Home Blog Page 435

SSD storage rulers of the world, unite!

The EDSFF organisation has standardised short and long and 3-inch ruler form factors for SSDs, with specs available from the Storage Networking Industry Alliance. 

EDSFF stands for Enterprise and Data centre SSD Form Factor and there is an EDSFF working group. The idea is to move away from 2.5-inch and 3.5-inch disk-bay sizes for SSDs, which now come in these sizes as well as M.2 gumstick cards and PCIe add-in card (AIC) formats.  There will be instead longer and thinner devices, the so-called rulers, which can provide 1PB of capacity in a 1U server, using 64-layer, TLC 3D NAND.

This mean more capacity in less space, leading to denser storage and server systems and less wasted space.

Suppliers involved in the EDSFF initiative include SSD makers Intel, Micron, Samsung, Toshiba and Western Digital, and server suppliers Dell EMC, HPE and Lenovo, which augurs well for wide industry adoption and the unification of the differing Intel and Samsung ruler formats.

There are short and long ruler standards with 9.5mm and 18mm widths. The short ruler is 5.9mm thick, 111.49mm long and 31.5mm wide. The long ruler can be 9.5mm or 18mm thick, 318.75mm long and 38.4mm wide.

An Intel slide shows its EDSFF plans:

The EDSFF is also working on a 3-inch form factor that is 76mm (3 inches) wide. This format has two thicknesses and two lengths: 7.5mm or 16.8mm thick and either 104.9mm or 142.2mm long. 

Download the format specs here.

Intel changes data centre SSD naming scheme

Intel is tidying up its naming scheme for data centre solid state drives, ranging from slow, entry-level drives to high-speed Optane near-memory speed products.

Data Centre drives from Intel are currently branded as SSD DC P4000 and SSD DC P3000 products, with Optane versions called Optane SSD  DC P4000. There are also SSD DC S4000 and SSD DC S3000 products. It’s a bit of a mess.

Anandtech  has published a slide showing the new five-level scheme:

The DC P4000, P3000, S4000 and S3000 scheme is transitioning to a D1, D3, D5, D7 Series and a separate Optane DC SS Series.

The entry level products will use SATA and PCIe interfaces and are renamed SSD D1 Series, or D1 Series for short. 

The D3 Series will be SATA interface products for mixed workloads, and two exist already – the D3-S4610 (240GB to 3.84GB) and D3-S4510 (240GB to 3.84GB.)

The D5 Series products will be capacity-optimised, using QLC (4bits/cell) flash and NVMe interfaces, and are intended for read-intensive workloads. Intel has recently announced the D5-P4320, with 7.68TB of QLC flash capacity, using 64-layer 3D NAND. As 96-layer NAND is in development we can expect a 50 per cent capacity uplift to around 12TB (11.36TB ) in future.

D7 Series products will are designed for mixed workloads and use TLC (3bits/cell) flash with an NVMe interface.

The 3D XPoint-based Optane products are henceforth named the Optane DC SSD Series. This is bound to become the Optane series of products, until some other solid state technology base is used alongside Optane 3D XPoint.

SoftRoCEr Attala provides NVMe over standard Ethernet

NVMe over Fabrics (NVMe-oF) provides direct-attach SSD access latencies for external, shared flash storage, side stepping Fibre Channel and iSCSI network delays. RoCE provides remote memory access-like speeds across data centre-class Ethernet for NVMe-oF.

But it has a need for the underlying networks to be lossless. Composable storage infrastructure supplier Attala Systems says this is relatively easy for rack-scale deployments, but is difficult to achieve on conventional layer three (L3) leaf-spine networks. By hardening the open-source SoftRoCE initiator, Attala’s own devices and providing SSDP-based discovery, Attala Systems has enabled its RoCE-based networking capability to operate on standard leaf-spine networks and with ordinary NICs. 

This allows cross-rack and ubiquitous deployment of NVMe-oRoCE’s low latency and high performance on Ethernet networks at speeds from 10GbitE to 100GbitE and beyond.

Taufik Ma, Attala Systems CEO, said: “NVMe over RoCE has always been seen as the lowest latency, highest performance flavor of NVMe protocols. However, NVMe over RoCE’s need for lossless networks and specialized initiators has been somewhat of a limitation to its deployment. Attala Systems’ approach allows their solution to be deployed on standard L3 leaf-spine networks without any special considerations.”

Attala hardens the  open-source SoftRoCE initiator driver and its own devices to gracefully handle packet-drops in a L3 leaf-spine networks – even those which do not support priority flow control (PFC) for achieve losslessness. 

