68: GreyBeards talk NVMeoF/TCP with Ahmet Houssein, VP of Marketing & Strategy @ Solarflare Communications

In this episode we talk with Ahmet Houssein, VP of Marketing and Strategic Direction at Solarflare Communications, (@solarflare_comm). Ahmet’s been in the industry forever and has a unique view on where NVMeoF needs to go. Howard had talked with Ahmet at last years FMS. Ahmet will also be speaking at this years FMS (this week in Santa Clara, CA)..

Solarflare Communications sells Ethernet communication gear, mostly to the financial services market and has developed a software plugin for the standard TCP/IP stack on Linux that supports both target and client mode NVMeoF/TCP. That is, their software plugin provides a complete implementation of NVMeoF across TCP Ethernet that extends the TCP protocol but doesn’t require RDMA (RoCE or iWARP) or data center bridging.

Implementing NVMeoF/TCP

Solarflare’s NVMeoF/TCP is a free plugin that once approved by the NVMe(oF) standard’s committees anyone can use to create a NVMeoF storage system and consume that storage from almost anywhere. The standards committee is expected to approve the protocol extension soon and sometime after that the plugin will be added to the Linux Kernel. After standards approval, maybe VMware and Microsoft will adopt it as well, but may take more work.

Over the last year plus most NVMeoF/Ethernet we encounter requires sophisticated RDMA hardware. When we talked with Pavilion Data Systems, a month or so ago, they had designed a more networking like approach to NVMeoF using RoCE and TCP a special purpose FPGA that’s used in their RDMA NICs and Mellanox switches to support client-target mode NVMeoF/UDP [updated 8/8/18 after VR’s comment, the ed.]. When we talked with Attala Systems, they had special purpose FPGA that’s used in RDMA NICs and Mellanox switches to support target & client mode NVMeoF/UDP were using standard RDMA NICs and Mellanox switches to support their NVMeoF/Ethernet storage [updated 8/8/18 after VR’s comment, the ed.].

Solarflare is taking a different tack.

One problem with the NVMeoF/Ethernet RDMA is compatibility. You can use either RoCE or iWARP RDMA NICs but at the moment you can’t use both. With TCP/IP plugins there’s no hardware compatibility issue. (Yes, there’s software compatibility at both ends of the pipe).

SolarFlare recently measured latencies for their NVMeoF/TCP (Iometer/FIO) which shows that the with the protocol running it adds about a 5-10% increase in latency versus running RDMA NVMeoF/UDP-RoCE-iWARP.

Performance measurements were taken using a server, running Red Hat Linux + their TCP plugin with NVMe SSDs on the storage side and a similar configuration on the client side without the SSDs.

If they add 10% latency to 10 microsec. IO (for Optane), latency becomes 11 microsec. Similarly for flash NVMe SSDs it moves from 100 microsec to 110 microsec.

Ahmet did mention that their NICs have some hardware optimizations which brings down this added latency into something approaching closer to 5%. And later we discuss the immense parallelism opportunities using the TCP stack in user space. Their hardware also better supports more threads doing IO in parallel.

Why TCP

Ahmets on a mission. He says there’s this misbelief that Ethernet RDMA hardware is required to achieve lightening fast response times using NVMeoF, but it’s not true. Standard TCP with proper protocol enhancements is more than capable of performing at very close to the same latencies as RDMA, without special NICs and DCB switch configurations.

Furthermore, TCP/IP already has multipathing support. So current high availability characteristics of TCP are readily applicable to NVMeoF/TCP

Parallelism through user space

NVMeoF/TCP was the subject of 1st half of our discussion but we spent the 2nd half talking about scaling or parallelism. Even if you can do 11 or 110 microsecond latency at some point, if you do enough of these IOs, the kernel overhead in processing blocks and transferring control from kernel space to user space will become a bottleneck.

However, there’s nothing stopping IT from running the TCP/IP stack in user space and eliminating any kernel control transfer whatsoever. By doing so, data centers could parallelize all this IO using as many cores as available.

Running the plugin in a TCP/IP stack in user space allows you to scale NVMeoF lightening fast IO to as many users as you have user spaces or cores, and the kernel doesn’t even break into a sweat

Anyone could simply download Solarflare’s plugin, configure a white box server with Linux and 24 NVMe SSDs and support ~8.4M IOPS (350Kx24) at ~110 microsec latency And with user space scaling, one could easily have 1000s of user spaces connected to it.

