge offerings.

Webb said: “I believe Intel is still not allowed to make announcements during its quiet period (30 days before earnings). I expect [an] announcement on Thursday,” when Intel announces its Q3 2020 earnings.

SK hynix could buy Intel NAND business

The WSJ is reporting Korean DRAM, NAND and SSD maker SK hynix could buy some all of Intel’s NAND business operation for around $10bn.

Details are few but a deal could be announced next Monday. 

Intel is the sixth and smallest NAND maker in terms of revenue market share, having 11 per cent of the market, behind leader Samsung with 31 per cent, Kioxia with 17 per cent, Western Digital with 16 per cent. Micron with 13.7 per cent and SK hynix itself with 12 per cent.

A seventh supplier, China’s YMTC, has negligible market share.

In March, Intel CFO George Davis said Intel was unable to sell enough SSDs to make a profit from the 3D NAND chips it made in its Dalian, China, foundry. 

Intel SSD.

Wells Fargo senior analyst Aaron Rakers told subscribers that Intel NAND operating income was -$340m for the trailing 12 months ending 2Q20, according to DRAMeXchange.

Intel was, Davis said, exploring options such as closing its foundry and buying in chips, or selling chips to third parties or buying in SSDs from a third party.

A Chinese media report posted in July last year said: “Hynix plans to acquire Intel’s Fab 68 factory in Dalian, China.” The paper quoted a person familiar with the matter as saying: “According to the current progress, Hynix is ​​negotiating with Intel. Hynix wants to acquire the entire Intel Dalian factory and 3D NAND business. Intel only retains XPoint-related technologies.”

A Reuters report said Intel would sell both its foundry and its NAND SSD business to SK hynix but keep its Optane 3D XPoint business. That would shrink its entire Non-Volatile Solutions group business unit drastically.

It appears likely that SK hynix is negotiating to buy Intel’s Dalian fab. Intel could then buy in chips from SK hynix if it wants to continue selling SSDs. It is conceivable that Intel is considering selling its entire NAND business to SK Hynix but that would be a drastic step.

Rakers suggested a consolidation from seven to six NAND manufacturers would be a good thing in terms of overall industry profitability. He thinks a supplier consolidation move, like SK hynix buying Intel’s fab, would also provide opportunities for competitors to attack Intel’s SSD customer base.

Intel reports its 3Q20 earnings on Thursday and we may hear more then.

Your occasional storage digest with MariaDB, Chief Data Officers and more

The digest this week is cloud-centric with MariaDB updating its SkySQL product and Nutanix refreshing its Era database management offering. But we’re also startled to learn that Chief Data Officers are a thing.

Chief Data Officer

We note that a new BPI Network report, entitled “Bringing Simplicity to Multi-Cloud Complexity: Advancing Data Connectivity, Logistics and Value,” highlights the growing importance of the Chief Data Officer (CDO) in the C-suite and the need to expand the value and accessibility of data and analytics across distributed enterprise organisations. It says CDOs of large and mid-sized companies are gaining competitive advantage through analytics and activation of cloud-based data repositories.

Who knew CDOs existed? It’s yet another example of Chief-something- Officer title creep away from CEO and CEO to lesser corporate responsibility beasts such as CIO, CTO, CSO (Strategy), and CPO (Product or People).

Is there no end to this? Will all EVPS and SVPs end up with a Chief-something-Officer title? Will I someday be called the B&F CWO – Chief Word Officer? Title-creep is fun to watch.

MariaDB updates SkySQL

MariaDB’s SkySQL cloud database has its core software updated to Maria DB Platform X5 to add distributed SQL and more capabilities.

Michael Howard, MariaDB’s CEO, issued a quote about first generation cloud databases: “The current landscape requires a smorgasbord of cloud services to get a single job done – AWS RDS for simple transactions, Aurora for availability and performance, Redshift for cloud data warehousing and Google Spanner for distributed SQL. SkySQL gives you all these capabilities in one elegant cloud database that delivers a consistent MariaDB experience regardless of the way you deploy it.”

