Author Archives: cking

Intel Optane – Bringing 3D XPoint Memory to Market

By Charles King, Pund-IT, Inc.

Intel’s new Optane SSD DC P4800X Series offerings are an interesting example of how complex technologies come to market. Interesting in one sense because of the potential value Intel’s new SSD solutions offer business customers and data center applications and workloads.

But also interesting is the somewhat circuitous road that Intel and its partner Micron traversed for the nearly two years it took to arrive here. In July 2015, the pair announced plans to develop 3D XPoint (pronounced: three dee cross point), a non-volatile memory (NVM) technology that would result in solutions that were “1,000X faster than flash with up to 1,000X flash’s endurance and 10X denser than conventional memory.”

How did they do? Like many or even most new technologies, the Optane SSD DC P4800X offerings don’t quite reach their developers’ original targets. However, they do deliver significant improvements over past solutions and benefits that should attract enterprise customers. As a result, Optane should help Intel achieve its larger goal: to fundamentally transform storage and memory architectures.

Memory’s promise and problems

Memory technologies have essentially transformed data center solutions and markets due to two issues: 1) falling prices of conventional DRAM (volatile memory) that make it more financially feasible to use it to speed applications and workloads, including in-memory databases, and 2) substantial performance boosts that NAND flash solid state drives (SSDs) deliver to data storage operations and processes.

However, the promise of commercial memory technologies is offset by significant problems. Yes, DRAM is cheaper than ever before, but it still costs more than NAND. Plus, the sensitivity volatile memory (where a loss of power causes data to be lost) makes it less than perfect for business-critical applications. In NAND’s case, low prices are offset by latency, endurance and density issues. Add in a crowded marketplace that makes it difficult for NAND vendors to be competitively profitable, and you can see why alternatives are attractive.

What could a vendor achieve with new products that addressed the shortcomings of NAND, allowing it to enhance both memory and storage applications? That’s what Intel and Micro were after with 3D XPoint memory, and what Intel says it delivered with the Optane SSD DC P4800X Series.

Taking off the wrapper: Optane SSD DC P4800X

What does the new solution offer out of the box? According to Intel, Optane delivers “an industry-leading combination of high throughput, low latency, high quality of service and high endurance” that allows the new solutions to be “deployed as blazing fast storage or caching tier.”

How does that claim look in real world terms? Though Intel compared the new 375GB Optane DC P4800X drives against its 400GB DC P3700 solutions, the performance numbers were notable and even startling.

In low queue depth (where most business applications generate their storage workloads) Optane SSDs performed 5X to 8X better than the DC P3700. But Optane’s random read IOPS performance is near 10X better than what DC P3700 delivers, and is about 4X the performance offered by competing NVM drives from Samsung and Seagate.

In practical terms, Optane has a significant lead in important metrics. Plus, the new solutions also offer better endurance than competing NAND SSDs. How much better and why is that important?

Let’s take the second point first. Though SSD failure rates are comparable to or better than mechanical HDDs, the materials in NAND flash SSDs degrade as they are used so their accuracy eventually falters. As a result, NAND flash SSDs’ working lifetimes are dictated by the number of write operations (known as program/erase (P/E) cycles) they can endure. SSD endurance metrics mean to produce lifespan numbers to guide end users’ purchasing decisions.

So what does that mean for Intel Optane? The company says DC P4800X products can support as many as 30 drive writes per day (DWPD) compared to the 0.5 to 10 DWPD offered by competing SSDs. That also enables the DC P4800X to endure up to 2.8x more total bytes written to it over five years than a DC P3700 SSD (which supports 17 DWPD).

As a result, the DC P4800X appears to fulfill Intel’s claims about its value as a notably fast and durable storage solution, but how about its applicability in memory caching scenarios? That’s where the company’s new Memory Drive Technology comes in, which Intel says “transparently integrates the drive into the memory subsystem and presents the SSD as DRAM to the OS and applications.” Thus, organizations can use Optane to create capacious, cost effective “pools” of shared memory for enhancing business application and workload performance.

Intel’s new Optane solutions are currently available to customers via an early ship program. MSRP for the pair is: Intel Optane SSD DC P4800X, 375 GB AIC: $1520, and Intel Optane SSD DC P4800X with Intel Memory Drive Technology: $1951 Additional capacities and form factors will be available in the second half of 2017. Intel noted that no changes to the OS are required by Optane but that the new SSDs are only supported by Intel Xeon processors.