They’re going to need need faster pipes!

The podcast runs ~39 minutes. Ahmet was very knowledgeable about NVMe, NVMeoF and TCP.  He was articulate and easy to talk with.  Listen to the podcast to learn more.

Ahmet Houssein, VP of Marketing and Strategic Direction at Solarflare Communications 

Ahmet Houssein is responsible for establishing marketing strategies and implementing programs to drive revenue growth, enter new markets and expand brand awareness to support Solarflare’s continuous development and global expansion.

He has over twenty-five years of experience in the server, storage, data center and networking industry, and held senior level executive positions in product development, marketing and business development at Intel and Honeywell. Most recently Houssein was SVP/GM at QLogic where he successfully delivered first to market with 25Gb Ethernet products securing design wins at HP and Dell.

One of the key leaders in the creation of the INFINIBAND and PCI-Express industry standard, Houssein is a recipient of the Intel Achievement Award and was a founding board member of the Storage Network Industry Association (SNIA), a global organization of 400 companies in the storage market. He was educated in London, UK and holds an Electrical Engineering Degree equivalent.

60: GreyBeards talk cloud data services with Eiki Hrafnsson, Technical Director, NetApp

Sponsored by:In this episode, we talk with Eiki Hraffnsson (@Eirikurh), Technical Director, NetApp Cloud Data Services.  Eiki gave a great talk at Cloud Field Day 3 (CFD3), although neither Howard nor I were in attendance. I just met Eiki at a NetApp Spring Analyst event earlier this month and after that Howard and I had a chance to talk with him about what’s new in NetApp Cloud Data Services

This is the fourth time NetApp has been on our show (see our podcast with Lee Caswell and Dave Wright,  podcast with Andy Banta, & last month’s sponsored podcast with Adam Carter) and this is their second sponsored podcast.

Eiki came from a company NetApp acquired last year called GreenQloud whose product was QStack. Since then, QStack has become an integral part of their Cloud Data Services.

NetApp has a number of solutions under their Cloud Data Services umbrella and his area of specialty is NetApp Cloud Data Volumes, soon to be available in the MarketPlace on AWS, already in public preview an Microsoft Azure Enterprise NFS and as of 7 May 2018, in private preview as NetApp Cloud Volumes for Google Cloud Platform.

NetApp Cloud Data Volumes

NetApp’s Cloud Data Volume is a public cloud based, storage-as-a-service that supplies enterprise class NFS and SMB (CIFS) storage on a pay as you go model for major public cloud providers. That way your compute instances can have access to predictable performance, highly available file storage in the  cloud.

One advantage that Cloud Data Volumes adds to the public cloud is performance SLAs. That is customers can purchase Low, Medium and High performance file storage. Eiki said they measured Cloud Data Volume IO performance and it achieved almost 10X the public cloud normal (file) storage performance. I assume this was HIGH performing Cloud Data Volume storage, and no information on which storage type was used as the cloud alternative.

Cloud Data Volume customers also get access to NetApp Snapshot services which can create, space efficient, quick read-only copies of their cloud file storage. Cloud Data Volume storage can be purchased on a $/GB/month basis. Other  purchase options are also available for customers who prefer a pre billed amount rather than a consumptive model.

Behind the scenes, Cloud Data Volumes is actually NetApp ONTAP storage. They won’t say what kind or how much, but they do say that NetApp storage is located in public cloud data centers and is fully managed by NetApp.

Customers can use the public cloud native services portal to purchase Cloud Data Volume storage (for Microsoft Azure and GCP) or the NetApp Cloud web portal (for AWS). Once purchased, customers can use an extensive set of native cloud APIs to provision, access and tear-down Cloud Volume storage.

Other NetApp Cloud Data Services

Eiki mentioned that Cloud Data Volumes is just one of many offerings from NetApp’s Cloud Data Services business unit, including:

  • NetApp Private Storage– colocated NetApp storage owned by customers that is adjacent to public clouds.
  • ONTAP Cloud – software defined ONTAP storage system that run in the cloud on compute services using cloud storage to provide block storage.
  • Cloud Sync – data synchronization as a service offering used to replicate data from onprem NAS and object storage to the public cloud.

Probably a few others I am missing here and my bet is more offerings are on the way.

Another item Eiki mentioned with the open source,  NetApp Trident Plugin (GitHub repo). Containers are starting to need persistent state information and this means they need access to storage.

