Wednesday, May 18th, 2016

Time	Event
5:50p	This IoT Startup Wants to Break Down Data Center Silos If the Internet of Things is a network of devices that generate data and share it over the network, than there is nothing new about the concept, even if the term is new. That’s especially true for data centers, whose operations teams have for many years relied on connected temperature and humidity sensors, networked PDUs, or power strips to manage their facilities. In Scott Noteboom’s view, however, that network is the “legacy Internet of Things,” which is different from the Internet of Things that’s being built today. Noteboom is a founder and CEO of LitBit, one of the companies doing the building. Electrical and mechanical systems in data centers are a perfect example of legacy IoT, he says, operating in silos, isolated from the IT systems they support. That isolation is the decades-old legacy, used to this day as the only method of securing these critical systems from intrusion. LitBit is targeting the data center market, and its initial beta customers are data center operators, but Noteboom’s ambitions extend far beyond the data center. The three-year-old San Jose, California-based startup has created a platform that aims to enable any networked device to generate data and share it in encrypted form, while giving users the tools to define what types of data they need and what they want to do with it. VCs Are Excited about IoT Startups Recently, LitBit raised $7 million in a second round of venture capital funding, following a $10.5 million round in 2013. That’s something the company probably wouldn’t have been able to do, had it defined itself as a data center infrastructure management startup. Scott Noteboom, founder and CEO, LitBit (Photo: Yevgeniy Sverdlik) IoT is a hot space, and there’s a lot more VC money available to companies in that space than in data center facilities. “Generally, investors don’t get excited about [physical data center management],” Noteboom, whose resume includes top infrastructure roles at Yahoo and Apple, says. “There isn’t a whole lot of startups that deal with that space.” Data center facilities management, generally, is not an innovative space, he explains. It’s a space where people largely do the same thing for the duration of their career, and if 25 years of experience doing the same thing is of value in an industry, that means “it doesn’t change too much.” “If you’re looking to serve anything on the machine side of the world that is not completely disruptive, it’s probably pretty challenging to raise venture capital.” VCs are also excited about IoT because it’s on the CIO’s radar, Noteboom explains. CIOs see it as a new challenge, with new budgets and new strategies. Connect Everything LitBit’s goal can be summed up as enabling more devices to communicate with each other and with people. Its platform, RhythmOS, is what aggregates and translates data generated by devices that use a variety of protocols, be they Modbus and BACnet, the widely used protocols in building management and automation systems, or SNMP, the protocol used for managing networked IP devices. Device manufacturers can contribute their own software for translating their protocols to the platform, and users can build their own applications on top of it, although LitBit does build and sell applications of its own. At the core of the platform is iota, a LitBit-led open source project that’s currently in incubation phase with the Apache Software Foundation. Incubation is the first step an open source project takes in the process of becoming an official Apache project. The software LitBit has built on top of RhythmOS so far has two general goals: providing users real-time data of their choice and analyzing a collective, central pool of operational data to help users manage infrastructure more effectively, be it infrastructure in a data center, a manufacturing plant, or an office building. Image: LitBit A colocation data center provider, for example, can easily share with a customer their power usage, or temperature and humidity in their colocation cage. They can set the software up to alert a user when they are approaching 80 percent utilization of their available power (which is not recommended) while at the same time alerting a sales rep who can contact the user with options to expand their deployment. Those are just two of many examples. The key is that the user defines the data sources they want to track, and those sources can be anything, from temperature to sound. A cooling unit may make a particular sound when it’s running low on refrigerant, for example, and if the system identifies the sound as abnormal, it may be set up to alert the operator. Combining Human and Machine Knowledge According to Noteboom, LitBit’s software uses machine learning techniques to gain insight from the data it collects, but it’s a different take on machine learning from machine learning in the typical sense, where a system is trained from scratch. “If you’re that data center facilities person who has 25 years of experience, and there are 10 of them in the data center, that collective 250 years’ worth of experience is really valuable and can make a machine a lot smarter,” he says. Related: Google Uses Machine Learning to Boost Data Center Efficiency If the system detects an unusual