SkySQL supports the latest versions of MariaDB Enterprise Server, advanced database proxy MaxScale and smart engines ColumnStore and Xpand and offers:

  • MariaDB Platform for distributed SQL: Xpand is a new smart engine that delivers distributed SQL through MariaDB Enterprise Server. This functionality is now also available in SkySQL with elastic scale and it automatically rebalances data hotspots for optimum performance.
  • MariaDB Platform for analytics: SkySQL includes a distributed cloud data warehouse that provides massively parallel processing (MPP) for scalability and high availability on large datasets. 
  • End-to-end security: SkySQL now enforces secure SSL/TLS connections for any database access, avoiding exposure of data due to insecure defaults or configuration choices.
  • Reduced complexity for application development: SkySQL provides a single connection point for applications rather than exposing individual database instances, primaries or high availability replicas. It manages read/write-splitting, failover and application session migration.
  • Expanded monitoring: SkySQL monitoring shows the status and all vital metrics for database instances and is highly customisable. The monitoring tool is updated to support the new topologies enabled in this new release of SkySQL such as Xpand for distributed SQL.

SkySQL’s pricing is all-inclusive with no up-charges for high availability setup, failover, backups or a database proxy for a single connection point.

Shorts

Cyber-and-data-protector Acronis has released Acronis Cyber Infrastructure 4.0 with more than 350 new capabilities and enhancements. These include erasure coding which improves writing performance and latency for hybrid and all flash tiers up to 50 per cent. Enhancements to storage networking have increased the performance of multi-thread workloads – with random reads improving up to 40 per cent. 

DDN, announced Brookhaven National Laboratory, one of 10 national laboratories overseen and primarily funded by the Office of Science of the U.S. Department of Energy (DOE), has selected its A3I AI400X all-NVME flash appliance storage for its Advanced Computing Lab. The AI400X will be used to provide a collaborative environment for scientists and technologists from government laboratories and academia agencies. 

Long-term retention supplier Falconstor has a support and migration program for IBM ProtecTIER customers. IBM is end-of-lifing its ProtecTIER virtual tape library (VTL) snf yjhr process has been made difficult by the Covid-19 pandemic.  FalconStor’s StorSafe product will send deduped backups previously handled by ProtecTIER directly to IBM on-premises storage systems and IBM COS.

Taiwan’s Gigabyte Technology has announced six new rack and high-density servers with AMD EPYC processors, PCIe Gen 4, boosted flash capacity, and support for tri-mode Broadcom storage adapters. They support PCIe, SAS or SATA devices in bays connected via an adapter in a PCIe slot. 

Memory cached disk drive array supplier Infinidat says it has nearly 400 global channel partners. It has held over 60 virtual events for the channel during the last six pandemic-affected months with a pivot to 100 per cent virtual and launched a new partner portal.

Digi-Key Electronics has added Netlist to its Marketplace. Netlist provides Digi-Key customers with a range of 2.5-inch U.2, half-height, half-length (HHHL) and M.2 NVM Express (NVMe) SSDs.

Nutanix has announced v2 of its Era database management offering which runs cross clouds and clusters and enables database to be delivered as a service. Era supports support for Oracle, Microsoft SQL Server, MySQL, and MariaDB, and now has support for Postgres and SAP HANA.

Object storage supplier Object Matrix announced a reseller agreement with Content Networks in Poland.

The UK’s Storage Made Easy (SME), announced its renewal for the Crown Commercial Service’s G-Cloud 12 scheme. Employees of government affiliated bodies can continue to access existing data sources and in-house servers and exploit public services such as Amazon S3, Azure, Office 365; remotely, safely and securely.

People

MariaDB Corporation announced the appointment of Amir Ameri as Chief Financial Officer, and Paul Jenkinson as Board Member and Audit Committee Chair. This news comes after MariaDB’s $25m cash raise in July to grow its SkySQL database-as-a-service platform.

Hybrid cloud data warehouse company Yellowbrick Data has hired Mark Cusack to be its CTO. He was previously VP for data and analytics at Teradata where he spent 6 years.

ESG validates WekaIO’s file system with benchmark trio: NetApp and Intel have individual wins

Analyst and consultancy firm ESG has validated WekaIO’s scale-out, parallel file system performance in three benchmarks – SPEC SFS 2014, IO-500, and STAC-M3 – finding it generally goes a lot faster than competing systems. But Intel has actually trounced Weka on the IO-500 test.

B&F has seen an ESG document: “WekaFS: Storage for Modern Exascale Workloads in the Enterprise,” which describes ESG’s findings.

SPEC SFS 2014

There are five different SPEC SFS 2014 tests reflecting different application environments in which file storage is used: SW Builds, VDA (streaming data), EDA (Electronic Design Automation), VDI and Database (OLTP database consolidation). Not all vendors submit results in all five categories. 

