While HP splits up, married entrepreneurs build something together

It’s a Structure Show of two extremes this week, kicking off with a discussion of the huge decision to split HP into two separate companies. The move is interesting for many reasons, including the possibility that it opens the door for a merger between HP (the server company) and storage kingpin EMC. The suggestion has also been floated that HP will now purchase Rackspace to quickly expand the footprint of it cloud computing business.

There are such a lot of angles here, and Barb Darrow and I break down all — OK, some — of them.


This week’s show concludes with an interview with Ann and Bobby Johnson, the husband and wife at the back of a new analytics startup referred to as Interana. We covered the company earlier this week, but dive even deeper in the podcast, including with a discussion about why, for the Johnsons at least, it was once easier to start a business together than many might consider.

But the truly interesting a part of Interana is the technology and the vision, much of which stems from the work Bobby and third co-founder Lior Abraham did all the way through their tenures at Facebook. Interana is a custom-built engine for storing, analyzing and visualizing massive amount of event data, and it’s meant to be usable by large numbers of employees.

So listen up. You might just learn something about analytics, corporate divestitures and even … love.



Hosts: Barb Darrow and Derrick Harris

Download This Episode

Subscribe in iTunes

The Structure Show RSS Feed


Ericsson mind melds with Apcera; Talko takes on communications; and what’s up with Watson?

Launching open-source projects is easy; keeping them running well not so much

What on the earth will HP do with Eucalyptus? The company’s cloud chiefs give us a hint

The trials and tribulations of the API economy

Here’s why the democratization of big data truly truly will have to excite you. Yes, you.

Red Hat and Mirantis: The gloves are off

Once upon a time Red Hat and Mirantis got on very well. Just last year, Red Hat invested in and partnered with the OpenStack systems integrator, but then Mirantis released its own OpenStack distribution and buddied up with Canonical, and when Mirantis raised $100 million last month Red Hat was once most definitely not a participant.

Also, in June Red Hat bought French systems integrator eNovance to kinda do for it what Mirantis was once supposed to be doing. So, with all that activity in latest months, how are Red Hat and Mirantis getting on nowadays? Funny you ask.

“We took an investment in Mirantis under the pretense that was once going to be a consulting [partnership],” Red Hat tech chief Paul Cormier told me today at the OpenStack Summit in Paris. “We were on the lookout for a consulting partner and they determined to get into the product space. That’s their prerogative.”

“So we went out and got a better consulting partner, eNovance. They’ve got in reality interesting management technologies that we’re integrating into our products now.”

But hang on … Mirantis CEO Adrian Ionel (pictured above) has a very different take on what happened between the two companies.

“There is a contract on the books that had our software and product story as a part of the agreement with Red Hat, that Red Hat signed as a part of the investment,” Ionel told me. “We signed an alliance agreement at the time to integrate our product story with theirs. They didn’t honor it.”

Now, as you’ll recall, earlier this year there was once a bit of a kerfuffle around Red Hat’s unwillingness to fully make stronger non-Red-Hat OpenStack deployments on Red Hat Enterprise Linux (RHEL) – an arguable storm in a teacup that nonetheless painted a picture of the company as one that wants to achieve some lock-in in the supposedly anti-lock-in OpenStack game.

Bearing that in mind, back to Ionel: “Red Hat asked us to completely lock ourselves into their operating system as a condition for honoring the agreement.”

Was that in the contract? “No, because we were supposed to be free,” he replied. “So they changed their position from saying, ‘Let’s make Red Hat a first-class citizen on par with everybody else,’ and they wanted to change it to be the only choice. After the deal Red Hat came back, Paul specifically, and he said Red Hat is the only citizen, not a first-class citizen.”

Wowzer. So, with this much bad blood, does Mirantis want out of its tattered partnership with Red Hat, as was once reported? “We’re going to let it run and that’s it,” Ionel said. “We wish to work with Red Hat, we wish to work on Red Hat, but we will be able to’t make Red Hat the only one.”

Fast-growing DigitalOcean brings in MakerBot vet to head finance

Larry White, who helped build 3D printing pioneer MakerBot into a market leader, is now  VP of finance at DigitalOcean, a hot New York-based IaaS provider that claims TaskRabbit and Amazon —  yes, Amazon! — among its customers.

DigitalOcean’s challenges now are very similar to those MakerBot faced: how to grow from a small group of “brilliant entrepreneurs” into a big business, and how to construct the infrastructure and team to make that happen, White said in an interview.

