November 23, 2020

The 37th IT Press Tour is around the corner

Covid impacts a lot of business and IT vendors have difficulties to maintain their visibility levels. Conferences and expos disappear and companies suffer from this.

The IT Press Tour, already recognized and picked by 240 companies for several years, benefits from the situation and accelerates.

The coming December edition will invite top US and European journalists to meet 10 hot technology companies:

  • Atempo, European leader in data protection,
  • Data Dynamics, reference in data management,
  • DDN Storage, largest private storage provider,
  • iXsystems, innovator in open source storage,
  • Kaseya, a comprehensive and ambitious IT software vendor,
  • Minio, fastest growing object storage company,
  • Robin.IO, emerging solid player in Kubernetes-based storage,
  • SoftIron, a Ceph-based disruptor,
  • StorCentric, a new kind of storage solutions player,
  • and StorPool, European software-defined storage actor.
I invite you to follow us on Twitter with #ITPT and @ITPressTour, my twitter handle @CDP_FST and journalists' respective handle. It will rock again.


October 30, 2020

StrongBox signs a huge deployment

StrongBox Data Solutions, a leader in Data Management, just announced a mega deployment at DKRZ. The deal is huge for the company both in tern of data capacity managed but also in terms of revenue and will last 5 years.

The project is two fold: management 120PB per year of HPC workflow and 150PB of migration form the previous system based on HPSS. StrongLink will replace proprietary HPSS services with LTFS and S3. The goal is to manage all data horizontally across many storage units with network file systems, LTFS and S3 with intelligent transparent tiering. Initially 150PB has to be migrated from previous HPSS configuration and manage 120PB per year of new data. It means that this organization generates on average more than 320TB every day.

In details, StrongLink will create a virtual namespace over tier 0, 1 and 2 environment from HPC storage, 1PB of tier 1 cache and 5 tape libraries equipped with 90 tape drive sand 75,000 cartridge slots. This project is implemented and supported by Cristie Data.

This deal confirms the scalable approach StrongBox Data can offer and will probably serve as an active reference inviting similar project to consider StrongLink.


September 9, 2020

New era for Ugloo

Ugloo, one of the P2P software storage innovators, just announced the recruitment of Gonzague Dupont as its COO. Gonzague came from Zerto, Commvault, Splunk, Veritas/Symantec and has a strong sales experience with small and established players.

Ugloo develops a dispersed storage infrastructure based on commodity servers. These servers could be already deployed in the environment or new and dedicated to the storage service. It offers a high durability level around 11 9s with an advanced erasure coding implementation installed on one site or across several sites. The storage is exposed as a S3 interface and thus could be integrated into many third party software. Among them, this storage service is perfect for secondary storage and use cases with backup/restore or archiving. The product is already supported by Rubrik and Atempo among others.

Ugloo recently published a white paper you can download here.

July 24, 2020

Dead end, OVHcloud swallows OpenIO

On July 23, OVHcloud surprised the Lille, France microcosm with a storage acquisition.

The original press release is available here.

This move gives a strange feeling and raises one big question essentially on the OVHcloud side. Why did OVHcloud, a cloud service provider, leader in Europe, buy a confidential open source storage software company? What is the logic behind this acquisition?

A few immediate answers come to mind:
  • OVHcloud can use the product without buying the company but it become difficult to continue to use a product without a company or a team behind it,
  • Following its venture round and even later, OpenIO didn’t really get more visibility, got small adoption limiting its development capabilities and perspectives,
  • The French firm had difficulty to raise a new round as its business is just weak, so thinking about this exit in the Lille ecosystem was just a way to save the technology and keep things in France.
One of the reason of the limited success comes from the absence of a real CEO – Laurent Denel being in fact a CTO – a good one well respected, and even if it’s a good model to start, it has to evolve with the company goal and ambition. Pressure from board’s members, even friends and other external people didn’t persuade Denel to move to this position to help the business grows. CEO position requires a different DNA.