Final analysis

That last point—that Optane is only supported by Intel Xeon processors—is both unsurprising and key to Intel’s larger strategy. Timing-wise, the company is planning commercial volumes of Optane in 2H 2017 which parallels the reported availability for its newest Xeon processors and server platform. Adding Optane to those next gen solutions should be a powerful attractant for system vendors and their customers. But the availability of unique Intel options for enhancing storage and memory would also steal some of the thunder of competing offerings from NVIDIA, AMD and ARM vendors whose data center ambitions are on the rise.

So what’s the key takeaway here? In essence, with its Optane SSD DC P4800X solutions, Intel has created a robust new technology that is appropriate for both fast storage and cost-effective pooled memory scenarios. The details the company shared, along with the comparisons to its own DC3700 SSDs, suggest that Optane’s price/performance qualities should attract enterprise buyers’ interest and enthusiasm.

Is there anything that might derail these ambitious plans? That depends on whether Intel is successful in ramping Optane production to meet commercial demand, a process that can be replete with unpleasant events and unplanned delays. Leaving aside those points, the company deserves applause for what it has achieved to date. Absent unforeseen events, IT markets should soon see robust, competitively-priced Intel Optane solutions for the data center.

© 2017 Pund-IT, Inc. All rights reserved.

Lenovo’s DX8200D – Strategic Partnering to Make SDS Simple and Affordable

By Charles King, Pund-IT, Inc.  March 22, 2017

Lenovo’s recent launch of its new DX8200D software-defined storage (SDS) appliance with SANsymphony from DataCore Software is worth considering on its own but also offers an opportunity to see how the company’s broader Data Center Group (DCG) partner strategy is evolving.

In short, the DX8200D is designed to harness the capabilities of existing SAN arrays by pooling and virtualizing internal, direct-attached and external heterogeneous storage infrastructures via its SDS capabilities. Lenovo describes the DX8200D as a pre-integrated turnkey appliance that promises to greatly simplify deployment and reduce management expenses while providing a single point of support for SDS applications and workloads.

Lenovo noted that the DX8200D’s centralized interface supports data protection, replication, de-duplication, compression, and other enterprise storage capabilities at a much lower price point than traditional SAN arrays. That enables Lenovo customers to use the new solution to scale storage infrastructures, increase the life and ease the replacement of their existing storage assets and simplify the transition to SDS.

Built on Lenovo’s System x3650 M5 server, the company noted that the DX8200D is designed to deliver the highest levels of response times, availability and utilization. In addition, Lenovo will work with DataCore to pre-test appliances to reduce deployment risks and increase time-to-value.

The DX8200D is designed for medium to large enterprises with existing heterogeneous storage infrastructures, and Lenovo believes customers in education, healthcare, industrial and public sector organizations will find the solution particularly valuable. Along with its SDS capabilities, the DX8200D should be optimal for accelerating transactional databases, analytics, ERP and CRM workloads, and for supporting business continuity and disaster recovery for business-critical applications.

Final analysis

To begin, the DX8200D looks fully capable of supporting all the capabilities and delivering the full IT and business value that Lenovo claims. The new solution is based on the company’s well-known and proven System x3650 M5, a rack server introduced early in 2015. Able to support up to 1.5TB of memory and a numerous HDD and SSD options with its 26 drive bays, the x3650 M5 is an appropriate choice for a wide variety of storage and business use cases, including server-based storage virtualization.

How does the new solution reflect on Lenovo’s partner strategy? As I’ve noted in previous commentaries, the company has charted an intriguing course that differs somewhat from its competitors. While all system vendors engage in strategic partnerships, quite often those relationships result in acquisitions designed to effect competitive dominance. HPE’s recent purchase of SimpliVity is a good example of that point.

In contrast, Lenovo has pursued and developed relationships with numerous like-minded partners, particularly infrastructure software players, including DataCore, Cloudian and Nexenta for SDS, and Nutanix and SimpliVity (at least until HPE took it off the board) for hyperconverged systems. As a result, Lenovo has been able to increase the variety and value of its own portfolio, and also maximize the options customers can choose to meet their business and compute requirements.