It says this unshackles NVMe-over-RoCE solutions from the confines of a single rack and specific servers. Attala’s system also includes an autonomous Simple Service Discovery Protocol (SSDP) based discovery mechanism for newly added end-points, solving a common issue experienced by IP networks being utilized for scale-out infrastructure deployments. 

The community-hardened open-source SoftRoCE initiator driver runs on any ordinary NIC, enabling 100 per cent compatibility with any Linux-based host server across the data centre.

This looks good. We look forward to hearing what customers say.

“Attala” by the way is the Sanskrit word for “watchtower” and you can check out Attala’s offerings here.

DDN sets up Server Virtualization division

Data Direct Networks (DDN) has set up  a Server Virtualization, Analytics, VDI, Container and DevOps Division (SVAVCDOD?) and plans to staff it with 100 new hires by the end of September. It says dozens of enterprise storage experts from companies such as Tintri, Pure Storage, IBM and Intel have joined already.

The services arm will advise clients on how to get the most value out of  flash, virtualized and containerized hybrid cloud installations, and  it represents a breakout beyond  DDN’s traditional HPC and big data analytics base.

DDN says its recent acquisition of Intel’s Lustre file system business added the world’s most advanced file system technology to DDN’s already substantial flash, analytics and cloud product portfolio. DDN also entered into an Asset Purchase Agreement with Tintri on July 18, 2018, whereby DDN has offered to acquire substantially all the assets of  the stricken company offering enterprise cloud infrastructure built on a public-cloud like web services architecture and RESTful APIs.

+Comment

This is a multi-million dollar initiative by DDN, which will bring it into competition with general enterprise server system vendors such as Dell, HPE, IBM, NetApp and Nutanix. If the Tintri acquisition goes through it has a shared external array product supply and customer base to build on. What it will do with that product technology, in this developing era of QLC flash, NVMe-oF, containerisation, storage-class memory, AI and data management is an intriguing question.

A bigger splash: Attala dives into hot data lake

Attala has announced Attala Data Lake (ADL) multi-tenant software, designed to pour a massive pool of data in a tailored shared hot shared cache box.

Attala produces FGPA-driven SSDs for direct NVMe-oF access to targeted drives and a composable NVMe storage resource. The concept for its new product is store a data lake on NVMe flash for faster-than-disk and SAS/SATA SSD access by multiple real-time applications. A shared flash cache as it were.

Deep Storage founder and chief scientist Howard Marks has provided a prepared quote for Attala: “Users can, and have, installed NVMe SSDs directly into their compute nodes, but that increases costs and complexity. Attala Systems’ new multi-tenant hot data lake capability provides shared high-performance NVMe storage to real-time analytics applications, simplifying deployment and reducing costs.”

The ADL software module provides centralised management and provisioning of the hot data lake. Attala says the result is ” low-cost, high-performance solution for real-time analytic applications.” 

Yes, but not exactly low-cost; SSDs in sufficient capacity have to be bought and that won’t be cheap, unless we are talking about data puddles and not data lakes.

The ADL module is available for evaluation by select customers and partners, and will be generally available in the fourth 2018 quarter.

You can use cheaper ethernet components for NVMe-oF – and here’s how, says Toshiba

There’s no need to use expensive RoCE-class Ethernet for NVMe-over Fabrics, according to Toshiba.

Toshiba’s Kumoscale software delivers an aggregated multi-SSD block storage pool across an NVMe over Fabrics (NVMe-oF) connection and it works with the NVMe over TCP protocol. Toshiba has tested this using Marvell FastLinQ 100GbitE network interface cards (NICs), which employ a TCPOffload engine.

The company says NVMe-oF on TCP enables a good price/performance balance because standard Ethernet components can be used instead of  data centre class versions required by, for example, RoCE-based NVMe-oF, as well as existing Ethernets.

Tosh has roped in IDC’s research VP Eric Burgener to provide a supporting quote: “Solutions like Toshiba’s KumoScale which use TCP as a transport give customers an NVMe over Fabric option that requires no custom hardware or software on the server side, making this higher performance host connection available to any and all servers without additional cost.”

So there is, Tosh implies, no need to go down the RoCE road. It says the NVMe organisation is actively working on a binding specification for NVMe-oF over TCP, and it is ready for the upcoming certification process.

Tosh and Marvell will jointly demonstrate KumoScale with TCP acceleration at Flash Memory Summit booth #307 (Hall A) in the Santa Clara Convention Centre from August 7-9.

Actifio retains unicorn status in latest funding round

Four years after picking up a $100m E-round, Actifio has picked up another $100m, valuing the company at more than $1.3bn. Total investment stands at $311.5m, according to our records.