Investors finally admit now what we said 3 years ago about the miracle of the $5 million raising campaign around the end of 2017 confirming that the acquisition multiple can’t be high and could be interpreted as a mayday. On the other hand, OpenIO technology stays in France which is a key strategic consideration for OVHcloud and France globally. There were some recent talks about the incredible choice made by French authorities to consider and sign with Microsoft Azure for the French Health Data Hub. In that case, OVHcloud should have been the preferred choice. We also see the GAIA-X project to build an European cloud initiated by Germany and France and OVHcloud with other cloud service providers like Orange, Scaleway or Outscale are also present. This is a clever idea as we still don’t understand why France has started several years ago Cloudwatt and Numergy, 2 sovereign clouds, that finally went nowhere.

Also it’s worth mentioning that Covid-19 has an impact of the business climate. Some normal companies became fragile and some fragile even disappeared.

All this invites us to ask a second question. Does OVHcloud really need to own its object storage solution? And why? Like many users, when your provider becomes fragile, the question to keep, own or acquire the solution is crucial to continue to offer the service powered by this technology. And if OVHcloud wishes to continue to use the French open source product, there is no alternative, they have to buy OpenIO. With French cloud project and sovereignty around it, it would have been strange to swap OpenIO by other American open source peers. Just impossible and out of reality. For OVHcloud, this is the first acquisition (real first one was vCloud Air in May 2017 acquired from VMware) that help them achieve different goals at the same time and it is a bargain.

OVHcloud comment is not very positive, claiming the desire to build the best object storage product on the planet. So it immediately means the current product is far to be that and even if the ambition is respectable, it reveals the total ignorance of that storage segment.

We understand that the OpenIO brand will disappear and we doubt that the solution will be promoted outside. How an OVHcloud product could be installed elsewhere? In other words, OpenIO becomes a proprietary solution available as open source but no more acting as an ISV. Of course, OVHcloud is not an ISV. But what will OpenIO users will become? Will OVHcloud really support them and especially as some of them are competitors like Scaleway? What will be the choice for them? Swap to other open source solution such Ceph or MinIO? And finally adopt a non French product…

Like many second tier cloud providers such OVHcloud, Outscale or Scaleway, their storage offering is interesting representing 2/3, being optimistic, of what the big players – AWS, Azure and GCP or even Alibaba Cloud – offer. Their storage service page claims to offer a comprehensive storage services but it’s essentially block and objet storage and they completely omit file storage. In other words, you don’t find file storage services like AWS EFS, Azure File Storage or GCP Filestore or just a simple NAS service to name a few.

Being and acting as an ISV is real mission and the OpenIO team comes from a IT services company with different expertise and human behavior. For a commercial ISV, the strategy is pretty obvious but not easy, plenty of hurdles, you raise money, you iterate the product development for use cases you identified, do marketing and build a ecosystem and if that works you can expect to take off. For an open source ISV, the story is different, you potentially raise a small amount of money as the return is lower and you try to build a momentum in the open source community, battling for credibility, adoption and visibility. But here the game was over with 2 other famous products: Ceph and MinIO.

OVHcloud and OpenIO mentioned that they contributed to OpenStack Swift but this service was a pure failure even OpenStack globally. The Swift journey is perfectly illustrated by SwiftStack who hit a wall a few months ago trying to survive for several quarters. The company, promoter of Swift, was swallowed by Nvidia early 2020 and Nvidia never mentioned the acquisition, no press release was issued neither a blog by the GPU giant. The product is not mentioned at all on their website, only SwiftStack mentioned it. It means it was a perfect bargain for a negligible piece of software acquired for a very small amount of dollars.

Coldago Research has published at the end of last year its annual Map that positions 22 object storage vendors. This map confirms the difficulty that OpenIO had to transform good ideas into real product traction then into revenue. For both 2 dimensions Execution and Capabilities and Vision and Strategy, the software vendor appears to be very weak.

Past software-defined storage acquisitions, and especially a subset of it, illustrate the difficulty to survive in a crowded segment with lack of differentiators. It was not the case 5 years ago when not all companies were supporting S3 (imagine that some company thought about an alternative interface, brilliant prediction…) or erasure coding or file interface. Today you can try to list vendors and their offering and even try to establish features differentiators, just that, good luck. You can even describe a product to users by its features and they will miss to recognize the product as they’re so close. Differences come from different areas and one of them is the ecosystem, ecosystem of partners with integration. This acquisition is bad news for the segment confirming the difficulty of small players in a crowded segment. Therefore it reinforces leaders and innovators like IBM, NetApp, Hitachi Vantara, Cloudian, Pure Storage, Vast Data and MinIO.