The evolution of individual partnerships is also intriguing, as a closer look at Lenovo and DataCore will reveal. At VMworld in September 2015, DataCore demonstrated its then-new parallel I/O software, certifying record-breaking performance in the SPC-1 benchmark with a Lenovo System x3650 M5. The following January, the two companies launched three new solutions based on the x3650 M5; an all-flash array and appliances for application acceleration and storage virtualization.

In February 2016, Lenovo and DataCore announced plans to develop a single stack, adaptive infrastructure solution for software-defined storage leveraging System x3650 M5 servers and SANsymphony software. To underscore the value of the effort, DataCore’s CEO said the company would invest 20 percent of its total revenues in the R&D effort, focusing on parallel I/O processing, virtual I/O, universal data services, and simplicity and automation.

Lenovo’s DX8200D with DataCore’s SANsymphony is the result of those strategic efforts.

Overall, the DX8200D seems entirely capable of providing customers the full range of software-defined storage features, functions and value that Lenovo and DataCore envision. But Lenovo’s DX8200D also qualifies as a notable example of the innovation that can result from deeply considered and fully developed strategic collaborations.

© 2017 Pund-IT, Inc. All rights reserved.

Intel Acquires Mobileye – Paving the Way for Autonomous Driving

By Charles King, Pund-IT, Inc.  March 15, 2017

Intel jolted more than a few industry watchers this week with its planned $15.3B acquisition of Mobileye. That’s a tidy sum for a company that is mostly recognized for developing innovative Advanced Driver Assistance Systems (ADAS) that support passive alert and active safety systems capabilities for automobiles and trucks. Those technologies have led Mobileye into successful relationships with all Tier 1 automotive industry suppliers and 27 OEMs that are implementing the company’s solutions in the serial production of 313 car models.

But as Mobileye notes on its company site, active safety system features, like Adaptive Cruise Control (ACC), Lane Keeping Assist (LKA), Lane Centering (LC) and Traffic Jam Assist (TJA) “comprise the building blocks of semi/fully autonomous driving.” That point was given top line billing in Intel’s announcement concerning the deal and in letters that both companies’ senior executives (Intel’s CEO Brian Krzanich and Mobileye’s co-founders Amnon Shashua and Ziv Aviram) wrote their employees.

Autonomous driving is an active focus point for numerous other IT industry heavyweights, including Google, NVIDIA, Uber and Apple (reportedly), and with good reason. As Intel’s Krzanich noted, the estimated market for these solutions will be $70B by 2030, representing a rich target for tech vendors preoccupied with the Next Big Thing. Perhaps the larger questions, then, are what Intel and Mobileye together will bring to the party, and why the pair believes they have what it takes to beat the competition.

Let’s take a closer look at that.

Car to cloud to car

The largest strategic factor to consider regarding Intel and Mobileye is that autonomous driving is essentially an infrastructure play. That is, in autonomous driving scenarios Mobileye’s advanced onboard cameras and related software and other system components are heavily reliant on and gain significant additional value from the memory, storage, wireless networks and backend cloud data centers required for many autonomous functions and processes.

That’s partly due to an issue that Krzanich brought up in his letter to employees: An average autonomous car will create and utilize four terabytes of data for every day it is on the road—about the same amount of data created by 3,000 people. Or as Intel’s CEO eloquently noted, “Put just one million autonomous cars on the road and you have the data equivalent of half the world’s population.”

That data has to be collected, initially analyzed, transmitted to high performance cloud facilities for additional analysis and then returned to the vehicles that generated it, often in near-real time. If you think that’s complicated, add in the catastrophic, even fatal problems that could occur if systems glitch, networks misfire or clouds disperse. In any case, it’s easy to see why vendors pursuing autonomous driving need to be as serious about dependability as they are about innovation.

Intel gets that point in a very big way and has a history of developing innovative, trustworthy technologies that millions of individuals and organizations bet their lives on both professionally and literally. The company focused clear attention on the value that the innovative FPGAs, 3D Xpoint memory and 5G modems will bring to the strategic initiatives that Mobileye will drive as the core organization and leadership in Intel’s new Automated Driving Group (ADG).

But without Intel’s dependably proven Xeon processors and other foundational cloud and data center technologies, autonomous driving solutions would be stranded by the side of the road with the night coming down.

