Jul 13, 2016

E8 Storage to lead the new NVMf wave

E8 Storage (www.e8storage.com), future leader of the new generation of storage arrays based on NVMe over fabrics, aka NVMf, recently participated to the IT Press Tour end of June and Zivan Ori, CEO and co-founder, based in Israel has made the trip to meet the press team. For references, I already wrote 2 times about E8 Storage and I invite to read May and early July posts.
First let me refresh minds with company info. E8 has less than 2 years of existence, funded in November 2014, having raised so far 2 rounds for a total of $18M from Accel - where we did the meeting in Palo Alto - , Vertex and Magma. The company splits responsibilities between Tel-Aviv for R&D and Santa Clara for Product, Marketing and Management. Zivan Ori came from IBM XIV and previously from Stratoscale with a strong background in software engineering and security. Alex Friedman, CTO and co-founder, came from XIV as well and Valentin Ossman, VP Product Management, came from PMC-Sierra, who co-present the session with Zivan.
To illustrate the shift in the demand for a new storage array approach, Zivan uses the famous RPC - Reliability, Performance and Cost - model also known as the 3P - Protection, Performance and Price. The RPC model was used to position local SSD, AFA, NVMf/JBOD and rack scale HW with also market needs and we realize that all these approaches has limited answers and don't solve all problems.
E8 insists on 2 dimensions - IOPS and TCO - of the solution and the slide #13 illustrates the result against AFA such XtremIO or Pure Storage and PCIe SSD. In 2 words: 10x performance and half the TCO. Wow super promising. Even the product was not available at the time of the presentation, it will be launched in less than one month at Flash Memory Summit in Santa Clara, specifications and characteristics were of course shared with us, here is an overview with this slide.

In addition, NVMe provides several advantages:
  • Low latency,
  • High Bandwidth and high IOPS,
  • Reduced CPU loads,
  • Device consolidation with high components density
We understand that E8 will shake the industry established position very soon now and the battle is on among the few new players such Apeiron, Mangstor, DSSD, Excelero and even AFA leader like Kaminario. Interesting again that this new storage development comes again from new and small dedicated actors not big ones that are just followers, again...

Jul 12, 2016

New release for StorPool

StorPool (www.storpool.com), a block SDS gem I discovered a few years ago and who has participated to the IT Press Tour in June 2014, just announced its last product release. Release after release, StorPool makes progress optimizing CPU consumption, increasing IOPS here by 30% and now able to support capacity beyond 1PB. The key thing about such product is how a user can integrate and use it. StorPool understands for a long time that the ecosystem is one of the key of the adoption and with that release, CloudStack is added alongside OpenStack, OnApp, LVM (hopefully) and LXC and Docker. Also among various new features, the SDS engine is now capable to manage 20,000 volumes and snapshots per cluster, a single volume can reach 30TB and larger ones can be created with LVM and performance got boosted. I just need to comment that there is no mention of throughput (GB/s) and no release number is associated with this new flavor of the product.

Jul 11, 2016

Versity develops a new Scale-Out File System

Versity Software (www.versity.com), innovator in data archiving software, announced during the last IT Press Tour a new project and initiative with the design of a Scale-Out File System. With a strong DNA in file system with 2 successful products - QFS and SAM-FS during the LSC time - Harriet Coverston, CTO and co-founder of Versity, has designed a new solution to address new storage challenges. The 2 challenges met are related to the scaling of Posix file systems and the scaling of archiving applications and the solution imagined by Harriet seems to be a new open source file system. The solution developed for 8 months will be beta in 4 months and GA before the end of 2016 with a new name and real launch. A good idea for the December IT Press Tour for sure.

Jul 1, 2016

E8 Storage promotes NVMe over fabrics

E8 Storage (www.e8storage.com), innovator in low latency storage, has driven and delivered an interesting SNIA tutorial "High Availability for Centralized NVMe" during the last DSI conference mid June. Zivan Ori, CEO and co-founder of E8 Storage, did a tough exercice to cover a topic and not to do any promotion for its own product that play in this segment, aligned with SNIA rules to present a neutral tutorial. The idea of this talk was to introduce the need for HA when you consider a centralized NVMe storage array, like in the past SAN storage array, but here with the constraint of the latency, that changes the solution.