OpenIO arrived very late, Veritas did same mistake as it entered into the segment too late to finally gave up a few months later. Veritas had time and money to insist but they did not. It proves that Symantec froze Veritas development in storage and data management. Some companies disappeared like Formation Data Systems, Coho Data, Skylable, some made a pivot like Igneous and reminds the OpenIO initiative in the similar domain. Quietly, OpenIO killed its Ethernet drives initiative finally.

MinIO is clearly the leader in open source object storage with tons of integration and adoption. The company even discovers some integrations made by users and partners with well documented solutions. Therefore the market traction is high, proven and recognized. Among this various integrations, there are Cisco, Datera, HPE MapR, Humio, iXsystems, McKesson, Nutanix, Pavilion Data Systems, Pivotal, Portworx, Qumulo, Robin.IO, Splunk, Ugloo and VMware.

None of the object storage vendors ever made an IPO and we don’t see and expect anyone doing so even if some had dreamed about it reaching some level of utopia when you compare reality, facts and projections. And such behaviors didn’t help their leaders, reputation and the segment as a whole. The more realistic goal is to find an exit but it became more and more difficult as the number of potential acquirers shrinks by itself, many of these players having a solution developed internally or by external operation. So the alternative is to become a zombie i.e a company that never got public, never got a stellar trajectory, still not profitable and continue to raise money, dilute capital to extend artificially their life… What a plan, we see companies on the market that were founded 10 or even 15 years ago. What’s the goal here, to get at the end a few $millions in their pocket if they ever find an acquirer? Avoid personal reputation disaster (it’s already the case)…? This survival mode is pathetic.

The object storage segment has seen different trends with S3 everywhere, on top of NAS, as gateway to tape, in HCI and even in unified storage confirming that object storage is just a S3 interface today. Read the article titled S3 eats the world published in June 2019.

As an unstructured data storage solution, it’s the perfect companion of file storage solution and we see more and more file storage product adding a S3 interface. So is there a place for pure object storage? This is a fundamental question as only a few of the market solutions got adoption, users prefer file + object, multiplying use cases on the same data platform.

At the end of the day, the OpenIO product and the team come from an IT services company and return to a new kind of player, a cloud service provider.

We suspect who will be the next one…

July 20, 2020

StorCentric unveils 3rd generation of Unity product line

StorCentric, holding structure controlling Nexsan, Drobo, Vexata and Retrospect Software, recently announced a new generation of Nexsan Unity product with model 3300 and 7900.

These products represents the 3rd iteration and support hybrid and all flash media and expose block and file protocols.

Two new key features appear, the first Data Migration helps adoption and installation of the array with the capability to move data to the new storage device smoothly. The second is the Cloud Connector supporting 18 public clouds from top well known to regional ones for tiered storage.

Unity is also able to connect to Assureon, the immutable storage solution coming for Nexsan and before Evertrust.

Unity is a very rich unified product providing high availability, snapshots, ESXi integration, inline compression, FASTier™ caching, asynchronous and synchronous replication, data at rest encryption, and advanced checksums and background scrubbing to avoid silent data corruption.

Both systems embed 2 controllers,
  • Unity 3300 is a 3U chassis with 120 HDDs or 24 SAS SSDs maximum with the expansion representing respectively 1.68PB and 983TB, 256GB of DRAM, 1.6TB of SSD for FASTier write cache and same capacity for read cache.
  • Unity 7900 is a 2U chassis with 480 HDDs and 216 SAS SSDs again with expansions chassis for 6.72PB and 3.31PB respectively, 448GB DRAM, 16BG of NVDIMMs for write cache and 1.6TB of SSD for read cache.

The Unity family represents an interesting alternative to classic unified storage from Pure Storage, Infinidat, NetApp, Dell EMC or recent Pavilion Data Systems even of some of them like Pavilion offers a third interface like S3.

July 13, 2020

Recap of the 35th IT Press Tour

This report was published for the 1st time on StorageNewsletter July 6th.

The recent 35th edition of The IT Press Tour, this time operated online, has been the opportunity to virtually meet 13 companies and their executives around IT infrastructure and data management. Several of them launched products and did announcements.