Trident provides dynamic, API driven provisioning of storage volumes for containers under Kubernetes.  Container developers define environmental characteristics which dictate operational environment and now with Trident, can also specify needed storage volumes. That way, when Kubernetes fires up a container for execution, NetApp storage is provisioned just-in-time to support container stateful execution.

The podcast runs ~25 minutes. Eiki was very knowledgeable and was easy to talk with especially on cloud technologies and how NetApp fits in.  Listen to the podcast to learn more.

Erikur (Eiki) Hrafnsson, Technical Director, NetApp Cloud Data Services

Erikur (Eiki) Hrafnsson is an entrepreneur, dad, singer. founder of GreenQloud and maker of QStack, the hybrid cloud platform, now part of NetApp Cloud Data Services. Eiki brings deep public cloud integration knowledge and broad experience in cloud automation and APIs.

39: Greybeards talk deep storage/archive with Matt Starr, CTO Spectra Logic

In this episode, we talk with Matt Starr (@StarrFiles),  CTO of Spectra Logic, the deep storage experts. Matt has been around a long time and Ray’s shared many a meal with Matt as we’re both in NW Denver. Howard has a minor quibble with Spectra Logic over the use of his company’s name (DeepStorage) in their product line but he’s also known Matt for awhile now.

The Pearl

Matt and Spectra Logic have a number of customers with multi-PB to over an EB of data repository problems and how to take care of these ever expanding storage stashes is an ongoing concern.  One of the solutions Spectra Logic offers is the Black Pearl Deep Storage, which provides an object storage, RESTfull interface front end to storage tiering/archive backend that uses flash, (spin-down) disk, (LTFS) tape (libraries) and the (AWS) cloud as backend storage.

Major portions of the Black Pearl are open sourced and available on GitHub. I see several (DS3-)SDK’s for Java, Python, C, and others. Open sourcing the product provides an easy way for client customization. In fact, one customer was using CEPH and they modified their CEPH backup client to send a copy of data off to the Pearl.

We talk a bit about the Black Pearl’s data integrity. It uses a checksum, computed over the object at creation time which is then verified anytime the object is retrieved, copied, moved or migrated and can be validated periodically (scrubbed), even when it has not been touched.

Super Computing’s interesting (storage) problems

Matt just returned from the SC16 (Super Computing Conference 2016) in Salt Lake City last month. At the conference there were plenty of MultiPB customers that were looking for better storage alternatives.

One customer Matt mentioned  was the Square Kilometer Array, the world’s largest radio telescope which will be transmitting 700TB/hour, over an 1EB per year.  All that data has to land somewhere and for this quantity (>eb) of data, tape becomes an necessary choice.

Matt likened Spectra’s  archive solutions to warehouses vs. factories. For the factory floor,  you need responsive (AFA or hybrid) primary storage but for the warehouse, you just want cheap, bulk storage (capacity).

The podcast runs long, over 51 minutes, and reveals a different world from the GreyBeards everyday enterprise environments. Specifically customers that have extra large data repositories and how they manage to survive under the data deluge. Matt’s an articulate spokesperson for Spectra Logic and their archive solutions and we could have talked about >eb data repositories for hours.  Listen to the podcast to learn more.

matt-starrMatt Starr, CTO, Spectra Logic

Matt Starr’s tenure with Spectra Logic spans 24 years and includes experience in service, hardware design, software development, operating systems, electronic design and management. As CTO, he is responsible for helping define the company’s product vision, and serves as the executive representative for the voice of the market. He leads Spectra’s efforts in high-performance computing, private cloud and other vertical markets.

Matt served as the lead engineering architect for the design and production of Spectra’s TSeries tape library family. Spectra Logic has secured more than 50 patents under Matt’s direction, establishing the company as the innovative technology leader in the data storage industry. He holds a BS in electrical engineering from the University of Colorado at Colorado Springs.

33: GreyBeards talk HPC storage with Frederic Van Haren, founder HighFens & former Sr. Director of HPC at Nuance

IMG_6319In episode 33 we talk with Frederic Van Haren (@fvha), founder of HighFens, Inc. (@HighFens), a new HPC consultancy and former Senior Director of HPC at Nuance Communications. Howard and I got a chance to talk with Frederic at a recent HPE storage deep dive event, I met up with him again during SFD10, where he was talking on behalf of Kaminario, and he was also at HPE Discover conference last week.