The company targets newbie developers and hopes they’ll stick with its inexpensive products and services as they gain proficiency. Professional and enterprise developers don’t seem to be a key focus, “despite the fact that we won’t turn them away,” White said.

“Our products and services are up and running in a couple of minutes, even as with AWS you want to be a console guru,” he said, referring to the AWS Management Console. 

DigitalOcean prices aggressively, starting at $5 per month for one CPU core, 512MB of memory, 20GB SSD disk and 1TB of data transfer. “Look to these guys to see where the price war is going,” said 451 Research analyst Carl Brooks.

White if truth be told joined DigitalOcean in September, despite the fact that the news used to be not publicly announced then. At that time, head count used to be 30; it’s now up to 40. Earlier this month, DigitalOcean said it had rolled out 500,000 virtual servers running on 5,000 physical machines — that’s a drop in the bucket in comparison to Microsoft and Amazon, but it’s not nothing, either.

The biggest challenge for the company, which garnered $3.2 million in seed funding last August, is keeping up with demand. White cited revenue growth of 30 percent month over month and said DigitalOcean averages 800 new customer sign-ups per day.

Facebook’s answer to serving 700TB of graph search data is lots of SSDs

Facebook’s graph search feature requires finding and serving the right data fast, and from a database that currently houses more than a trillion posts and 700 terabytes of data overall. In a Thursday morning blog post, engineer Ashoat Tevosyan dove into one of the most challenges of building infrastructure that can handle these demands.

One decision stood out to me was that Facebook opted for solid-state drives to store among the data, saving only the most-steadily accessed stuff for RAM. This wasn’t a problem until Graph Search began including users’ posts, which drastically bumped up the size of the indexes it was dealing with. According to the post

“[S]toring more than 700 terabytes in RAM imposes a considerable amount of overhead, as it involves maintaining an index that is spread across many racks of machines. The performance cost of having these machines coordinate with each other drove the Unicorn [search infrastructure] team to look into new solutions for serving the posts index.”

SSDs have been a key part of Facebook’s growth strategy for a even as as an option for preserving the performance users require but saving on the high costs of storing data in RAM. In January, it unveiled a new all-flash server call Dragonstone for just such a purpose. In March, the company detailed a system it had built called McDipper, which is an SSD-based implementation of the popular memcached caching layer for RAM.

However, just because Facebook is the use of flash and SSDs a lot more incessantly, that doesn’t mean the company is all the time happy about. As VP of Engineering Jay Parikh told me throughout the company’s Open Compute Summit earlier this year, if hard disks are like minivans and current flash drives are like Ferraris, Facebook is on the lookout for the Toyota Prius of storage that delivers the right balance of speed, efficiency and cost.

Check out the rest of Tevosyan’s post for more details on building the Graph Search indexes in HBase, harvesting user data from its MySQL cluster without throttling its performance and the use of Facebook’s new Wormhole technology to update the Graph Search index as changes happen to the MySQL data.

How the Industrial Internet is bridging the Rust Belt and Silicon Valley

Until recently, Jon Sobel and Sight Machine had a perception problem. “Someone once said we were too Michigan for Silicon Valley and too Silicon Valley for Michigan,” Sobel joked right through a recent phone call.

He can laugh about it now that companies like GE have made “Industrial Internet” a household term, but it’s still a little true. You see, Sight Machine, which is co-headquartered in San Francisco and Ann Arbor, Mich., is attempting to sell big manufacturing plants — from the ones building cars in nearby Detroit to those producing packaged foods — on the idea that they need to upgrade their computerized vision systems for quality control to Sight Machine’s cloud-based platform. It might end up being a transformative experience, but try telling that to prospective customers.

Whether they’re selling computer vision platforms or highly instrumented forklifts, companies trying to push the current generation’s hot technologies on users who don’t read the tech business press can end up playing a distorted version of the $100,000 Pyramid: The two contestants both want to get to the same place, but the contestant doing the selling can’t just throw a bunch of words out and hope they mean something to the buyer.

The trick to winning is establishing a strong sense of consider. That means knowing the technology they’re selling cold, but knowing the customers even better.

Translating tech speak into factory speak

“If you say I’m from Silicon Valley [and] I’m here to help,” Sobel said, “they’ll throw you out.”

It’s not because the manufacturing sector is afraid of technology (it spends hundreds of billions on it every year) or biased against Silicon Valley geeks who have never gotten their fingers dirty actually building something. It’s because the manufacturing sector can’t find the money for to replace its tried-and-true systems — then again flawed — with promises of the next big thing.