Bamboo Systems
The firm, emerging ARM server vendor, introduced the B1000N Series with 4 independent server in 1U with 1 or 2 blades. A full 2 blades server embeds 8 nodes with 128 cores, 512GB RAM and 64TB NVMe SSD at 24GB/s read with 4 x 40Gb QSFP. It promotes a new metric for server with the energy consumption so kilowatt hours instead of processor frequency or CPU cores. The team promotes a new design named PANDA for Parallel Arm Node Designed Architecture. Future direction indicates a more dense serve going to 12 blades for 48 nodes.

Fujifilm Recording Media
It reintroduced its Object Archive, known under this name in North America and Japan, and as Software-Defined Tape elsewhere. The idea is pretty similar to the VTL for tape libraries a few years ago but now the solution is more universal with a S3 interface. With the product, it introduces a new open source tape format named OTFormat (Open Tape Format) unifying metadata and data and supported from LTO7. The pricing model is based on capacity as a subscription grouping the software and LTO tapes for multi petabytes.

it launched a Cloud Service Provider program fueled by Protégé. CSPs can promote the service as a co-branded solution for data protection, data migration and disaster recovery. Users can leverage the service to adopt a multi-cloud approach without any risk. Thanks to an open API, Protégé offers the capability to be integrated into a more global cloud based solution for managing data and infrastructure and finally deliver a consistent data management across different IT environments.

It continues to expand its product portfolio by acquisitions fueled by a rapid market penetration. IT Complete is very frequently enhanced and improved by adding new services in endpoint management, security, backup and DR, cloud management, audit and compliance plus a service desk and an automated IT documentation. For backup, IT Complete integrates Unified backup, Cloud Backup and 0365 Backup leveraging Unitrends and Spanning data protection technologies. Fred Voccola, CEO of Kaseya, said that there is a pending acquisition in the security space and the next big event for the company should be an IPO in 2021.

The company re-insisted on its product strategy and especially the enterprises aspect of it. Exposed as a NAS in central sites or remotes offices, deployed on top on object storage, it develops an internal distributed file system named UniFS. It is an internal file system as it is not visible by users but really is the cornerstone of the product. The file solution spans primary and secondary storage thus avoiding silos and complexity. The company is now an unicorn.

It launches its product, the Services Processing Unit (SPU), as a core of its Cloud-Defined Storage product (CDS) strategy. The SPU is a PCIe board designed to offload data services from central CPUs and can build clusters of 32 systems with this horizontal data centric model. The infrastructure is managed by a cloud-based application named Nebulon ON.

Remember that Salesforce will stop its data recovery service July 31s, 2020. As the #1 backup and recovery solution in the AppExchange with 2,000 customers, OwnBackup introduced its Enhanced Sandbox Seeding offering to accelerate development, testing and training. The team also develops a Governance Plus package that leverages Etherum blockchain to prove backup integrity. Beyond backup, OwnBackup Archiver offers the possibility to restore archives directly from Salesforce.

Pavilion Data Systems
It took time to pitch its positioning as the 3rd wave of storage fueled by a fundamental element i.e end-to-end NVMe. The idea of Pavilion Hyperparallel Flash Array (HFA) is to provide fast, dense – in 4U – storage services for demanding environments thanks to a very scalable architecture with 20 controllers for 72 SSDs and 40x100GbE ports.

The firm launched its Shift product and unveils a strategy with the facility to move file data stored on Qumulo clusters, on-prem or cloud-based, to AWS S3. The idea is to leverage the ton of AWS applications which consume data through a native S3 connection. It offers to its users a way to subscribe to AWS, compute data with new services, while continue to operate Qumulo “classic” cluster in different places.

It develops a cloud native platform to provide end-to-end automation for Kubernetes-based applications. From a storage point of view, Robin delivers snapshot and clones, QoA, replication, backup, data re-balancing, tiering, thin provisioning, encryption and compression. The company has found a specific niche, 5G networks, to deploy its technology with radio, edge and core sites.

UK-based company leads the Ceph-based storage infrastructure wave with the design of its own appliance optimized for energy consumption and performance. Leveraging open source with a clear direction towards consistent performance at scale while energy is minimized with dense systems, HyperDrive targets demanding storage environments such research centers like the recent Sandia National Labs reference. Beyond storage, SoftIron has designed, developed and built 2 others products: HyperSwith based on SONiC and HyperCast for transcoding device based on FFmpeg.

