Technology:Unpacked

This week Chris and Martin chat to Steve Fingerhut, CMO at Pliops. Pliops has developed a “Storage Processor” in the form factor of an AIC (add-in card) that offloads storage functions from application software. This discussion looks at why the technology is needed and how Pliops has implemented the solution to drop into an existing server.

The Pliops Storage Processor acts as a high-speed key-value store, implementing in the first instance features such as data compression. As a hardware accelerator, the technology reduces CPU core overhead, saves on storage and can reduce application licensing costs. One interesting factor is the implementation of this hardware compared to other SmartNICs, in that the data path remains within the server itself.

You can find out more at https://pliops.com/

Elapsed Time: 00:34:04

Timeline

* 00:00:00 – Intros * 00:01:30 – What is the need for SmartNICs? * 00:04:00 – The gap is now flash/CPU not SSD/HDD * 00:07:00 – Pliops accelerates workloads within the server * 00:08:55 – Offload processing is not a new thing * 00:10:30 – Pliops presents as a block device * 00:14:40 – What happens with resiliency/redundancy? * 00:17:00 – How are benefits of Pliops quantified? * 00:21:00 – Offload can save application licensing costs * 00:22:24 – Right-sizing servers is hard * 00:25:40 – Is the product shipping? * 00:28:45 – hardware is more relevant again * 00:31:00 – How will composable affect these solutions? * 00:32:53 – Wrap up

Related Podcasts & Blogs

* VMware Project Monterey – First Impressions* Fixing the x86 Problem* #177 – SmartNICs and Project Monterey

Copyright (c) 2016-2020 Storage Unpacked. No reproduction or re-use without permission. Podcast episode #e62o.

Everyone thinks cloud storage is cheaper than on-premises, but is that really true? When cloud storage like AWS S3 was first introduced, vendors dropped their $/GB prices on a frequent basis. However, AWS hasn’t reduced the base price of S3 in five years, instead preferring to offer cheaper tiers of storage. In this week’s episode, Chris and Martin debate the reasons for the lack of reductions, especially in light of continuing growth in HDD capacities and reducing costs.

Elapsed Time: 00:30:58

Timeline

* 00:00:00 – Intros* 00:02:00 – Is cloud storage cheaper than on-premises?* 00:05:05 – What new storage services has AWS introduced?* 00:07:00 – There are complex rules around data data retrieval* 00:08:40 – Cloud is driving the need for FinOps skills* 00:10:33 – Data mobility can be lost with egress charges* 00:12:30 – Access patterns (like sub-object) are hard in cloud* 00:13:10 – Why have base storage prices not declined?* 00:17:00 – Cloud providers could be tiering behind the scenes* 00:18:52 – Should you build your own storage cloud?* 00:20:00 – IOPS (block) data needs to be treated separately * 00:21:40 – On-premises storage lets you sweat the hardware assets * 00:23:00 – Data is fixed, compute is ephemeral * 00:26:30 – Understand your data profiles * 00:28:00 – Cloud forces you to manage your data * 00:30:00 – Wrap Up

Related Podcasts & Blogs

* Is AWS passing on the benefits of storage media price reductions?* #166 – Infinidat Elastic Storage Pricing with Eran Brown* #55 – Storage for Hyperscalers

Copyright (c) 2016-2020 Storage Unpacked. No reproduction or re-use without permission. Podcast episode #6illj.

This week, Chris has a great conversation with Bill Schmarzo, Chief Innovation Officer at Hitachi Vantara. Bill maintains a consultancy practice within Hitachi that helps customers build processes and identify data that can be used to create business value within an organisation. In this discussion, Bill outlines the process for identifying opportunities, capturing the data and building governance around translating data into income.

Aside from the insights in how businesses can generate new revenue streams from the data in their organisation, this podcast highlights the challenges of determining which data to keep and discard, while attributing value to data being protected. We assume all data needs to be protected equally, but perhaps there needs to be an alignment between data management and data value.

Bill has three books in print with a fourth (The Economics of Data Analytics and Digital Transformation) expected in November 2020. During the podcast, Bill references the Prioritisation Matrix. This blog post provides more information on the concept and process (explaining what we couldn’t visualise on an audio podcast).

You can catch up with Bill on LinkedIn and Twitter. Here’s a link to Bill’s second book, The Big Data MBA, referenced at the end of the podcast.