These are production systems that move fast. Sometimes, they’re building very expensive machinery or components for very expensive machinery. You can’t come to them with a beta system and expect to just apologize if something goes wrong, or worse yet, come in pitching a half-baked idea that’s great in theory but has never really been tested in the field.

Vendors also can’t come in bandying buzzwords about. That might work in the C suite or in Silicon Valley, where some terms get people so excited they border on pornographic, but not so much on a factory floor in middle America.

“People on the plant floor don’t really think about buying software, they’re used to buying equipment,” Sobel explained. “… We don’t use words like cloud or big data, it doesn’t mean anything to them. … They say, ‘Does it work?’”

Although he’s not coming from the Bay Area, Jim Gaskell, of New Bremen, Ohio-based forklift manufacturer Crown Equipment, can vouch. His business has evolved from one that sells big forklifts — or lift trucks, as they’re known in the industry — to companies running large warehouses and distribution centers into one that sells big forklifts and a cloud platform to manage data from all the sensors Crown is installing on them.

“They don’t buy it because it’s a neat gadget that makes them feel better,” Gaskell said. “They buy it because they’re trying to accomplish a goal.”

But it is high technology …

In the case of Sight Machine, the technology Sobel is attempting to sell is a fusion just those buzzwords he warned against lobbing at customers — cloud computing, big data, computer vision. Speaking to me, a tech journalist, Sobel went so far as to describe Sight Machine’s platform as “Ruby on Rails for vision.” It’s a software framework for real-time computer vision, but it’s a lot more than just vision.

After the system identifies whether units moving across the assembly pass or fail inspection, it stores and the images in the cloud and generates data based on the units’ dimensions, time stamps and other points. Users can then access that data from their desktops, analyzing everything from time-series data on the whole line to how many standard deviations an individual unit’s dimensions were from the ideal. It doesn’t matter what algorithms are programmed into the camera’s processor — or even supposing the camera has an embedded processor — because all the work is done elsewhere.


It’s the kind of easy and detailed data analysis that somebody covering cloud computing and big data for a living might expect. Sobel said his company is even engaged in discussions with companies like Microsoft and Google about the cutting edge in image-recognition algorithms. He co-founded Sight Machine together with Slashdot Co-founder Nathan Oostendorp, who, it turns out, is also a skilled industrial engineer.

But it’s likely none of this matters much to the guys writing the checks that pay Sight Machine’s bills. To some of them, the idea of using a cloud platform might be an utterly new concept. They’re just trying to escape from the Stone Age — some customers were previously keeping quality control data in three-ring binders — without fear that upgrading to 21st-century technology will render their operations, then again antiquated, ineffective.

And “that’s the norm, not the exception,” Sobel noted, because many computer vision systems in place today are really just designed to identify abnormalities and make pass-fail calls on the fly. Once that call is made, the data is ceaselessly erased by the camera’s embedded processor. That’s why some factories turn to those binders or other jerry-rigged systems that allow them to keep some records, even supposing accessing or analyzing them is nigh impossible.

“There’s nothing revolutionary about [what we’re doing] except that it just hasn’t been done in manufacturing,” Sobel acknowledged.

Teaching moments in the world of forklifts

Crown Equipment has experienced the same sort of situation, even though its customers are ceaselessly coming from a place of even less emphasis on data analysis. Five years ago, Gaskell said, people used to ask Crown Equipment about advice on managing their fleets of forklifts and they wouldn’t even know how many they had. They’d ask, for example, if $1 million was a reasonable cost to manage their forklifts for a year without any real context around that number.

So Crown’s first move into the data space was just to help customers get a handle on their fleets: Figure out how many they have (“A lot of customers end up having too many trucks,” Gaskell said) and the degree to which they’re using them (“They think they’re using the truck 24-7 … they find out they’re using them 4 to 5 hours a day,” Gaskell said), and then figure out a plan for making that more efficient.

Now, the company is managing sensor data from its lift trucks so companies can figure how they’re being operated and ensure they last, in part by changing drivers’ behavior. The sensors are measuring things like speed, whether the machine is sitting idle, force of have an effect on and even whether the wheels stopped moving (this, it turns out, can be a sign of whether someone ran over a bump in the floor or crashed into something). “Once the operator realizes he can’t fool you … that’s gonna start changing operator behavior,” Gaskell said.

But in order to let customers actually benefit from all this data, Crown has had to help them evolve their cultures into ones that value data over words. He estimates 9 of 10 customers would probably not take full advantage of Crown’s cloud platform, called InfoLink, if left to their own devices. So the company educates them not only on how to use it, but also on the fact that the data is telling the truth even supposing, for example, a driver said he just ran over a twig.

