Thursday, August 16, 2018

Google Cloud Platform: Building a Hybrid Render Farm

Google Cloud Platform posted an online guide for building a render farm using its scalable architecture.

Render farms, which are needed by content creators for outputting digital productions, can be very expensive to build on-premise and require a lot of electricity to run.

GCP is now enabling render applications to scale from 2- or 4-core VMs all the way to a 160-core VM with up to 3844 GB of RAM. Up to 8 GPUs may be attached to any VM to create a GPU farm.

https://cloud.google.com/blog/products/gcp/building-a-hybrid-render-farm-on-gcp-new-guide-available


Gartner highlights emerging risks for cloud computing, including GDPR

In its latest quarterly Emerging Risks Report, Gartner highlights a number of new risks for enterprises in using cloud resources, including new risks from cybersecurity threats, information disclosure leaks, and General Data Protection Regulation (GDPR) compliance. The report is based on a survey of 110 senior executives in risk, audit, finance and compliance at large global organizations.

Gartner identifies social engineering and GDPR compliance as "most likely to cause the greatest enterprise damage if not adequately addressed by risk management leaders." Enterprises are also continuing to struggle with cybersecurity.

Nevertheless, the research consultancy remains bullish on the growth of cloud services, predicting cloud computing to be a $300 billion business by 2021.

https://www.gartner.com/en/newsroom/press-releases/2018-08-15-gartner-says-cloud-computing-remains-top-emerging-business-risk


Berlin Commercial Internet Exchange deploys Coriant Groove G30

The Berlin Commercial Internet Exchange (BCIX), Berlin’s leading Internet Exchange Point, has selected the Coriant Groove G30 Network Disaggregation Platform to scale its distributed peering infrastructure. BCIX will use the Coriant platform to connect major data centers across Berlin. High-capacity neutral exchange services will include 100G Data Center Interconnect (DCI).

“The German economy continues to experience strong growth. The year over year doubling of IP traffic volumes we are seeing in the nation’s capital is further evidence of a robust business environment,” said Christian Kroeger, Managing Director, BCIX Management GmbH. “To meet the evolving high-speed connectivity demands of our customers, we are boosting the scalability of our backbone network with the Coriant Groove G30. After rigorous testing and analysis of competitive offerings, the Coriant solution demonstrated superior performance and flexibility, and met our stringent performance requirements for scalable, resilient, and low latency transport.”

Coriant said its flexible plug-and-play modular architecture and the recent introduction of a new Coherent Multiservice Sled, enables its Groove G30 solution to provide BCIX with a seamless and cost-effective path to higher interconnect speeds, including 400G client side and 600G line side transmission – without the need for forklift upgrades or costly and complex flexi-grid ROADMs.

“Berlin is an important internet hub not only for Germany, but also for the rest of Europe, and BCIX plays a critical role as a nexus point connecting many of the world’s most recognized brands and tech giants,” said Ronald Van der Kraan, Managing Director, Europe, Coriant. "We are excited to be working closely with BCIX as they enhance their customer value proposition and extend their market leadership as a premier provider of high-capacity internet exchange services.”

Coriant debuts 1.2 Tbps modular sled for its Groove G3

Coriant introduced a new Coherent Multiservice Sled for its Groove G30 Network Disaggregation Platform.

The new 1.2 Tbps modular sled leverages advances in integrated photonics and the latest 16nm coherent DSP.

Coriant said its new Groove G30 Coherent Multiservice Sled enables the platform to achieve new performance benchmarks, including:

  • Highest density – 9.6 Tbps of capacity in a compact and highly modular 1RU, a 2X improvement over the closest comparable solution; delivers unprecedented OpEx savings as networks scale
  • Lowest power consumption – the Groove G30 Coherent Multiservice Sled requires only 0.16 watts per gigabit of power, representing up to half the power consumption of competing solutions
  • Advanced programmability – programmable baud rate (30-70G baud), FEC (0-27%), and modulation depth (QPSK to 64QAM) enable operators to fine-tune and cost-optimize any transport application for the optimal spectral efficiency, lowest latency, and required transparent reach
  • Unmatched speed and reach – 600 Gbps single wavelength transmission for DCI and metro applications (up to 38.4 Tbps of capacity per fiber); 400 Gbps up to 1,000 kilometers, and 200 Gbps up to 4,000 kilometers for long-haul applications

Telstra posts declining profitability, impact from nbn

Telstra reported flat  FY18 revenue of AU$26 billion, with EBITDA declining by 5.2 percent to AU$10.1 billion. Net profit was AU$3.5 billion, down by 9 percent from AU$3.9 billion in FY17.

The company attributed its declining profit to the National Broadband Network (nbn) and mobile competition.

Telstra CEO Andrew Penn said the results showed strong customer growth for the year and good progress on Telstra’s productivity program, however, the continued downward pressure on EBITDA and NPAT caused by the further rollout of the nbn and lower Average Revenue per User (ARPU) clearly reinforced the importance of the T22 strategy.