Elapsed Time: 00:32:53

Timeline

* 00:00:00 – Intros * 00:02:16 – How is data monetised? * 00:03:30 – What data is worth retaining or discarding? * 00:05:00 – Data creators and owners can be very diverse * 00:06:30 – Create the business need then locate the data * 00:08:10 – Pick one idea and prove it out – Proof of Value * 00:09:45 – How is governance managed in a Data Lake? * 00:11:00 – How is data management cost aligned with data value? * 00:14:00 – Feedback loop – review the prioritisation matrix * 00:16:00 – Economies of learning are greater than economies of scale * 00:18:50 – Data is a unique business asset * 00:20:00 – How is valuable data protected? * 00:24:00 – Data is the new oil? Almost but it’s infinite * 00:26:00 – Why teach? * 00:28:30 – How does AI empower people in the workplace? * 00:30:40 – Wrap Up

Related Podcasts & Blogs

* #103 – Data Management and DataOps with Hitachi Vantara* #81 – Storage or Data Asset Management?* The $2 billion Gamble on Data Management

Bill’s Bio

Bill Schmarzo is

regarded as one of the top Digital Transformation influencers on Big Data and

Data Science. In addition to his role as Chief Innovation Officer at Hitachi

Vantara, Bill is a University of San Francisco School of Management (SOM) Executive

Fellow and an Honorary Professor at the School of Business and Economics at the

National University of Ireland-Galway. His career spans over 30 years in data

warehousing, BI and advanced analytics. Bill formerly served as CTO of Big Data

This week Chris and Martin look at SmartNIC technology and the announcement of Project Monterey. SmartNICs are offload devices that provide networking, storage and security functions with additional benefits such as centralised management. VMware has announced Project Monterey, a preview solution that takes SmartNICs and offloads storage and networking tasks from the ESXi hypervisor.

It’s clear from the discussion that SmartNIC technology needs to provide a “10x” benefit to the data centre. The technology is already in use by hyperscalers to deliver Public Cloud. With Monterey and ESXi 7, the private data centre can use a similar centralised management approach. How will this work with existing infrastructure? VMware has ported ESXi to ARM and intends to run services on ESXi on the SmartNIC card. There are so many questions this raises, so join us for a fascinating discussion on this technology.

Elapsed Time: 00:36:13

Timeline

* 00:00:00 – Intros * 00:01:30 – Project Monterey – not the IBM AIX port… * 00:03:00 – What is a SmartNIC? * 00:04:05 – Are SmartNICs new or reinvention of existing technology? 00:04:40 – Mainframes did offload first * 00:08:00 – Division of responsibilities – team wars? * 00:11:00 – What benefits do SmartNICs bring? * 00:14:25 – Digging deeper into Project Monterey * 00:15:31 – Why put ESXi on ARM? * 00:18:00 – How will hot pluggable work? * 00:22:00 – Here’s where Project Monterey could be useful * 00:24:00 – Could SmartNICs aid the adoption of Arm for primary workloads? * 00:27:40 – Could Composable see a big benefit from SmartNICs? * 00:30:00 – Return of the mainframe? * 00:31:55 – On reflection, are SmartNICs worthwhile? * 00:35:00 – Wrap Up

Related Podcasts & Blogs

* #96 – Discussing SmartNICs and Storage with Rob Davis from Mellanox* ESXi on Raspberry Pi* VMware Project Monterey – First Impressions* Fixing the x86 Problem

Copyright (c) 2016-2020 Storage Unpacked. No reproduction or re-use without permission. Podcast episode #8j46.

This week, Martin and Chris catch up with Chris Mellor in his “pandemic infused isolation”. Without a particular theme, this episode looks at the unchanging landscape of tape and media, Dell EMC building another new storage platform and Nutanix still not making a profit.

In the future zone, we speculate on AMD and Micron getting together to build an Optane challenger, the chances of Kubernetes storage being successful and whether Nebulon has a new architecture with cloud-defined storage.

Naturally the conversation is interspersed with many tangents and diversions, but hey, it wouldn’t be opinion without it! This recording is way longer than normal and for that we sincerely apologise, and hope you enjoy it – keep listening to the end though, as every minute is worth your time. As usual, if you disagree or have any comments, just get in touch.