“It started off about selling hardware,” Gaskell said, looking back on Crown’s business. “What it’s turned into is providing a service.” And as a service provider, he added, “There’s a different quality level … that I have to provide for my customer now.”

A question of trust

Back in San Francisco, Sight Machine’s Sobel is attempting to develop with its computer vision customers the type of relationships an established company Crown has with its forklift customers. And he thinks he might have it figured out. “Everybody thought … that it would be impossible to sell to manufacturing clients,” he said. “I have found … that it’s incredibly straightforward.”

It all boils down to establishing consider, which means being up front about what the technology can do for customers today, as well as what it can’t do. There are no exaggerations or promises of amazing capabilities coming down the pike. “That,” Sobel said, “is the one thing that’s fatal with these guys.”

Cisco, Dell, HP, IBM and EMC have most to lose in China post NSA-gate: Report

We all know China is a large — or potentially huge — market for U.S. tech goods. We also know that geopolitical realities have restrained trade with China — which is why Huawei has very little presence in the U.S. and also why Cisco has seen its market share in China decline prior to now few years.

Revelations that the U.S. National Security Agency and its “Five Eyes” analogs in the U.K., New Zealand, Australia and Canada now pose additional hurdles to these U.S. suppliers as they are attempting to sell in China. According to a new research note by Sanford Bernstein’s hardware and software analysts:

“While spying has occurred across many companies, governments and corporations, we imagine U.S. technology companies face the most revenue risk in China by a wide margin, followed by Brazil and other emerging markets.”

The degree of risk faced by U.S. providers in China depends on how much domestic competition they face. For that reason the researchers see enterprise software companies as quite secure because there are fewer domestic offerings to compete with them in China — no less than for now. So I guess that leaves companies like Microsoft and Oracle in the catbird seat (also, for now). Microsoft, by the way, is offering Azure and Office 365 products and services in a preview by way of a partnership with Via21net in China. But, networking hardware vendors including Cisco face considerably more risk because China steers domestic businesses to Huawei. Cisco has pushed the U.S. government to keep Huawei out and it’s clear that effort isn’t lost on China or on Huawei.  (Cisco insists that it has not “lobbied” the U.S. on this, but tomato-tomahto.)

The conclusion? Vendors most at risk are:

“Cisco; Dell; HP (particularly its commercial PC and enterprise businesses); IBM (particularly its hardware business, which accounts for ~15% of total company revenues); and EMC. All of these businesses are transactional and play in markets with Chinese and/or other competitors that could act as substitutes. EMC’s business is transactional and has competitive and substitution risk, but its China exposure is quite low.”

Apache Mahout, Hadoop’s original machine learning project, is moving on from MapReduce

Apache Mahout, a machine learning library for Hadoop since 2009, is joining the exodus away from MapReduce. The project’s community has made up our minds to rework Mahout to give a boost to the increasingly popular Apache Spark in-memory data-processing framework, as well as the H2O engine for running machine learning and mathematical workloads at scale.

While data processing in Hadoop has traditionally been done the use of MapReduce, the batch-oriented framework has fallen out of vogue as users began demanding lower-latency processing for certain types of workloads — such as machine learning. However, nobody actually wants to abandon Hadoop entirely because it’s still great for storing lots of data and many still use MapReduce for most of their workloads. Spark, which was developed at the University of California, Berkeley, has stepped in to fill that void in a growing number of cases where speed and ease of programming actually matter.

H2O was developed one after the other by a startup called 0xdata (pronounced hexadata), even supposing it’s also available as open source software. It’s an in-memory data engine specifically designed for running more than a few types of types of statisical computations — including deep learning models — on data stored in the Hadoop Distributed File System.

“[H2O] looks like a actually good technology layer to drive a large number of what Mahout’s been missing and remove the artificial constraints that have been in Mahout’s way,” Ted Dunning, project management committee member for Apache Mahout and the chief application architect at Hadoop software vendor MapR, told me. “A combination of H20 and Spark could actually be something,” he added.

SriSatish Ambati, the founder and CEO of 0xdata, noted that the data science community isn’t married to one computational framework over another as long as they get the job done. It’s the higher-level stuff, including how people program models, that actually matters. H2o natively supports the R programming language, for example, which is moderately popular and would be a new capability for Mahout, Dunning said.

One could argue that the Mahout community had to embrace Spark, at least, if it wanted to remain relevant. Already, Cloudera is working on its Oryx machine learning framework that was designed in order to overcome Mahout’s shortcoming and will be ported to Spark sooner or later. The Spark community itself is also working on a set of machine learning libraries called MLlib.