Zivan Ori, CEO and co-founder, E8 Storage
NVMe drives, introduced in 2014, has very limited deployment models usually embedded in servers or laptops in a DAS classic way. In that case, NVMe drives are super fast with very low latency. Even with some problems and challenges like redundancy and resiliency to name one key point, NVMe is still very interesting especially when you consider sharing drives between servers. But now, when you do that, the remote latency diverges from local one I mean when the drive was locally attached. Second point is about pooling NMVe drives in a shared array with the bandwidth/throughput explosion challenge. The obvious idea would be to connect a NVMe array via a SAN but immediately we found 2 problems with the I/O stack and the storage network. For these reasons the industry has worked, specified and published in June an official specification of NVMe over fabric aka NVMf. The key component here is RDMA available on IB, Ethernet named ROCE (RDMA over Converged Ethernet) or TCP/IP. But this NVMf approach solves only half of the problem only the latency challenge with remote NVMe drives but not the resiliency and other features represented by the I/O stack. To solve that, 2 choices exist:
  1. implement data services on the host with a sort of very thin storage agent. It allows scale-out offering with a SDS model with features implemented outside the array, the logic is running on the host. But some questions exist such coordination between agents et still the resiliency of the platform. But among the 2 modes, this one is the most scalable of course.
  2. run these in the controller itself that make possible only scale-up products pretty similar to "classic" controller-based storage appliance and as mentioned you scale "behind" the controller.

The page 25 of the presentation below shows a good table that compares the 2 modes. Data Centers are evolving and NVMe and NVMf are a reality with different offerings on the market, each promoting its own value and differentiator. If you wish to read some vendors collateral, go to Apeiron Data Systems, DSSD, E8 Storage, Excelero, Kaminario, Mangstor or Mellanox.

Jun 28, 2016

Backblaze B2 is GA with interesting partners

Backblaze (www.backblaze.com), the famous consumer backup provider, continues to iterate its Cloud offering and now officially announces B2 as GA. The company already participated to 2 IT Press Tour and they were recently invited by OpenIO to jointly announced their partnership covering Hybrid Cloud. First let me summarize what is B2 to then introduce OpenIO integration.
B2 is a Cloud Storage offering like Amazon S3, Google Cloud Storage or Microsoft Azure at a fraction of the cost i.e $0.005/GB/month. It is easy to use via 3 methods: WebUI with drag and drop, API integration with dedicated product and CLI.

OpenIO is the only Backblaze partner dedicated to enterprise, service providers or cloud providers and the product supports natively now with the 16.07 release the option to migrate data (it's a move) to a new storage tier, cloud based, represented by B2. It means that OpenIO did the job to integrate the B2 API additionally to the tiering logic implemented in the storage policy. OpenIO is deployed on-premise and users can integrate an external cloud tier completely transparently. The namespace is fully maintained by OpenIO, acting as an in-band logic via a specific gateway service, and the application never interacts directly with Backblaze B2.

Interesting for both companies, OpenIO receives a new argument for US presence and penetration and same from Backblaze not yet present in Europe as the only data center is in Rancho Cordova, near Sacramento, in California. But as of today, I'm not convinced that European OpenIO clients will migrate content to California-based B2 center via the internet... With this announcement, Backblaze also supports 4 other partners - CloudBerry Lab, Cyberduck, Dropshare and Synology - as source of traffic for B2 meaning that this a SMB or departemental play. These 5 partners have made some commitment and integration with the B2 API but I'm still don't understand why B2 didn't pick Amazon S3, the only viable cloud storage interface and API.

"S3 is the De Facto Cloud Storage Standard serving as the backbone for 90% of today's web companies.
In 2 words, the game is over globally.

With S3, immediately and by default users will be able to send traffic to B2, generate revenue and penetrate the market. For further info and even wish to integrate B2, please refer to https://www.backblaze.com/b2/cloud-storage.html page with step-by-step guide, examples or codes.
Is it a try for a few months, is it a choice to control traffic and avoid any overload to the infrastructure ? Still a mystery. But, choosing a specific API late after S3, just new with no market presence is super risky. I let you imagine the success of the solution if only 5% of each installed base from these 5 partners will extend usage with B2... So in 2 words, I wish and expect that Backblaze will move rapidly to S3 to boost the market adoption.

Jun 24, 2016

Noobaa super simple scalable S3 storage for masses

NooBaa (www.noobaa.com), a new storage Israeli ISV, continues its incremental visibility effort on the market. I had the privilege to meet Yuval Dimnik, CEO and co-founder, and Mike Davis, CMO, a few days ago in Palo Alto for a briefing. The first feeling I had was great as they plan to introduce a real new storage animal to the market. Let me summarize what I learned without telling any secret as the web site already exists. First, NooBaa has a strong DNA in storage and networking with many leaders having worked for Exanet, NooBaa's CEO and Guy Margalit, CTO, have worked there. The company, today with approximately 15 people and a HQ in Boston and R&D center in Rehovot in Israel, was founded in 2013 and is backed by JVP and Ourcrowd for an unknown amount, OurCrowd has ingested $934,504. The product they design and develop belongs to the object storage category with key differentiators, it is a full SDS philosophy dedicated to unstructured data.