Elapsed Time: 00:54:08

Timeline

* 00:00:00 – Intros * 00:01:00 – Reading the weather on Countryfile * 00:02:20 – Some things never change * 00:02:40 – LTO9 slowing the capacity paradigm * 00:07:00 – 3480 cartridges were in megabytes * 00:08:10 – Pushing 20TB drive capacities * 00:09:19 – Martin takes it back to tape * 00:10:00 – How do tapes take such a beating? * 00:12:00 – Pollution and computer equipment * 00:13:30 – Another Dell EMC storage platform? * 00:19:00 – We need to play Storage buzzword bingo! * 00:19:30 – Is Intel making a profit on Optane? * 00:22:00 – Could AMD pair up with Micron on 3D-XPoint? * 00:26:00 – Could AMD/Micron succeed in the console space? * 00:28:00 – Is a new Nutanix CEO change the cost of new money  * 00:30:00 – Now emerging solutions/technologies * 00:30:05 – What does the Pure/Portworx acquisition mean? * 00:33:00 – Could Kubernetes be the next OpenStack? * 00:35:30 – NetApp has rebranded! * 00:37:00 – Storage on a Stick anyone? (Nebulon) * 00:40:02 – Most storage is the same generic technology * 00:45:00 – Stop the press!  Charlie’s closing down Pure! * 00:54:00 – Wrap Up

Related Podcasts & Blogs

* #158 – Midrange Teardown with Chris Mellor* #138 – Storage Predictions for 2020 (Part I)* #139 – Storage Predictions for 2020 (Part II)

Copyright (c) 2016-2020 Storage Unpacked. No reproduction or re-use without permission. Podcast episode #5cgb.

This week Chris catches up with Ralf Colbus from IBM to talk about the evolution of FlashSystem. The FlashSystem platform is an enterprise-class block-based storage solution that scales from SMB/SME offerings to high-end all-NVMe and SCM capable devices. At the heart of the design is the software behind Spectrum Virtualise, the SAN Volume Controller or SVC.

FlashSystem now offers a standardised portfolio based on SVC. This starts with the 5010 and 5030, both SAS-based solutions that can be all-flash or hybrid flash & HDD. The 5100 upwards (including 7200 and 9200) introduces NVMe and IBM’s FlashCore modules.

FlashCore is based on technology developed from the TMS acquisition. FlashCore is effectively a custom SSD, based purely on hardware and with in-built FPGAs that deliver features which include hardware-based encryption. FlashCore also uses MRAM to eliminate super-capacitors (see episode #159). Capacities scale from 4.8TB to 38.4TB per modules and up to 2PB in 2U.

An interesting aspect in the evolution of FlashSystem is the strategy to put pricing online. We’ve talked about transparent pricing recently (episode #173) and it’s good to see IBM both being open about the cost of solutions, while offering new, flexible terms, including “pay as you go”.

To learn more, check out the IBM Storage Digital Platform.

Elapsed Time: 00:41:30

Timeline

* 00:00:00 – Intros * 00:01:00 – Where did SVC & FlashSystem develop from? * 00:03:06 – Storwize brought a brand name and some IP * 00:03:55 – IBM Standardized the portfolio around SVC * 00:05:00 – Low-end FlashSystem can be hybrid storage (HDD & SSD) 00:06:26 – 5100 models upwards use NVMe in place of SAS * 00:07:20 – 5100 onwards introduces hardware-based encryption * 00:08:06 – FlashCore – custom flash SSD modules based on TMS IP * 00:10:00 – FlashCore is all hardware with FPGAs * 00:13:00 – SCM/PMEM (Optane) can be used as a tier or for pinned LUNs * 00:17:00 – FlashSystem scales from two to eight nodes (4-way) * 00:21:00 – SVC is used within FlashSystem as the software layer * 00:23:00 – SVC (and so FlashSystem) offers the capability of 100% uptime * 00:25:00 – FlashSystem supports CSI and automation tools like Ansible * 00:28:45 – How has IBM changed their FlashSystem pricing strategy? * 00:32:00 – IBM offers flexible “on-demand” pricing * 00:35:00 – How has IBM evolved Storage Resource Management tools? * 00:39:00 – IBM is using Watson to create Storage AIOps insights * 00:40:25 – Wrap Up

Ralf’s Bio

Ralf Colbus is Chief Storage Strategist for IBM Systems, designing secure and cost-effective data storage solutions for customers and business partners in the DACH (Germany, Austria, Switzerland) region. He has over 20 years of experience working in data storage across several industries, and is very passionate about helping clients understand the power of data in supporting strategic decisions which will give them competitive advantage. 
Outside the world of data, he enjoys music, good wine, playing golf, and advocating for environmental causes such as climate change.

Copyright (c) 2016-2020 Storage Unpacked. No reproduction or re-use without permission. Podcast episode #gc78.