Nuance is the backend speech recognition engine for a number of popular service offerings. Nuance looks very similar to a lot of other hyper-scale customers and ultimately, we feel may be the way of the future for all IT over the coming decades.  Nuance’s data storage journey since Frederic’s tenure with the company holds many lessons for all of us in the storage industry

Nuance currently has ~6PB usable (~16PB raw) of speech wave files as well as uncountable text and other files, all inside IBM SpectrumScale (GPFS).  They have both lots of big files and lots of small files. These days, Spectrum Scale is processing 2-3M files/second. They have doubled capacity for each of the last 9 years, and today handle a billion new files a month. GPFS stripes data across storage, provides data protection, migration, snapshotting and storage tiering across a diverse mix of storage. At the end of the podcast we discussed some open source alternatives to Spectrum Scale but at the time Nuance started down this path,  GPFS was found to be the only thing that could do the job. This proved to be a great solution as they have completely swapped out the underlying storage at least 3 times and all their users were none the wiser.

The first storage that Frederic talked about was Coraid (no longer in business) and their ATA over Ethernet storage solution. This used a SuperMicro with 24 SATA drives/shelf and they bought 40 shelves. Over time this grew to 1000s of SATA drives and was easily scaleable but hard to manage, as it was pretty dumb storage. In fact, they had to deploy video cameras, focused on drive shelves, to detect when drives failed!

Overtime, Nuance came to the realization that they had to do something more manageable and brought in HPE MSA storage to replace their Coraid storage. The MSA was a great solution for them which had 96 SAS drives, were able to support both faster “SCRATCH” storage using fast SAS 300GB/15KRPM drives and slower “STATIC” storage with slower SATA 760GB/7.2KRPM drives and was much more manageable than the Coraid solution.

Although MSA storage worked great, after a while, Nuance’s sprawling FC environment which was doubling yearly, caused them to rethink their storage once again. This led them to swap out all their HPE MSA storage, for HPE 3PAR to consolidate their FC network and storage footprint.

For metadata, Nuance uses a 76 node, Hadoop cluster for sophisticated search queries as doing an LS on the GPFS file system would take days. Their file meta-data is essentially a textual, row by row database and they use queries over the Hadoop cluster to determine things like which files have american english, spoken by females, with 8Khz recording.  Not sure when, but eventually Nuance deployed HPE Vertica SQL over Hadoop for their metadata engine and dropped average query from 12 minutes to 73 sec.(!!)

Nuance, because of their extreme growth and more open environment to storage innovation, had become a favorite for storage startups and major vendors to do Proofs of Concepts (PoC) on new storage offerings. One PoC, Nuance did was for Kamanario storage. There is a standard metric that says a CPU core requires so many IOPS, so that when CPU cores  increase,  you need to supply more IOPS. They went with Kaminario for their test-dev environment and more performance intensive storage. Nuance appreciates Kamanario’s reliability, high availability and highly predictable performance. (See the SFD10 video feed for Frederic’s session)

We talked a bit about how speech recognition’s Hidden Markov Chain statistical model was heavily dependent on CPU cores. Today, if you want to do a recognition task, you assigned it to one core and waited until it was done, a serial process dependent on the # of CPU cores you had available. This turned out to be quite a problem as you had to scale CPU cores if you wanted to do more concurrent speech recognition activities. Then came GPUs and you could do speech recognition work on a GPU core. With the new GPU cards,   instead of a server having ~16 CPU cores,  you could have a server with multiple Graphic cards having 3000-GPU cores. This scaled a lot easier. Machine learning and deep neural nets have the potential to parallelize this, so that it will scale even better

In the end, HPC trials, tribulations and ways of doing business are starting to become  mainstream. I was recently talking to one vendor that said, most HPC groups start out in isolation to support one application but over time they either subsume corporate IT or get absorbed into corp. IT or continue to be a standalone group (while waiting until one of the other two happen).

The podcast runs ~41 minutes and  covers a lot of ground about one HPC organization’s evolution of their storage environment over time, what was driving some of that evolution and the tools they chose to master it.  Listen to the podcast to learn more.

0F2A7849 - Copyv2-resizedFrederic Van Haren, founder HighFens, Inc.

Frederic Van Haren is the Chief Technology Officer @Highfens and known for his insights in the HPC and storage industry. He has over 20 years of experience in High Tech providing technical leadership and strategic direction in Telecom and Speech markets. Frederic spent the last decade at  Nuance Communications building large HPC environments from the ground up. He is frequently invited to speak at events to provide his insights on the HPC and storage markets. He has played leading roles as President of a variety of technology user groups promoting the use of innovative technology. As an Engineer he enjoys working with the engineering teams from technology vendors providing feedback on new and upcoming products.