grinding noise in a data center, for instance, it will take a few seconds for an experienced data center manager to identify it as a broken bearing in a failed power supply. In this example, the system would alert the engineers of an unusual noise it is not familiar with, and one of the engineers would identify the problem. The next time the system detects a similar noise, it will know what it means. It’s human-assisted machine learning, or “a human knowledge base that applies AI and machine learning on top of that,” Noteboom explains. He expects this operational knowledge base to grow over time and be available to all LitBit users. Customers will have the option to anonymously share this data with a central data repository. The more users share their data over time, the more intelligent the system becomes. This data will be used not just for predictive operational intelligence, but to orchestrate infrastructure to fine-tune its performance. Facilities systems will be managed based on data collected from IT and facilities systems across many data centers over time. The silo will go away, including the silo within the four walls of a single data center. (Comment on this)
5:56p	Global Warming of Data Eric Bassier is Senior Director of Datacenter Solutions at Quantum. It already reached 90 degrees in Seattle this year. In April. I’m not complaining – yet – but I’m definitely a believer that global warming is happening and that we need to make some changes to address it. But this article isn’t about climate change – it’s about data. Specifically, it’s about the growth of unstructured data and the gloomy fate ahead if we continue to deny the problem and ignore the warning signs. Sound familiar? It’s hard to argue with the evidence of unstructured data growth. Estimates and studies vary, but the general consensus is that there will be 40-50 zettabytes of data by the year 2020, and 80-90 percent of that will be unstructured. What’s Driving Unstructured Data Growth? Data growth comes from many places. Of course there are sources like 4K HD movies and TV shows, and movies, pictures, and images that all of us take on our smartphones every day, but unstructured data growth is much broader than that. There are also vast amounts of data generated everyday by machines and sensors across a wide variety of data-driven industries like research, engineering and design, financial services, geospatial exploration, healthcare, and more. Video surveillance alone is creating almost an exabyte of unstructured data every day as camera resolutions and retention times have increased. These diverse datasets share some common characteristics. Typically, they are: Comprised of large file sizes; Un-compressible – i.e., techniques like deduplication are not effective at reducing the data; Valuable to the company, department, or users that created the data; Stored for years. The Parallels with Global Warming So how is unstructured data growth like global warming? People behave like this problem doesn’t exist: Every day companies are spewing out more and more unstructured data into their IT environments, but when it comes to managing this growth, it is business as usual. Despite all evidence to the contrary, many businesses are still attempting to manage and store unstructured datasets using the same approaches to data storage they’ve always used – they put it all on disk. This approach is starting to break down in the face of both the size and scale of this data. Beyond growing costs, the ability to ingest the content into a storage system quickly enough degrades over time, and traditional backup approaches are no longer sufficient to protect the data. For these massive machine- and sensor-generated datasets, clearly a different approach to storing and managing this data is required. Data that has been thought of as “cold” is starting to “warm up”: A really interesting dynamic is appearing across multiple industries. With all of these datasets, the data is generated, processed and then archived. But now more and more examples are surfacing where companies can get additional value out of this “cold” data: For video content generated for movie or TV studios, it can be repurposed and redistributed – think “behind the scenes” episodes of your favorite reality TV show. Retail companies are analyzing video surveillance footage to track shopping patterns, and using the insights to increase sales. Scientists are able to run analyses on datasets generated years ago to gain new insights and advance new innovations in their fields. Autonomous car developers are using video and sensor data generated during early test drives to make autonomous cars safer and more efficient. The list goes on, but the point is that for these types of datasets, as cold data becomes more valuable or “warms up,” the storage approach for that data needs to change. Even archived data needs to remain accessible to the users. There’s a need to act now. Before you place that next large order for more disk storage, the time is now to stop and consider other alternatives. Sticking with the status quo is the easiest approach, but also one that leads to excess storage costs and inefficiencies. What’s the Solution? To tackle this problem, let’s first introduce what might be a new term: data workflow. In some industries this is a common