ESG looked at Weka’s performance in four of them, leaving out VDI, and comparing Weka to the next best result:

ESG chart showing four SPEC SFS 2014 category results.

We have a B&F chart showing more suppliers’ scores in the five categories:

B&F chart of composite SPEC SFS 20-14 results.

ESG said that, although NetApp scored higher than Weka in the SW Builds benchmark, 6,200 vs Weka’s 5,700, Weka produced its builds faster, having a 0.63ms average latency at 5,700 builds vs NetApp’s 2.8ms at 6,200 builds.

NetApp’s overall response time was 0.83ms while Weka’s was 0.26ms.

1O-500

The IO-500 benchmark computes a score using file data and metadata access tests for high-performance computing (HPC) systems. ESG said: “IO-500 scoring is derived from the square root of the product of metadata IOPS and throughput” and compared Weka’s result to three other systems: 

ESG chart of Weka IO-500 scores.

It declared: “Weka’s IO-500 results on AWS are significantly higher than other file system solutions whether on-premises or running in the cloud.”

Not so fast ESG. Intel scored higher still with an August 2020 result using its DAOS filesystem, 1,792.98 vs Weka’s 938.95 – nearly twice as fast. Intel’s score was helped by it using Optane SSDS and Optane Persistent Memory (DIMMs). Intel’s previous best IO-500 result was 933.64. 

BF chart of IO-500 scores with added Intel DAOS top line.

Weka is relying on parallelism for its IO-500 result; it used 345 client nodes. The Intel DAOS Optane score needed just 52 nodes. If Weka wants to get past this 1,792.98 IO-500 score it looks as if it will have to use Optane Persistent Memory and SSDs to do it.

STAC-M3

The third benchmark ESG looked at was the STAC-M3 suite, used in the financial services industry. The STAC acronym stands for the Securities Technology Analysis Center. 

ESG STAC-MC chart. Shorter lines are better.

ESG deployed an an eight-node WekaFS cluster on the AWS public cloud using 8, 10, 12 and then 16 i3en.24xlarge all-flash instances. Throughput scaled linearly as nodes were added to the cluster.

The all-flash NAS in the chart was a Dell Isilon F800 with PowerEdge servers. The NVMe-oF San refers to a Vexata VX-100 all flash array with Intel 22-core Xeon E5-2699 servers. The Optane SSD server was a Lenovo ThinkSystem SR950 with 4 x Xeon Platinum 8180 CPUs and 6TB of Intel DC P4800X (Optane) SSD capacity bulked out with 56TB of Intel DC P4510 (3D NAND) SSDs.

The ESG document also looks at extending a file system with object storage where Weka gives a good account of itself. Expect Weka to publish the ESG document quite soon.

Optane in the membrane: How Intel’s memory-storage technology faces two ways

Janus
The Roman God Janus

Analysis Like the Roman god Janus, Intel’s Optane technology faces two ways, to memory and to storage, and that has made adoption harder, because it is neither memory fish nor storage fowl, but both.

Why should this matter?

There is a memory/storage hierarchy forming a kind of continuum from high-capacity/slow access/low-cost tape at the bottom to low capacity/ extremely fast access/expensive level 1 processor cache at the top. Jim Handy, a semiconductor analyst at Objective Analysis, has created a log-scale diagram to show this with technologies placed in a space defined by bandwidth and cost axes.

Object Analysis’ Memory/Storage hierarchy diagram by Jim Handy.

The technologies form two types: storage with tape, disk drives and SSDs , and memories, with DRAM, and three levels of CPU cache. The chart shows PM, persistent memory or 3D XPoint, at the point where the storage devices give way to memory devices.

Memory devices are accessed by application and system software at the bit and byte-level using load/store instructions with no recourse to the operating system’s IO stack. DRAM memory devices are built in DIMMs (Dual Inline Memory Modules).

Storage devices are accessed at the block, file or object level, meaning groups of bytes, through an operating system’s IO stack. This takes a lot of time compared to memory access.

Optane Persistent Memory.

Whenever new storage devices come along, such as a new type of disk or a new type of 3D NAND SSD, they are accessed in the same way – through the storage IO stack. Storage device accesses generally use SATA, SAS or NVMe protocols across a cable between them and the host’s PCIe bus or, with NVMe, a direct link to the PCIe bus.

New types of memory, such as DDR5, use memory protocol, the load/store instructions, like their predecessors – DDR4, DDR3 and so on.