New Relic acquires Ducksboard and rolls out a bunch of new data products

New Relic, the application performance monitoring startup, plans to announce Wednesday at its FutureStack conference that it has acquired Barcelona-based startup Ducksboard and is unveiling a raft of new products. Financial terms of the acquisition were not disclosed.

Ducksboard is an eight-person company that can hook into different cloud services and products like GitHub, Salesforce.com and Twitter, consolidate their data and convert it into a dashboard that a user can configure. With Ducksboard, New Relic now has a Barcelona office and access to the startup’s technology that allows it to integrate with other data stores, said New Relic CEO Lew Cirne (pictured above).

Regarding its own lineup of services and products, New Relic is rolling out the ability for business users to create their own applications to analyze and interpret the data New Relic collects.

For example, a customer beef up team member will have to be capable of easily build an application based on the organization’s gathered data that could help the remainder of the team discover why customers are is also experiencing particular problems, explained Cirne. This new feature will have to be to be had in 2015.

“In the past, customers would send half-a-dozen developers to build an app,” said Cirne in regard to how users with limited development experience can craft helpful internal applications the use of the New Relic interface.

The company may be introducing a couple new features for developers. New Relic Browser helps coders better consider why their JavaScript applications is also having performance issues. New Relic Synthetics is a test automation tool that can spot problems like a malfunctioning content delivery network (CDN) that can be hampering an application’s speed.

New Relic Browser is to be had to the public as of Wednesday and New Relic Synthetics will be released later in the quarter.

Sponsored post: Introducing the Infinera Cloud Xpress

Whether it’s used for search, social networking or enterprise business applications, the cloud, a distributed computing model built upon multi-server and multi-data center infrastructure, is becoming ubiquitous. In fact, enterprises find immediate value in the cloud as it’s three times more economical to rent a cloud service somewhat than to own a server. Demand for these applications has led to exponential growth of the cloud, which in turn is dramatically transforming IT and network architectures. The rise of compute-and-storage virtualization has been the technology driver to make the cloud viable, but the network is the critical glue that in the long run makes a cloud a cloud.

Cloud growth creates tremendous demand for bandwidth to interconnect the data centers. Interconnecting these sites requires a transport network, which is a fundamental element of any cloud network. Providers need the ability to hastily scale their infrastructure to meet the bandwidth demands lately and accommodate future growth.

The Infinera Cloud Xpress is a purpose-built, optimized platform circle of relatives for point-to-point hyper-scale bandwidth interconnect application across regional, metro and campus environments. It is the first optical platform to deliver 1Tb/s of input and output capacity in just two rack units in a data center rack-and-stack form. The industry’s highest density solution in one-third the space of its competitors, it uses 50% less power than the current market leader, and will also be provisioned in a simple three step process in only some minutes.

To learn more about Cloud Xpress, read our brochure or consult with www.infinera.com/go/cloud.

“Coopetition” reigns supreme in Microsoft-Dropbox alliance

Dropbox and Microsoft both offer free or near-free storage to customers. But now they’re making it easier for Microsoft Office users — especially those who use Office for iPad — to keep the usage of Dropbox storage, which might seem counterintuitive since Microsoft pushes its own OneDrive as its repository of choice for Office users.

In a partnership reported by The Verge Tuesday, Kirk Koenigsbauer of the Microsoft Office Engineering team said that Dropbox get right of entry to was once a key wishlist item for iPad for Office users. Accordingly, the Dropbox icon will show up next to Microsoft’s OneDrive icon at the iPad for Office screen. The deal appears targeted at mobile Office users, not the more traditional, desk-bound knowledge workers who generally tend to make use of Word and Excel in their offices.

In return, Dropbox will encourage its users to turn to Microsoft Office applications to edit and create their documents in the first place. Microsoft has some other alliance with Box to ease co-existence of Office 365 and Box cloud storage and file sync software.

Here is the Microsoft blog post outlining the plan, and here’s Dropbox’s take.

Microsoft, under new CEO Satya Nadella, is steering a tricky course between pushing its own applications against rival ISVs and wooing those ISVs over to make use of Microsoft Azure and other resources.

Dropbox and Box, either one of which began out as cloud storage and file sync-and-share companies that work with everyone’s applications, are both seeking to broaden their reach and turn their products into more of a “platform” for third parties. That’s because nearly everyone, from Apple to Google to Microsoft, now has their own cloud storage-and-sync story to sell.