This week, Chris and Martin chat to Phil Bullinger, Senior VP and General Manager for the Data Centre Business Unit at Western Digital. As storage media capacities increase, recording methods are introducing challenges to maintaining resiliency and performance. SMR (Shingled Magnetic Recording) and ZNS (Zoned Namespaces) are two techniques that have developed to address the scaling issues in modern media devices.

SMR is a technique for hard drives that overlays the recording area of tracks on storage media to gain increased areal density. This results in a requirement to re-write entire blocks or zones of disk that could be tens of megabytes in size.

ZNS divides solid-state media into zones to improve I/O performance and increase device endurance. This is achieved through reduction in write amplification and reduced garbage collection.

With both media, the recording techniques continue to help reduce the cost of devices and maintain the downward trend of $/GB for modern media. This discussion introduces us to the benefits and some of the challenges. If you want to learn more, Western Digital sponsors an independent website at http://zonedstorage.io/.

Elapsed Time: 00:45:08

Timeline

* 00:00:00 – Intros * 00:02:00 – What are today’s storage & data challenges? * 00:04:20 – HDDs & SSDs continue to grow in capacity but have challenges * 00:06:40 – I/O density represents a big problem for today’s media * 00:09:00 – What is zoned storage? * 00:12:41 – HDD media grains are driving new ways of reading/writing * 00:14:00 – SMR overmap * 00:16:50 – Zoned storage is designed for sequential workloads * 00:19:05 – What is ZNS – Zoned Namespaces? * 00:21:00 – SSDs try to emulate hard drives * 00:23:40 – ZNS can simplify the operation of reads and writes * 00:26:00 – ZNS can make SSD performance more deterministic * 00:28:40 – SMR has drive managed, host managed or host aware * 00:32:00 – Data centre builds are al about TCO, including storage * 00:34:34 – Will SDS see greater benefits from zoned storage? * 00:38:50 – Most data is written once and never amended * 00:42:00 – ZNS will improve the endurance of QLC * 00:43:00 – Wrap Up

Related Podcasts & Blogs

* #113 – The Expanding Storage Hierarchy* #55 – Storage for Hyperscalers* #161 – Seagate MACH.2 Dual Actuator Drive Deep Dive* Managing Massive Media

Phil’s Bio

Phil

Bullinger is Senior Vice President and General Manager of the Data Center Business

Unit for Western Digital where he focuses on accelerating the growth and

performance of the company’s broad portfolio of data center disk and flash

products.  

Previously,

Bullinger was Senior Vice President and General Manager at Dell EMC, where he

was responsible for the Isilon product line, including product planning,

hardware and software engineering, production operations and customer support.

He also held executive positions at Oracle as their Senior Vice President of

SAN/NAS Storage, and at LSI as the Executive Vice President and General Manager

for the Engenio Storage Group.

Bullinger’s

Enterprise storage pricing has all the simplicity of a mobile phone tariff. Vendors love to obfuscate the costs, whereas prospective purchasers just like a good, honest price. Why does enterprise storage pricing have to be so complicated and can’t we just have pricing online? Chris and Martin chat to George Crump from StorONE about strategies for pricing from both the customer and vendor perspective.

Vendors mentioned in this podcast: StorONE, IBM, NetApp, Microsoft Azure, Pure Storage, Dell EMC.

Find more about StorONE at https://www.storone.com.

Elapsed Time: 00:35:05

Timeline

* 00:00:00 – Intros * 00:02:00 – Enterprise storage pricing is very ambiguous * 00:03:20 – Hotel room rates and storage pricing are made up * 00:04:50 – How do vendors set list price? * 00:05:40 – Sales cycles drive pricing on quarterly & annual basis * 00:07:00 – Do “good” customers get a better deal? * 00:09:30 – Is being transparent on price a weakness? * 00:11:30 – Storage value is in the software * 00:12:00 – Software pricing is generally done by capacity * 00:14:00 – Subscriptions allow vendors to add value through new features * 00:18:00 – Should features be inclusive or itemised? * 00:20:10 – Cloud is training people to adjust to itemised pricing * 00:27:00 – So what does pricing in the cloud look like? * 00:29:53 – Customers need to be ready to buy * 00:34:00 – Wrap Up

Related Podcasts & Blogs

* #166 – Infinidat Elastic Storage Pricing With Eran Brown* #164 – Introduction to StorONE S1: AFAn* #130 – Making Money in the Storage Business

Copyright (c) 2016-2020 Storage Unpacked. No reproduction or re-use without permission. Podcast episode #olcc.

This week, Chris and Martin talk to Shawn Meyers, Field CTO at Tintri about SQL Integrated Storage. The Tintri VMstore platform originally provided the ability to apply policy-based management to virtual machines on shared storage. This capability has now been extended to databases, in particular Microsoft SQL Server.