3D XPoint can be accessed either as storage, with Optane SSDs, or as memory, with Optane Persistent Memory (PMEM) products built in the DIMM form factor. That makes it two products.

Accessing it as storage is basically simple; it’s just another NVMe drive, albeit faster than flash SSDs.

Accessing it as memory is hard because it can be done in five different ways. It starts with Memory Mode or App Direct Mode (DAX or Direct Access Mode). DAX can be sub-divided into three options; Raw Device Access, access via a File API, or Memory Access. The File API method then has two further sub-options; via a File System or via a Non-volatile Memory-aware File System (NVM-aware).

These multiple PMEM access modes mean that applications using Optane PMEM have to decide which ones to use and then produce code and test it, which takes months of effort.

Let’s make a couple of extra points. As storage, an Optane SSD is neither fast enough, nor cheap enough (cost/GB) to replace NVMe SSDs. The 64-layer and 96-layer SSDs currently shipping are less expensive to make than 2-layer Optane SSDs. The coming 4-layer Optane drives should help to reduce the gap between Optane and NAND SSDs.

Optane SSD.

Optane PMEM is slower than DRAM and its appeal is based on servers having more total memory capacity with Optane (4.5TB vs 1.5TB) and running more applications/VM faster than a maxed-out DRAM-only server. That means Optane PMEM is a high-end application-centric choice and not a generic I-want-my-servers-to-go-faster choice.

The breadth of such application support is expanding all the time and this will broaden Optane PMEM’s market appeal.

In effect, Optane PMEM is not fast enough to be a generic DRAM substitute and various access mode tweaks are needed to use it to best effect.

Note. Janus was the Roman god of beginnings and transitions, dualities and doorways. He is often depicted as having two faces looking in opposite directions.

Flash Memory Summit 2020: Tech hints on computational storage, ruler drives and Smart NICs

The annual Flash Memory Summit has gone virtual and a freshly unveiled agenda provides tantalising hints about what’s coming and what may not be coming down the storage pike.

The first thing to catch our eye was session A-9: Computational Storage Increases System Throughput and Scalability (Computational Storage Track). The computational storage idea is to stick small processors (ASICs, FPGAs or Arm CPUs) and software on SSDs to carry out low-level and repetitive storage tasks such as encryption, compression, deduplication and video transcoding.

Andy Walls.

Startups such as Eideticom, NGD and ScaleFlux are developing products. The track is being organised by JB Baker, senior director of product management at ScaleFlux, and there are three presenters: chief scientist Chuck Sobey from consultancy Channel Science, Eideticom CTO Stephen Bates, and IBM fellow and flash storage CTO and chief architect Andy Walls.

It was intriguing to see Walls listed there, as IBM has expressed no public interest in computational storage and has no product yet. Is that going to change? Walls’ bio says: “He is currently defining next-generation products that can be used in traditional SAN environments and clouds, and also by emerging workloads.” 

The second eye catcher was Session D-8: Using the New EDSFF (E3) SSDs Effectively (SSDs Track). It’s a panel session discussing  adoption of ruler format SSDs – longer and narrower than 2.5-inch drives. Thera are three panel members: distinguished engineer Bill Lynn from Dell, advanced storage technologist Paul Kaler from HPE, and SSD industry standards director John Geldman from Kioxia.

Are we actually going to see Dell and HPE produce servers and/or storage chassis containing EDSFF drives made by Kioxia in 2021?

Session D-10: SmartNICs: The Key to High-Speed Converged Networks (Hyperscale Applications Track) has panel members from Nvidia, Broadcom, Pensando and Intel. The concept of offloading infrastructure processing tasks from application-oriented X86 servers in the data centre like security, networking and low-level storage service to smart NICs is gaining ground and we will hear more about this in 2021.

What’s not on the FMS2020 agenda is anything about Micron’s QuantX (3D XPoint) activities or penta-level cell (5 bits/cell) flash. There are sessions discussing Intel’s Optane and QLC (4 bits/cell) flash but nothing listed about PLC or Micron’s QuantX technology.

Two Rubrik execs depart, including CFO Murray Demo

Blocks and Files has learned that Murray Demo, Rubrik’s CFO, and Chief People Officer Jeff Vijungco are leaving the company.

Murray Demo

Rubrik told us: “After a long and successful career, Murray Demo will be retiring from his day-to-day leadership role as of early next year. This is part of a planned succession process as Kiran Choudary SVP, Finance & Strategy, is being promoted to CFO.” 