term, but for many industries it might be a new concept, albeit an intuitive one. All of these unstructured datasets I’ve mentioned thus far have a workflow associated with them. It looks something like this: data is generated or captured, ingested into a storage system, and stored and processed to reach some result (often collaboration between many users is required); then data is archived for long-term preservation and re-use. This process is more efficient using a storage system that is customized from the outset for specific dataset workflows. Workflow storage must handle high performance ingest when needed. Also key is the ability to share across the network to enable collaboration – as well as the ability to tier data to lower cost tiers of storage such as tape while preserving access on the network for the users and applications that need the data. This last piece is what really unlocks the ability to get more value out of the archived data in a way that doesn’t break the bank. This workflow-based approach to storage results in significant cost reductions compared to keeping all data on flash or spinning disk, and it enables other organizations to do more with their data. And, One More Parallel… By using tiered storage and keeping most of this data on low-cost, low-power storage like tape, you’re actually doing your part to help the environment, and fight global warming. Industry Perspectives is a content channel at Data Center Knowledge highlighting thought leadership in the data center arena. See our guidelines and submission process for information on participating. View previously published Industry Perspectives in our Knowledge Library. (Comment on this)
7:02p	Data Center REITs Thrive on Our Internet Addiction (Bloomberg) — This is not news to you. We all have a blossoming and apparently limitless addiction to the internet. But here’s a twist: That phenomenon is fueling the growth of real estate investment trusts specializing in data centers. Every time we stream video on Netflix, share photos on Instagram or listen to “Lemonade” on loop on Tidal, we add to the internet traffic glut — and to the pocketbooks of data REIT shareholders. That’s not going to stop anytime soon. Worldwide, internet traffic this year is expected to surpass a zettabyte (or 250 billion DVDs’ worth) of information, according to networking equipment company Cisco. The company expects current traffic to double by 2019. Read more: Data Center Network Traffic Four Years from Now — 10 Key Figures All that internet activity increasingly relies on the cloud or, in other words, servers located in data centers somewhere else). Research firm Forrester expects cloud adoption to accelerate this year. That’s good news for data center REITs, the real estate companies that house, power and cool data centers around the world. Read more: How Long Will the Cloud Data Center Land Grab Last? Share prices of the six, publicly-traded US data center REITs have risen 27 percent, on average, so far this year. Last week, Digital Realty’s stock jumped on news that it’s joining the S&P 500 (making it the second data center REIT to do so after Equinix, which is the biggest in its class by market cap). The six publicly-traded US data center REITs are (DCK): Equinix Digital Realty Trust CoreSite Realty Corp. CyrusOne DuPont Fabros Technology QTS Realty Trust Part of these companies’ successes stems from their ability to accommodate a range of data needs. Customers have to pay up if they want a data center REIT to host private servers or set up a private cloud for them, but those are appealing options for companies that want full control over their computing needs. Public cloud vendors like Amazon and Microsoft sell space on their servers to customers more cheaply (and often more conveniently) — but companies can’t manage systems as directly and self-sufficiently themselves when they do so. Many big companies use a mix of their own services and those from outside vendors (an option known as the “hybrid cloud”). Amazon itself is a customer of Equinix and Digital Realty. Equinix is located in 40 cities around the world and it’s the global leader in interconnection services (which allow different companies to connect directly with each other), according to Bloomberg Intelligence analyst Joshua Yatskowitz. Digital Realty has the most data center square footage (22.8 million square feet). It had mainly leased its data centers to wholesale customers, until its acquisition of Telx last year amplified its retail and interconnections businesses as well. Read more: Telx Acquisition Closed, Here’s Digital Realty’s Plan QTS and CoreSite are smaller than their peers but offer a range of data services, including interconnection and cloud service businesses. However rosy the stock appreciation has been for the sector, that largely appears to be a bet on future revenue, earnings, and market share growth. Revenue growth thus far for these major players has been strong. Earnings for all of the REITs have been heading in decidedly the wrong direction (a function, in part, of the large up-front capital expenditures and other large, one-time costs data centers require). Managing and sustaining growth will be a challenge for all of these REITs, especially in a sector in which, as noted above, new data centers are costly and can take one or two years to build and launch. Repurposing existing buildings for such uses is possible, but the inventory of buildings that meet data centers’ needs is limited. While REITs in general will benefit from greater investor attention after they get recategorized as their own market sector on major indices this summer, data center REITs will still have to prove that their futures are as bright as their recent past. This column does not necessarily reflect the opinion of Bloomberg LP and its owners. (Comment on this)