The  F-round was led by Crestline Investors and joined by North Bridge Venture Partners, 83North, Advanced Technology Ventures, Heritage Group, Andreessen Horowitz, and other existing investors.

In the 2014  funding Actifo was valued at $1.1bn – gaining it unicorn status as a startup valued at more than $1bn.  At the time  of the funding Techcrunch reported the copy data manager and converger vendor was preparing for an IPO and expected $100m annual revenue run rate for that year.

The IPO never got off the ground and in four years  the valuation has increased by just 18.2 per cent – admittedly that works out at an extra $200m.  But the company has an updated story to tell. It claims a total addressable market (TAM)  of $50bn – 13.6 per cent larger than the $44bn TAM calculated in 2014 – and more than 3,000 enterprise customers, compared with 300 in 2014. They include five of  the top 20 global financials, four of the top 10 energy companies, three of the top 10 healthcare providers, six of the top 10 service providers and four of the top 20 global retail organisations.

Actifio now refers to itself as a multi-cloud, data-as-a-service company. Competitors include Cohesity, Delphix and others in the data management space.  Let’s see how it uses this $100m cash injection,  which is needed to develop and grow its customer base, accelerate away from  competitors, and get revenues and revenue growth rate high enough to take another shot at an IPO. 

Nakivo extends backup product support

NAKIVO Backup & Replication v7.5 supports vSphere 6.7 and adds integration with storage vendors such as Dell EMC Data Domain, adding Boost support, and Netgear ReadyNAS.

It has bandwidth throttling to enable speed limits to be set for data protection processes to avoid network overload during business hours or chosen periods of time.

The software adds functionality for VM data recovery across platforms. Users can export VM data from VMware and Hyper-V backups into VMDK, VHD, or VHDX format. The exported files can then be used for recovery across different hypervisors, cross-platform VM data migration, long-term archival, or sandbox testing.

Nakivo says its software can be installed directly on Netgear’s ReadyNAS devices, creating a VM backup appliance. It is then a 5-in-1 offering featuring backup hardware, software, storage, built-in global data deduplication, and backup-to-cloud functionality.

A fully-functional free trial of NAKIVO Backup & Replication v7.5 can be downloaded at  www.nakivo.com.

Nakivo is a US business founded in 2012 which claims 20 consecutive quarters of double-digit growth and  more than 10,000 deployments worldwide. Customers include Honda, Coca Cola, China Airlines, Microsemi, and others.

NGD adds Newport computational storage drives

NGD

NGD has announced a Newport platform for computational storage, built on the Catalina-2 product base.

The 16 flash channel ASIC-based product runs 64-bit Linux in a variety of form factors including U.2, M.2 and EDSFF (ruler) with NVMe SSDs and an NVMe 1.3 PCIe Gen 3.0 x4 link to hosts. 

The software supports container virtualization and all Catalina-2 applications are supported.

NGD claims Newport allows computational storage to “cross the chasm” and enter mainstream market use cases. 

NGD CEO Nader Salessi claims: “The Newport platform further enables near-data processing for real-time analytics on large-scale data sets with improved power and density, both in watts per terabyte and terabytes per cubic inch.”

We have no information about capacity levels, pricing, on-board CPU power and customers.

The Newport platform and computational storage products will be showcased at the Flash Memory Summit, Aug 7-9, at the Santa Clara Convention Center. ®

NetApp ONTAP AI vs Pure AIRI

We have charted the numbers in the NetApp ONTAP AI story to show how NetApp’s A800-based AI box using Nvidia’s DGX-1 GPUs compared to its A700 version and the Pure AIRI system:

As you can see the A800 AI-in-a-box out-performed and out-scaled both the NetApp A700 system and Pure’s AIRI offering.

Tesla hires Seagate CFO

The surprise resignation of  Seagate CFO Dave Morton was due to a move to Tesla,  the electric car company has revealed.

Seagate has been Morton’s career home for 23 years. He will become the chief accounting officer at Tesla, reporting to CFO Deepak Ahuja.

It’s an interesting move, and one thing is certain: Tesla will be a more exciting employer than Seagate.

Tesla’s previous CAO, Eric Branderiz, left Tesla on 7 March. SVP Engineering Doug Field resigned from Tesla this month and Musk announced a 9 per cent Tesla employee layoff in June.

Tesla’s SEC filing said Morton, age 46, will have an annual base salary of $350,000 and will receive a $10m new hire equity grant (stock options effectively), which will vest over four years.

No doubt the company car scheme at Tesla  is also good.

Morton will be responsible for Tesla’s corporate financial reporting and global accounting, payroll, tax, and trade activities.

He resigned from Seagate on 25 July and is officially leaving on 3 August, seven working days later.