Rubrik is a six year-old startup funded to the tune of $552m, and valued at $3bn-plus. It produces backup and data management software and in February claimed to have a $600m run rate from more than 2,500 customers.

Demo came onboard at Rubrik in January 2018, after a six-year stint at Atlassian where he was the CFO. During his time there, the company IPO’d, raising $426m. There was speculation when Rubrik hired him that a IPO might be in the offing.

Choudary joined Rubrik in August 2018, also coming from Atlasssian, where he spent five years and became VP Finance and Strategy.

Rubrik’s spokesperson said: “Kiran brings an impressive background and experience to the role having led corporate finance at companies like Atlassian. Murray will remain in an advisory role with Rubrik following  his retirement and will continue to support the company through the remainder of the year in transition.” That advisory role will be a paid position.

Jeff Vijungco.

On the HR front, “Jeff Vijungco  will be departing Rubrik to spend more time with his family and begin his next chapter as a partner at an executive search firm. We have begun a search for a Chief People Officer and have a strong interim team in place that will continue to lead and implement  our talent, people development, benefit, and diversity initiatives. We wish Jeff great success in his future endeavours and thank him for his contributions.”

Bits and bytes in bacteria: DNA data storage in living cells

Chinese scientists in Tianjin University have stored and retrieved 445KB of digital data from living E. coli bacteria cells and say E.coli represents a stable DNA storage medium.

Up until now, digital data storage in DNA has used synthetic DNA stored in glass phials or similar containers. A Chinese team of research scientists has done it by inserting synthetic DNA into living cells and then retrieving the data after the cells have reproduced.

The researchers said artificially stored DNA, stored in vitro or glass, typically use short lengths of DNA with fragment lengths ranging from 100 to around 200 nucleotides. Their method, stored in vivo (living cells) uses much larger fragments, with up to 11,520 nucleotides.

Schematic for data storage using DNSA in bacterial cells.

As the cells reproduce they make new cells which carry the inserted digital data-carrying DNA fragments. The researchers said “the genomic maintenance mechanism of living cells ensures that the DNA molecules are replicated with high fidelity.” This means the fragments can be retrieved from a larger number of cells than the starting population, meaning “higher stability and longer storage periods could be expected.”

The details are in a Nature paper and the researchers looked at oligos, short for oligonucleotides, which are single short strands of DNA. They used more than 10,000 of them encompassing 2,304 kilo bit pairs (kbps). This is the unit of length measurement for  DNA equal to 1,000 base pairs. A base pair is two of the four DNA nucleobases combined, meaning adenine with thymine, and cytosine with guanine.

The source binary data was encoded into the 4-letter DNA alphabet formed from the four nucleobases; A, C, G and T. There was an encoding redundancy of 1.56 per cent at the software level to tolerate the physical loss of some oligos. They built oligo pools of up to 11,520 distinct oligos. These were assembled into plasmid vectors, twin-stranded circular DNA molecules which replicate independently within a cell. These were assembled in a redundant fashion and stored in a mixed E. Coli culture on solid plates or in liquid medium. There, the cells reproduced and formed colonies.

Use of plasmids in DNA-based data storage in cells.

This number of oligos remained stable in the mixed culture of E. coli cells even over multiple divide and split operations in which a culture in one medium was divided into two, left to grow again, and then divided once more. Up to five such splits, “passages” in the scientists’ terminology, were demonstrated.

The cell’s digital data contents were read via DNA sequencing and found to be correct. The reading process involved plasmids carrying the digital information being isolated from a large liquid culture. Then a large number of oligos was recovered following a digestion process. There was little contamination from the host cell’s own DNA and that was removed using bio-reagents.

In summary the researchers concluded: “DNA storage inside cells has distinct advantage in terms of stable DNA maintenance for long periods of time and very low cost of replication.”

They claimed their work “is the largest scale archival data storage in living cells reported so far, paving the way for biological data storage taking advantage of both in vitro synthesis capacity and the biological power of living cells in an economical and efficient way, which is crucial for developing practical cold data storage on a large scale.”

The paper is entitled “A mixed culture of bacterial cells enables an economic DNA storage on a large scale” published in Communications Biology, volume  3, Article number: 416 (2020). It is credited to Min Hao, Hongyan Qiao, Yanmin Gao, Zhaoguan Wang, Xin Qiao, Xin Chen & Hao Qi.