7:42p	Google: AlphaGo Powered by Custom AI Chip Last October, a computer system beat a professional human player at the ancient Chinese board game Go. The AI system, AlphaGo, was built by Google and trained using machine learning techniques. Google built the hardware that powered AlphaGo in-house, as it does with most of its infrastructure components. At the core of that hardware is the Tensor Processing Unit, or TPU, a chip Google designed specifically for its AI hardware, the company’s CEO, Sundar Pichai, said from stage this morning during the opening Google I/O conference keynote next to Google headquarters in Mountain View, California. This is the first time Google has shared any information about the hardware backend that powers its AI, which will play a central role in the company’s revamped cloud services strategy, announced earlier this year. TPUs will be part of the infrastructure that supports its cloud services. Related: Google to Build and Lease Data Centers in Big Cloud Expansion Pichai shared little detail about the TPU, saying only that its performance per watt was “orders of magnitude higher” than any commercially available CPU or GPU (Graphics Processing Unit): Google CEO Sundar Pichai on stage at Google I/O 2016 (Source: Google I/O live stream) “Tensor Processing Unit (TPU) is a custom ASIC for machine learning that fits in the same footprint of a hard drive, and was the secret sauce for AlphaGo in Korea,” Google said in an emailed statement. TPU gets its name from TensorFlow, the software library for machine intelligence that powers Google Search and other services, such as speech recognition, Gmail, and Photos. The company open sourced TensorFlow in November of last year. (Comment on this)
7:53p	HPE Makes $100M Bet on Startups (Bloomberg) — Meg Whitman, CEO of HPE, sat down with her top lieutenants last month at the Palo Alto headquarters to evaluate a technology storage company and a data center tools startup. She analyzed product details, asked about capital expenditures, and wondered about energy efficiency. The businesses, though, aren’t the billion-dollar acquisition targets that her company has been known for. They’re startups hoping to win investments from HPE’s venture capital arm. Putting money into startups is a way for the company to contend with new technologies from rivals like Amazon and Google. It’s also an effort to end a checkered spending pattern on acquisitions in the past decade. The business is still making purchases—it acquired Aruba Networks for roughly $3 billion last year—but venture investments provide an opportunity to make cheap bets on promising companies. HPE’s VC arm aims to do about 10 to 12 deals annually and has already closed a couple this year. “The stakes have gotten very high for them,” said Crawford Del Prete, an analyst with IDC. “It’s just a low-risk way to see if those companies play with where HP’s strengths are.” The Hewlett Packard Ventures program, also known as Pathfinder, is targeting startups that focus on big data, security, and the cloud and data center. Lak Ananth, managing director of the group, said it intends to invest approximately $100 million this year. That would roughly match what it gave to startups in 2015, its first full year. The company might invest $5 million to $10 million in an expansion-stage round—and potentially invest more in later rounds, according to Ananth. Whitman is personally involved. Every quarter she hosts “Coffee With Meg” gatherings, listens to presentations from startups, asks pointed questions, and helps decide who will get the company’s money. Her engagement has been key for Florian Leibert, CEO of Mesosphere, a data center software provider that raised $73.5 million in March, led by HPE. “That was a big reason why we were really excited,” he said. “She continues to be really responsive.” Whitman’s VC push comes after witnessing huge writedowns that cost the company more than $15 billion early in her tenure. In 2011, under Whitman’s predecessor Leo Apotheker, HP announced it would spend $10.3 billion on British software maker Autonomy. A year later the company said it was writing down about 85 percent of Autonomy’s value. Also in 2012, it said it would write down about $8 billion after purchasing Electronic Data Systems four years earlier. In November, HP underwent a massive corporate split from what’s now called HP Inc., which sells computers and printers. Now, Hewlett Packard Enterprise—which provides servers, storage gear, and tech services—is counting on the smaller, nimbler structure to better navigate the fast-changing corporate tech market. Pathfinder is a way for HPE to be involved with more experimental ideas and products. It’s not aiming for VC-style returns, though it doesn’t want to lose money, either, said Chief Operating Officer