Building an industry standard stack

Those points also underscore the larger play regarding Mobile. Given the broad, enthusiastic adoption of its ADAS and other solutions among auto industry Tier 1 suppliers and OEMs, and Mobileye’s current place in over 300 car models, acquiring the company provides Intel a critical piece for creating an integrated, industry standard component stack for autonomous driving. That should open the door to numerous new business alliances and opportunities, but it also puts a thumb firmly in the eye of NVIDIA, Intel’s most ambitious challenger in this market.

It’s also worth noting that Intel’s broader integrative vision and strategy regarding Mobileye and autonomous driving is reminiscent of other past company initiatives. For example, beginning in the early 2000s Intel made significant strides in integrating wi-fi and WiMax wireless functionalities into its notebook-focused CPUs. That effort came to fruition in the Centrino-branded chips the company introduced in 2004 and then extended via a partnership with Nokia and the wireless roadmap it revealed at the 2006 Intel Developer Forum.

Centrino enabled Intel to establish a forward-looking leadership position in PC wireless chips, and allowed the company’s OEM partners to develop laptop systems that were thinner and lighter than ever before. But Centrino also displaced numerous suppliers’ plug-in PCMCIA and USB cards for wireless connectivity. In other words, Intel’s integrated, industry standard wireless stack helped the PC market evolve in a more orderly, sustainable manner than it had previously.

Final analysis

Whether integrating car-to-cloud-to-car functionalities via the Mobileye acquisition will have a similar effect on autonomous driving is impossible to say at this point. However, the deal reflects a strategic vision Intel has pursued and a road it has traveled to success many times in the past. Whatever destination the company’s new Intel Automated Driving Group pursues, Intel’s vision and its deal for Mobileye will fundamentally change the market and market dynamics for autonomous driving.

© 2017 Pund-IT, Inc. All rights reserved.

IBM Q: Evolving Quantum Computing and its Ecosystem

By Charles King, Pund-IT, Inc.  March 8, 2017

The IT industry is seldom good at describing how new products and entirely new classes of technology come into being. This is partly a matter of industry culture and partly due to vendors exerting control over their own brands and narratives. In other words, vendors prefer to be applauded for the brilliance and foresight that goes into successful products rather than acknowledge the blind alleys and rats’ nests that are a common part of the development process.

But now and again something comes along that is so thoroughly different than what has gone before and offers such clear opportunities for advancement that its evolution becomes part of common discourse and it “grows up in public.” Quantum computing, and the new IBM Q efforts are a good example of this. Continue reading

Dell IoT: Pushing the Edge of End-to-End Computing

By Charles King, Pund-IT, Inc.  March 1, 2017

There was a time when end-to-end computing – that is, delivering solutions from the desktop to the data center and everything in between – dominated system vendor thinking. That was the case in the 1990s when companies, including IBM, HP and Compaq all pursued end-to-end strategies, and even mainly data center-centric vendors that shunned desktop PCs, like Sun Microsystems, still offered signature workstation endpoints.

Things inevitably change, of course. HP bought Compaq and refocused its brand on budget PC lines. IBM sold its PC unit and then its x86 server business to Lenovo as part of a broad retreat from commodity technologies. Sun’s workstation business, along with the rest of the company, fell victim to increasingly robust Intel-based PCs and servers. Most recently, HP spun itself into two companies: one focused exclusively on data center offerings (HPE) and the other on PCs and printing/imaging products (HPQ).

But two other vendors have headed in a distinctly opposite direction. Continue reading

IBM Connect 2017: Nurturing and Augmenting Cognitive Collaboration

By Charles King, Pund-IT, Inc.  March 1, 2017

I’ve attended IBM’s annual partner and user conferences for as long as I’ve been an IT industry analyst. While I’ve seen more than a few shifts in the company’s emphasis and strategic intent around staging these events, few reflected the kind of fundamental realignment that I witnessed at the Connect 2017 conference last week in San Francisco.

Why is that the case? Because of the evolution of IBM’s collaboration and application business. Continue reading

Lenovo Delivers Enterprise Cloud for SAP HANA in China

By Charles King, Pund-IT, Inc.  February 22, 2017

Lenovo’s plan to develop and deliver a new “Enterprise Cloud Solution designed for SAP HANA” in China offers an interesting lesson in how a vendor can effectively parlay its strengths to pursue opportunities for its own and its partner’s benefit. In this case, Lenovo is exercising its muscular technological and regional leadership positions.

That’s well worth considering, but before that, what are the announcement’s core points? Continue reading