“We have seen strong subscriber growth, particularly in the second half of the year, adding 342,000 retail mobile customers, 88,000 retail fixed broadband customers and 135,000 retail bundles during FY18,” Penn said. “Despite this, the challenging trading conditions are expected to continue in FY19, including ongoing pressure on ARPU and further negative impact of the nbn network rollout on our underlying earnings.

Some highlights for the first half of the year

  • Mobile net adds - 235,000 retail mobile customers including 130,000 postpaid handheld; churn 10.9%
  • 21,000 Belong mobile; +118,000 wholesale mobile customers
  • underlying core fixed costs declined 7.2% 
  • Fixed EBITDA (ex nbn C2C1)-29% negatively impacted by growing nbn network payments and loss of wholesale margins. 
  • nbn impact in period $370m; $870m life to date 
  • >1m Telstra TVs
  • >1.5m AFL, NRL, Netball Telstra Live Sports Pass users
  • New nbn connections +454,000 (market share 51%, ex- satellite); retail bundle adds +57,000
  • Mobile service revenue -1.2% with post-paid 
  • handheld ARPU -2.9%
  • Global connectivity revenue +6.7% (LC)



NVIDIA posts 40% sales growth in Q2 but warns on crypto sales

NVIDIA reported revenue of $3.12 billion for its second quarter ended July 29, 2018, up 40 percent from $2.23 billion a year earlier, and down 3 percent from $3.21 billion in the previous quarter.
GAAP earnings per diluted share for the quarter were $1.76, up 91 percent from $0.92 a year ago and down 11 percent from $1.98 in the previous quarter. Non-GAAP earnings per diluted share were $1.94, up 92 percent from $1.01 a year earlier and down 5 percent from $2.05 in the previous quarter.

“Growth across every platform - AI, Gaming, Professional Visualization, self-driving cars - drove another great quarter,” said Jensen Huang, founder and CEO of NVIDIA. “Fueling our growth is the widening gap between demand for computing across every industry and the limits reached by traditional computing. Developers are jumping on the GPU-accelerated computing model that we pioneered for the boost they need.

However, Colette Kress, NVIDIA's CFO, warned that boom in sales for crypto mining applications is over. In a prepared statement on the company's investor call she said "“Our revenue outlook had anticipated cryptocurrency-specific products declining to approximately $100 million, while actual crypto-specific product revenue was $18 million.”

For its data centers category, quarterly sales amounted to $760 million, driven by demand from hyperscale customers.

Intel acquires Vertex.AI for deep learning

Intel has acquired Vertex.AI, a start-up based in Seattle, for its work in deep learning. Financial terms were not disclosed.

Vertex.AI has contributed to the PlaidML deep learning open source project, which Intel plans to transition to the Apache 2.0 license.

Vertex.AI is now part of Intel’s Artificial Intelligence Products Group.

FWD: Intel's data center connectivity ambitions

Intel's Data Center Innovation Summit this week in Silicon Valley, which coincided with the annual Flash Memory Summit across the street at the Santa Clara Convention Center, was the company's big event for resetting expectations about its strategic direction inside the data center.

Everyone knows that big data is the new hot commodity of our times, and that data centers are repositories where the new wealth is extracted and refined. Intel already dominates workload processing in this space with its well-entrenched Xeon franchise. Its new Optane storage promises to redefine the storage hierarchy in the data center. The third leg of the stool is communications, and while Intel has deep roots in Ethernet and network adapters, it has been burned before in trying to capture the switching and transport components of the data center.

Intel's ambition is now clear -- to offer an end-to-end data center connectivity portfolio, leveraging its FPGA technology (the Altera acquisition) with its homegrown silicon photonics technology, to capture the $11 billion data center connectivity opportunity.

The following slides were captured from Intel's daylong keynotes, which are now archived here:

http://intelstudios.edgesuite.net/180808_dcis/index.html
.




Intel has some of these pieces of the puzzle already, especially its Ethernet network interface cards and FPGA-based acceleration cards. Earlier this year, Intel confirmed that FPGA-powered accelerator cards will be a range of Xeon powered servers this year from major OEMs, including Dell EMC and Fujitsu. 



The new component announced this week is a family of SmartNICs designed to offload workloads such as firewalling or load balancing from the local server processor or inline dedicated appliance.  Intel plans to introduced a 2x25 Gbps SmartNIC, followed by a 100 GigE model.  

A number of vendors already offer SmartNIC products, including ones powered by FPGAs, and hyperscale cloud companies, including Microsoft and Baidu, have talked of their SmartNICs for supercharging their data center networks. Intel thinks this trend is just beginning and that the industry will soon see programmable NICs re-directing network traffic and orchestrating the movement of VMs and containers between available resources.



Intel's new expectation is that FPGA, whether in SmartNICs or acceleration boards, will become a standard part of the infrastructure for applications from AI to video processing.



The second major component of Intel's data center connectivity strategy is Silicon Photonics, where it claims an advantage gained from years of R&D into integrating and manufacturing lasers with Indium Phosphide.


See also