Zerto beefs up backup, DR and in-AWS-cloud protection

DR specialist Zerto is converging backup and disaster recovery (DR) and hopes to strengthen both sides of that equation with stronger backup and DR facilities, using an expanded continuous data protection engine, as well as previewing an in-AWS cloud backup service.

Zerto Enterprise Cloud Edition (ECE) is its core a DR product providing DR facilities for on-premises and cloud applications. The DR can be provided to the cloud, from the cloud and in-between clouds.  It features automation of both DR and backup functions and ECE includes the Zerto Data Protection (ZDP) continuous data protection technology.

ZDP

Gil Levonai.

Gil Levonai, CMO and SVP of product at Zerto offered a quote: “We … are now delivering a new offering that I personally believe will change the backup market—an industry that hasn’t evolved in more than 30 years. ZDP gives businesses a data protection strategy for all of their applications with significant TCO savings tailored to their unique needs.”

ZDP delivers local continuous backup for day-to-day backup restores. Its local journaling technology enables customers to recover without the data loss, downtime, or production impact that Zerto says are inherent to traditional backup systems ensuring business continuity and availability.  

In fact ZDP should, in Zerto’s view, displace traditional backup because it offers lower data loss rates and lower infrastructure costs in TCO terms, with an up to 50 per cent saving claimed. ZDP also provides long-term retention on-premises or in the public cloud; with both AWS and Azure as public cloud targets.

Positioning

Updated: Oct 16

Zerto has ZDP, ECE and Zerto 8.5. How do they fit together? Zerto told us: The core software, referred to as the Zerto platform has now moved from v8.0 to v8.5; “ZDP and ECE are the ways you consume/use the platform. They are the license types that you can have now with Zerto.”

Zerto told us “ZDP [which] is for backup and long term retention. … is a new offering, still based on Continuous Data Protection that is focused and priced for backup. The reason it’s priced for backup is because Disaster Recovery capabilities (failback, failover, DR testing, re-IP) are removed because customers don’t need those capabilities for “backup.”

In effect ZDP is ECE-lite. The company tells us: “This means that we will see most customers use ECE for the mission-critical applications and ZDP to back up the rest of their environment that only requires “backup’’ and not DR.”

Zerto 8.5

Zerto 8.5 is the latest version of Zerto’s core protection software, and is used in both the ECE and ZDP products. It follows on from the Zerto 7.0 and Zerto 8.0 versions. At the time of its release Zerto said v7 “converges backup and disaster recovery using hypervisor-based replication and journalling for short- and long-term retention.” And: “Its technology allows you to achieve RPOs of seconds using journal checkpoints up to 30 days ago, instead of a 24-hour recovery time frame.”

Zerto 8 advanced this, as it brought Continuous Data Protection (CDP) to VMware on Google Cloud.  The company said: “Zerto’s continuous data protection (CDP) offers the most effective protection for your business applications and data. Zerto automatically tracks and captures modifications, saving every version of your data locally or to a target repository, short and long term.

Zerto claimed its “CDP and innovative journaling technology removes the need for snapshots and thus eliminates production impact and lengthy backup windows. Its … recovery granularity reduces data loss to just seconds.”

The 8.5 version takes this a step further as it expands the CDP applicability beyond VMware on Google’s Cloud. It is also now suited for for lower tier application backup, and not only DR for upper tier applications. Zerto 8.5 includes;

  • Backup directly to Azure and AWS
  • Instant file and folder restore to production
  • VMware on public cloud disaster recovery and data protection for Microsoft Azure VMware Solution, Google Cloud VMware Engine, and the Oracle Cloud VMware Solution
  • CDP capabilities
  • Platform automation and lifecycle management features
  • Zerto PowerShell Cmdlets Module

The Cmdlets enable performing specific tasks by using a script and not from within the Zerto User Interface. This could be for retrieving information like the Virtual Protection Groups (VPGs) defined for a site, working with checkpoints and deleting a VPG.

Preview

Zerto also previewed an in-cloud data protection and DR product on AWS, which protects applications across regions with cloud-native resilience.

The company said it will extend its platform to offer more simplicity and orchestration across all use cases. This will cover businesses requiring a recent data restore due to user error or DR from an infrastructure outage, cloud-first businesses, or businesses just starting out in the public cloud. 

It wants to be the one-supplier-fits-all use cases across the data protection spectrum from SMB backup to large enterprise DR covering on-premises virtual servers, containerised servers and in-cloud applications in multiple public clouds.