SQL Integrated Storage (or SIS) works by exposing an SMB share from the VMstore platform onto which database files are stored. VMstore is provided awareness of the SQL database structure and can therefore manage the QoS and data management requirements of individual files that comprise a single database.

What’s interesting about this technology is the promise for further abstraction of application data away from virtual machines. The files of a database can now easily be mapped to a container or physical machine as required. Data can be cloned and replicated very easily. The technology can also be extended to other structured data solutions.

Learn more about SIS here – https://tintri.com/solutions/sql-integrated-storage/

Elapsed Time: 00:34:03

Timeline

* 00:00:00 – Intros * 00:02:00 – Tintri is now part of DDN * 00:03:00 – What is VMstore? * 00:05:20 – VMstore applies VM-level QoS * 00:07:00 – vSAN doesn’t do VM-based QoS * 00:09:00 – SQL Integrated Storage offers database-level QoS * 00:10:45 – How is SIS managing data? SMB * 00:12:50 – What databases are supported?  MS SQL Server * 00:15:00 – NFS faster than Fibre Channel?  Never! * 00:16:51 – What does SIS enable? * 00:19:30 – Separation of data and code * 00:20:30 – SIS is a data mobility solution * 00:24:00 – SIS could provide minute-based SQL snapshots * 00:25:43 – What is the customer feedback so far? * 00:27:40 – SIS could enable quick post-replication checks * 00:29:15 – Obligatory mainframe reference! * 00:31:09 – Call to action * 00:32:30 – Wrap Up

Related Podcasts & Blogs

* Private Cloud Storage and The Tintri Platform* #168 – Storage Unicorns* #110 – Storage Vendor Consolidations & Acquisitions* Whatever Happened to VVOLs?

Copyright (c) 2016-2020 Storage Unpacked. No reproduction or re-use without permission. Podcast episode #hlo4.

This week the team double down on the topics of in-memory computing and persistent memory. Chris and Martin talk to Charles Fan, CEO at MemVerge about Big Memory and using persistent memory technology (specifically Optane) to supplement system DRAM.

DRAM is expensive and as capacities scale linearly, the price of memory increases exponentially. Systems are limited by maximum addressable memory per socket. Persistent Memory in the form of Intel Optane provides the capability to massively increase the virtual memory footprint, using a combination of DRAM and Optane DIMMs. How is that memory managed – this is where MemVerge comes in.

MemVerge uses a feature called LD_PRELOAD to intercept memory management calls and direct requests to the MemVerge software. This means MemVerge can manage applications sitting across real DRAM and Optane. It also means the company can introduce neat tricks like memory snapshots and application rollbacks. This is a great conversation and serves as a great introduction in what we can expect of in-memory computing in the future. Find more on MemVerge at https://www.memverge.com/

Elapsed Time: 00:39:10

Timeline

* 00:00:00 – Intros * 00:02:15 – Real-time computing is driving in-memory computing * 00:03:00 – DRAM is not scalable and expensive * 00:05:30 – Applications can hit a cliff edge when in-memory capacity is exhausted * 00:07:45 – Persistent Memory (like Optane) is byte-addressable * 00:09:30 – Memory capacities scale linearly and costs increase exponentially * 00:11:00 – How are systems going to balance DRAM and PMEM access in a single server? * 00:13:00 – Application rewrites are not popular… * 00:15:30 – How is PMEM identified by the O/S? * 00:16:45 – MemVerge implements “software-defined memory” – LD_PRELOAD * 00:19:30 – How is MemVerge technology implemented? * 00:21:30 – What algorithms are used to manage DRAM and PMEM? * 00:24:45 – Test figures show performance can be better than DRAM * 00:28:00 – How do you use the persistence of PMEM effectively? * 00:30:00 – MemVerge offers application checkpointing – memory snapshots * 00:33:00 – Snapshots in memory and on disk are co-ordinated * 00:36:45 – Persistent Memory will be a game-changer for in-memory computing * 00:38:00 – Wrap Up

Related Podcasts & Blogs

* #169 – In-Memory Computing and Apache Ignite* #36 – The Persistence of Memory with Rob Peglar* #159 – Introduction to MRAM with Joe O’Hare from Everspin* Persistent Memory in the Data Centre* What are Storage Class Memory (SCM) and Persistent Memory (PM)?

Copyright (c) 2016-2020 Storage Unpacked. No reproduction or re-use without permission. Podcast episode #747k.