NooBaa product is a pure software approach, offering only S3 interface running on any compute resource available anywhere supporting heterogeneous network-driven and shared storage entities. The product offers also deduplication with a sliding window technique, compression, encryption and replication to protect data with by default 3 copies. Erasure coding is on the roadmap. NooBaa uses a asymmetric model with side metadata servers and data servers acting as chunk servers. Metadata servers use MongoDB. The demo the team has made was super efficient having demonstrated how it is easy and fast to deploy, run and operate. Just 15 minutes and we played with the system, ingest some data, stream a video. The pricing has 2 modes: Freemium and then a pay-per-user subscription base on real used capacity (real capacity for one data copy). The product is in Beta mode and should be GA early 2017. Great first feeling with the team and a super promising product, lots of potential and real interesting ideas.

Jun 22, 2016

Cloudian is the real S3 leader

Cloudian (www.cloudian.com), the best On-premise S3 Cloud Storage, did a fantastic presentation yesterday at the IT Press Tour. The business path for the company is pretty impressive with bookings multiplied by 3, no churn at all, 90% win ratio and some recent super deals beating some competitors that move to the marketing side of the market. All these success is the result of a superb recruitment and a precise market strategy, thanks to Paul Turner, John Ash and of course Michael Tso, among others. One of the top recent news is the oem agreement with Lenovo, who finally even with an implicit link with IBM, has chosen Cloudian. The key reason for that is simple: Cloudian is the best S3 product for on-premise cloud storage, with S3 the de-facto market standard. For Lenovo and end-users, standards are fundamental and help for the adoption.

Michael Tso - CEO and co-founder Cloudian
The second advantage of Cloudian versus the competition is the development and positioning of the platform promoting a new SDS acronym i.e Smart Data Storage with 3 dimensions behind it: Smart Storage Operations, Smart Storage Platform and Smart Storage Analytics. HyperStore is a very comprehensive storage platform, available as a software or as an appliance, and from the PAM - Protect Access and Manage - criteria, is among the few best ones.
When Cloudian needs to develop new feature, they adopt a common sense attitude and the company understands for a long time that the differentiator is not only the feature itself but the recognized technology behind it with a time to market dimension. Cassandra and Redis are 2 good examples for that, Cloudian uses the first DB solution for metadata and the second one for user credentials. Among the long list of new features, here is an extract:
  • Now for the metadata search, Cloudian has integrated ElasticSearch within the product and everything is masked for a superb result.
  • Same thing for the Analytics part with the integration of Spark. These 2 examples are also good illustrations of the power of open source at scale.
  • Storage policies at the bucket level.
  • QoS and Secure Storage with alignement with some regulations needs plus Authorization and Authentication methods such ACL...
  • We had the privilege during the session to receive a demo of the product, the GUI is just a beauty, super intuitive and very easy to use. There is also a panel with an hardware view where you can see the physical reality of different components, super useful when there is a failure especially when the cluster is big, very big.
  • Cloudian has also an interesting rolling upgrade method, provides a pro-active disk repair mechanism with Read-On-Repair operation and Smart Redirect during writes. HyperStore offers a Smart Disk Balancing mechanism to optimize the balancing period and address new available resources faster.
  • They also provide a Cloud to Cloud copy i.e HyperStore to HyperStore or HyperStore to AWS.
  • As a true believer of the ecosystem, I was instrumental in the success of some companies with that kind of program I built, Cloudian has made a good demo with CommVault Simpana via S3 and Veeam via NFS, super simple to operate.
  • Now in term of support, Cloudian has introduced a continuous monitoring solution named HyperStore Smart Support, a service pretty similar to Nimble Storage InfoSight that finally serves as a reference for many implementations and vendors. The service seems to be also very comprehensive and is also integrated with Salesforce.
It was a good session, we feel and saw progress made by Cloudian in many aspects: product, operation, sales, strategy... definitely one of the top object storage solutions of the market, no doubt.

Jun 21, 2016

OpenIO continues its technology journey

OpenIO (www.openio.io), alternative object storage player and recent market entrant, has been met yesterday by the IT Press Tour crew. The company has made some significant efforts for a few months to improve its visibility especially to better communicate technologies development. This is critical for the success and the market penetration.

Let me summarize a few key points about OpenIO. Market promotes Software-Defined Storage and OpenIO is a perfect illustration of that and it's a real one like Hedvig or Minio. The philosophy of the product confirms that the scalability in performance and capacity is well achieved with a asymmetric model meaning the use of side metadata servers and JBoCS (Just a Bunch of Chunk Servers) talking with client machines. The product doesn't use any consistent hashing-based algorithm, so there is no need of rebalance and new servers so storage are immediately available for production. You can start super small and for sure there is potentially no need for object at that size except if you wish to do remote access and above all grow fast, often and becoming big, very big. The solution can be also very easily and rapidly deployed, Paul Delaporte, Product Manager, did a live deployment in 5:43, really impressive.
About the news, OpenIO introduces 3 things: first, the Storage Tiering feature, supporting internal and external capability, moving content from one pool to a secondary one, second, a detailed new iteration for Kinetic integration and a 3rd one under embargo for a few days.