Frederic lives in Massachusetts,  USA but grew up in the northern part of Belgium where he received his Masters in Electrical Engineering, Electronics and Automation.

Greybeards talk car videos, storage and IT trends with Marc Farley

In our 30th episode, we talk with 3rd time guest star,  Marc Farley (@GoFarley), Formerly with Datera and Tegile. Marc has recently gone on sabbatical and we wanted to talk to him about what was keeping him busy and what was going on in storage/IT industry these days.

Marc is currently curating a car comedy vlog called theridecast.com. Apparently people, at least in California, are making comedy videos in their cars. They can be quite hilarious, checkout this episode of comedian in cars getting coffee.

While in the storage biz, the industry is getting battered by a number of trends: IT shrinking budgets, vendor proliferation, migration to cloud, and flash becoming old hat. Marc makes multiple points as to why the storage market is undergoing such a major transition these days:

  • Death to tech refresh, long live the cloud –  yes the cloud does upgrade hardware but  planned storage system obsolesce doesn’t happen in the cloud anymore. Cloud providers are  buying new SSDs, disks, white box servers, memory etc,  but not enterprise class storage, server or networking hardware.
  • AFA is boring, but selling – every vendor’s got one , two or sometimes three and they all know how to provide flash storage services. Customers pay extra for AFA, whether they need to or not, because they are swapping out old expensive, enterprise class storage for AFAs that often cost less but still provide better performance..
  • Tail IO latency becoming more important but it’s not understood – when IO response times go from 100µsec to 10msec, it hurts. It doesn’t matter if it’s every 1000 or 10,000 IOs, customers want less performance variability, which is a main reason they move to AFA in the first place. But not all AFA’s perform the same in tail latency and SSD controller/system architecture make a big difference.
  • Hybrid storage survives but only if you go big – hybrid storage economics makes sense only for large, diverse data repositories, that mix user directories, non-performance sensitive apps, and other structured and unstructured data in one data store.
  • Greenfield apps & secondary storage are moving to the cloud but migrating current apps to the cloud is difficult –  for new app development and archive storage, moving to or starting in the cloud is a no-brainer. Transitioning running enterprise class apps to the cloud is tough to do, that requires multiple skill sets and may never be successful. Hybrid  (cloud-on premises) enterprise class apps are too arduous to even contemplate.
  • Realtime analytics is emerging but data needs to be on flash – yes MapReduce is a batch activity which can uses lots of slow disk but there’s more to analytics than MR, and doing log analysis, in anything approaching realtime, one needs flash performance.
  • Optical’s persistence is great but who leaves data on the same technology for  20 years –  with magnetic and electronic storage densities going up every couple of years, who could afford keep data on the same optical technology that was 20 years old. Imagine using microfiche to keep PB of data today, inconceivable.

As for IT in general, one limiter of IT activity will become the lack of skilled engineers, specifically full-stack engineers and data scientists.

We ended our discussions on the economics of Samsung 3D NAND and Intel-Micron (IM) 3D Xpoint non-volatile memories. Both new semiconductor technologies are always long term investments. Today, Samsung is probably losing money on each 3D TLC NAND SSD it sells, but over time, as  fab yields improve, it should become cheap enough to make a profit. Similarly, 3D Xpoint may be costly to produce early on, but as IM perfect  their fab processes, the technology should become inexpensive enough to make oodles of $s for them. And there’s more technology changes to come.

The podcast runs just over 40 minutes and covers a lot of ground. Marc’s been in the IT almost as long as the GreyBeards and has a unique perspective on what’s happening today, having been with so many diverse, major and (minor) startup vendors throughout his tenure in the industry.  Listen to the podcast to learn more.

Marc Farley


Marc is a storage greybeard who has worked for many storage companies and is currently on sabbatical. He has written three books on storage including his most recent, Rethinking Enterprise Storage: A Hybrid Cloud Model and his previous books Building Storage Networks and Storage Networking Fundamentals.

In addition to his writing books he has been a blogger and podcaster about storage topics while working for EqualLogic, Dell, 3PAR, HP, StorSimple,  Microsoft, and others.

When he is not working, Marc likes to ride bicycles, listen to music, spend time with his family and dote on his cats. Of course there’s that car video curation…