Chris Hsu. “The purpose of this is for us to actually be in the market all the time, understanding the emerging technologies,” he said. Corporate venture capital is nothing new; Google and Intel both have investment arms. In 2015 there were 801 corporate venture capital units, up 79 percent from 2011, according to Global Corporate Venturing, a research group. That includes archrival Dell Inc. and HP Inc. Pathfinder has about 10 people and should have closer to 15 by the end of the year, Ananth said. That puts it among the larger players; three-fourths of corporate firms have about five people or fewer devoted to venture capital, according to Toby Lewis, chief analytics officer of Global Corporate Venturing. Ananth said the group is also looking at possibly expanding internationally and getting into new fields such as artificial intelligence. Already, HPE is seeing gains from its investments. One customer, Ananth said, was considering moving some of its spending to Amazon’s cloud. After an introduction to Pathfinder, the customer was intrigued to see HPE had newer options. The startup sales team is now working alongside the HPE account team for that customer. Requests from customers are helping drive which companies get investments, Ananth said. There’s a lot of competition to back hot startups, but HPE’s name stands out among Silicon Valley’s many VCs; software startups want to work with the storied company to sell products. Chef Software raised $40 million from HPE and others last year and has already seen benefits from the relationship. “They bring customers,” said Chef CEO Barry Crist. “They bring a lot of enterprise experience—and a lot of reach.” (Comment on this)
10:05p	Containers and Persistent Data Storage on Docker and CoreOS By The VAR Guy As containers from Docker and other vendors grow in popularity, so does the need for enterprise-ready data storage solutions that work well with containers. Here’s an overview of the challenges on this front, and how developers are solving them. You may be wondering why data storage for containers is an issue at all. After all, in our era of scale-out storage, automatic failover and redundant arrays, figuring out ways to store and protect data is not usually difficult. The Container Data Storage Paradox Containers, however, pose a sort of paradox when it comes to data storage. That paradox is this: One of the chief advantages of using containers is that they are ephemeral. They can be spun up or down quickly, which gives your data center scalability. Yet because containerized apps are not persistent, the data inside containers isn’t, either. In other words, you can’t rely on ordinary containers to store data over the long term. You need another solution. That makes containers different from traditional, VMware-style virtualization. With traditional virtual machines, you can easily create persistent data storage by configuring virtual disks. A related problem is that, because containers are designed to be isolated from one another, as well as from the host, there is no simple way for an app running inside one container to share data with an app in a different one. Container Data Solutions So far, the solutions to this conundrum mostly fall into one of two categories. The first is to create special containers whose purpose is to store data, rather than run apps. This is essentially the approach that Docker takes with Docker Data Volumes. The advantage of this method is that Docker does a lot of the dirty work required to share data between containers for you. With a few Docker commands, you can create and share data volumes between containers. The main disadvantage is that you’re still relying on containers to store your data, and those containers do not exist forever. Relatively speaking, data volumes provide more persistent storage than you’d get by storing app data inside containers. But it’s still not persistent in the full sense of the word. The second main approach to data storage is to create a networked or cloud file system and allow containerized apps to access it over the network. This is more or less the method that CoreOS, Docker’s main competitor, promotes, although Docker containers are also compatible with CoreOS. Sharing data via the network is nice because it’s familiar to anyone who has worked with clusters or the cloud in the last decade. But the drawback is that containerized apps themselves need to be written to work with networked data storage. Many do, but it’s not a sure bet that a given app image will support data shares delivered via the network instead of the local file system running inside the container. Future Storage Solutions If the outline of container data storage above sounds complicated, it’s because it is. Data storage in the container world has come far in the last two years. But there is certainly room for more elegant, enterprise-friendly solutions to emerge in this area. Those solutions will no doubt be key in convincing more enterprises to use containers in production environments. In turn, they’ll also provide another way for organizations to cope with the data deluge. This first ran at http://thevarguy.com/cloud-computing-services-and-business-solutions/containers-and-persistent-data-storage-docker-and-co (Comment on this)

<< Previous Day

2016/05/18 [Calendar]

Next Day >>