StrongBox Data Solutions
It continues to extend its data management solution with a second generation product, probably one of the most comprehensive in a very competitive market landscape. Dedicated to unstructured data management, it adds the support of LTFS and provide data movement/placement/copy/tier from any source to any targets as soon as the interface is a file, object, cloud or LTFS-based protocol. This product iteration represents a new dimension in the scalability aligned with high demanding environments needs.

It reaffirmed its disaggregated shared everything architecture leveraging 3D XPoint with Intel Optane, QLC flash and of course NVMe-oF. The goal is to "kill" HDD with a full flash file storage solution for all tier and any workload. To achieve this, the team has developed a wide erasure coding and data reduction techniques to deliver an compelling $/TB.


July 9, 2020

OwnBackup raised $50M

OwnBackup, the reference in SaaS applications backup, just announced a new round of $50M to accelerate its growth.

This round was led by Insight Partners and Salesforce Ventures is again in the loop. It means that the company reaches $100M and clearly presents itself as the #1 SaaS application backup vendor even if the solution is today limited to Salesforce but it's probably the best one.

Insight Partners owns capital in several data protection companies and we imagine the desire to build a software giant in the domain especially since they acquired Veeam. It could be even bigger...

Other serious players in Salesforce data protection are Kaseya Spanning, Datto Backupify, Asigra or Druva. This backup segment starts to be very hot with approximately today more than 20 players.

I asked the team several times, and especially during the recent the IT Press Tour, to expand to other SaaS applications like Workday, NetSuite... but also they need to address G Suite, Office 365... and it seems that the direction is now taken.

July 3, 2020

Fivetrran joins the unicorn club

Fivetran, a fast growing analytics data connection firm, just announced its series C of $100M at $1.2B valuation. The total raised so far is $163M with Andreessen Horowitz, General Catalyst, CEAS Investments and Matrix Partners.

The company continues to accelerate fast even during hard Covid period and counts so far more than 1,100 customers with 75% growth on this side. For readers who don't know yet Fivetran, the provides pipes to connect different data sources with data warehouse such Snowflake, BigQuery or Redshift..

July 2, 2020

Kaminario lost its CSO

Kaminario seems to be under some changes wave and one of them is the move of Eyal David, former CTO and more recently chief strategy officer, to Model9 as its new chief strategy officer. Model9 is an ISV dedicated to Cloud Data Management for Mainframe. The company has received a series A funding of $9M.

June 29, 2020

New ARM-based server for Bamboo

Bamboo Systems, formerly known as Kaleao, announced recently during the IT Press Tour its new ARM-based server, the model B1000N. Founded in 2015, the company has raised $4.5M in a Seed round.

The new system introduces a new architecture named PANDA for Parallel Arm Node Designed Architecture. PANDA combines multiple benefits for modern software needs and low energy consumption.

Microservices applications have initiated this new design associated with the trend for high density of components.

Bamboo targets specific demanding use cases around AI/ML, analytics, edge computing and Kubernetes environments.

B1000N is a 1U server with 1 or 2 blades and each blade embed 4 compute nodes plus a non-blocking embedded L3 switch exposing dual 40Gb QSFP ports. Usually a node operates Linux meaning 1U you can have 8 Linux servers, perfect density and computing power for lots of applications. But the strategy of Bamboo is multi-servers so a multiple of 8 servers finally.

In 4U, you obtain the power of 48 servers reducing drastically energy and footprint therefore TCO. A node uses NXP 2160A A72 Coretex 16 core processors so per blade it gives 64 cores and on 1U 2 blades server it means 128 cores. RAM is 64GB/node and 8TB PCIe/SSD/node so 64TB per 1U server. It's small for storage but perfect for edge computing.

ARM processors arrive at the right moment as a few key metrics such energy and parallel tasks to tackles big problems. The momentum has already started and shows 38.2% of increase for none x86 while x86 decreased 9.1% during the same Q1/2020 period.

Recently Apple also swapped Intel x86 to ARM CPU illustrating a real trend. Bamboo insists on the new metric introduced as the kilowatt-hours instead of frequency or number of cores for eon for instance.