AWS Announces New Headquarters and More
Amazon confirmed that New York City and Arlington, Virginia will be the homes of their second headquarters, HQ2, as its being monikered. This means the company will now have three global headquarters. Amazon says it will invest $5 billion, $2.5 billion in each location, and create more than 50,000 jobs across the two, with more than 25,000 employees each in New York City and Arlington. The company also announced several new cloud products, including the availability of a second GovCloud Region in the United States, and future data centers in Italy.
All of these announcements, and we’re looking forward to more, come just a little over a week before AWS re:Invent held from November 26-30 in Las Vegas. DoubleHorn’s Uttam Reddy and Paul Teich will be at the event covering for Liftr Cloud Insights. Check the Liftr Cloud Insights Twitter page and future Liftr Cloud Look Aheads to follow the event.
Tencent Earnings Report is In
Last Wednesday, the Chinese conglomerate Tencent, released its quarterly earnings report. Tencent’s reported revenue for the third quarter exceeded 80 billion Yuan. The company’s third quarter cloud services revenue more than doubled year-on-year and increased at a double-digit percentage rate quarter on quarter. Tencent appears to be successfully making the shift away from its core gaming business to other strategic opportunities such as advertising, digital content, payment and cloud services.
Google Experiences Major Outage
Last Monday, Google experienced a major outage that rendered its search and other key services unusable. The outage even affected Spotify users and other cloud customers not directly utilizing Google products. The issue was caused by network traffic redirections came in five distinct waves over a 74-minute period. The IP ranges from the redirection transmitted some of Google’s most sensitive communications, including the company’s corporate WAN infrastructure and the Google VPN, according to BGPmon on Twitter.
Although Google has said that the outage wasn’t a malicious attack, there is still suspicion surrounding it due to traffic redirection went through the networks of China Telecom, the Chinese government-owned provider. China Telecom was recently caught improperly routing traffic from several Western carriers through mainland China.
What does this mean for the security of Google Cloud? Google has stated they will be doing a thorough investigation of the incident.
If you tuned in last week, we reported that Alibaba had goals to double its 2017 sales of $25.3 billion on its globally renown Singles Day. While the company didn’t meet that goal, they still ended the day with record breaking numbers, coming in at $30.8 billion dollars; though, their sales growth percentage was only 27% in comparison to last year’s 39% growth.
Insights from Supercomputing 18
Our Principal Analyst, Paul Teich attended Supercomputing 18 in Dallas, Texas last week and brought back with him key insights and observations from the event. Microsoft’s agreement to host Cray supercomputers in Azure datacenters, announced last year, progressed, with Microsoft showing the specific Cray “Sentinel” configuration that will be collocated in Azure datacenters.
Google showed off NVIDIA’s Tesla T4 GPU-based add-in card at their booth. Google will initially target T4-based compute instances for deep learning inference as a service applications. Google uses its own first-generation TensorFlow processing unit, or TPUv1, for its internal inference-based services, such as various types of searches. However, hardening development tools and creating a broad ecosystem of support is very expensive. NVIDIA’s developer tools are the de facto industry standard for machine learning today. Google seems to be deploying T4 cards for external access instead of trying to compete with NVIDIA on devtools; at least for the time being.
Two products getting a lot of attention at SC18 were Xilinx’s new Alveo FPGA add-in cards and AMD’s EPYC server processor. AWS launched AMD EPYC instances the week before SC18, and AMD EPYC server designs were in a surprising number of system manufacturer’s booths. EPYC is getting attention from HPC hardware vendors and AWS is the first public cloud vendor to offer AMD EPYC to external customers.
The Alveo FPGA PCI-Express add-in cards were in almost every booth that NVIDIA GPU PCI-Express add-in cards were shown. This shows huge support by hardware and software vendors across the expo for offering machine learning and deep learning hardware acceleration options. We expect this trend to continue, especially for inference as a service applications.
If you are not attending AWS Re:Invent in person, make sure to sign up for its keynote and launchpad livestreams. Liftr Cloud Insights’ Principal Analyst Paul Teich will also be at KubeCon /CloudNativeCon in Seattle on December 11-13, so be sure to look for him there.