Jun 20, 2016

Versity Software, a confidential leader

Versity Software (www.versity.com), an innovative data management ISV founded in 2011, was a good surprise of the last IT Press Tour edition. I already wrote a few weeks ago a short post about Versity's presentation delivered during the MSST 2016 conference. I know the 2 founders - Bruce Gilpin, CEO, and Harriet Coverston, CTO - for some times and always consider seriously what they built and developed. Bruce Gilpin was our host for that session.

First the description of the company changed a bit since I saw a presentation from Versity last month, instead of "the only independent pure play HSM software company", the message is now "the only pure play data archiving software company", which is clearly broader as HSM really means specific things, has limitations and doesn't have always a good critic around it. The first comment is that the segment exists for decades with established leaders and the market has changed a bit the approach with the so famous ILM wave 10 years ago aggregating HSM, Tiering, Migration and Archiving together glued to the application.
For Versity the value resides in software, Bruce Gilpin promotes the open source as the long term run is critical for archive. The companion to that is of course commodity hardware with the need to really scale by unifying and aggregating resources horizontally. The other key differentiator in favor of Versity that provides a real value to users is the subscription model with no minimal contractual term with a protection and ability to read data for ever even if the contracts is ended. The long term investment protection for archive with Versity is also offered as VSM - Versity Storage Manager - uses an open source GNU TAR format. In term of business, Versity has signed a strategic deal with Cay that offers the product as TAS - Tiered Adaptive Storage -, the HPC leader has made a investment in the company in 2013. Recently, Versity signed with a key vertical player - Ovation Data Systems - for a Data Lifecycle solution based on VSM. The company continues to target vertical use cases such Finance, Oil and Gas, M&E or Healthcare/Medical and Genomics.

VSM was available in April 2014 with the 1.0 release. VSM is a proprietary software product constructed around SAM-QFS core technology thanks to the CDDL license model chosen by Sun Microsystem a long time ago. VSM software retains the same command set and administration methods used for SAM-QFS file system administration. The product has 2 modules, a kernel one with the Versity Shared File System, and a user land one with the Application Archive Manager layer with the policy engine, the media management and the tiered storage service. The new iteration of the product is the announcement of the object storage back-end via the de-facto standard Amazon S3 API run in production at Ovation Data Systems in Houston. This was a market need to adress new capabilities and avoid tape limitations such: fast to 1st byte, fast without the need of a fast and costly network and remote access. The idea is clearly the mix of the 2 worlds with Posix for applications and cloud object storage for secondary data entities glued by Versity. As Cloudian is the best S3 object storage, coupling the 2 solutions should provide a real comprehensive solution to the market. Think about that.

Jun 14, 2016

New iteration in Data Reduction

Ascava Inc., no web site yet neither a logo, is making progress in their mission to change the Data Reduction landscape. The company, founded in 2007 and incorporated in Delaware, has raised $1.59M in Nov. 2014, the team is currently located in Los Altos Hills, CA. I had the privilege to meet at SNIA DSI conference 2 key executives of the company: Harsh Sharangpani, CEO and CTO, and Rajesh Patil, VP Business Development and Operations. LinkedIn shows 5 people in addition to Rajesh not listed as Ascava.
As mentioned at the beginning of this post, the product is about data size reduction what the company named Data Distillation with ratio 1.5 to 2x better than the best DeDup + Compression ratio existing on the market. Ascava has made several innovations in that sector and is not ready yet to announce anything but developments are in sync with the plan.
The product is a software running as a separate standalone application you operate on any node able to shrink file size and return a super optimized reduced file. The outcome is obvious less consumed space for a fraction of additional cpus cycles. Of course you need a "reader" able to unpack the file to give you back the original content before any application reads it in its original content. We can imagine multiple usages of this such data archiving, data analysis and file tiering/HSM/migration stuff. It reminds me a bit what Ocarina Networks did in the past even if here Ascava is doing radical new things to achieve this reduction ratio. For whose who wish to read a few things about Ocarina, I wrote a few posts, sorry in French at that time, available here: Aug. 2007, Apr. 2008 and July 2010. Scava was a good meeting even a surprise and illustrates that some initiatives and companies continue to dig in that space. CPU and memories continue to be super fast and affordable, we then imagine that some innovators wish to move forward and propose something new, Ascava is one of these. Good Luck.