IBM has revealed more about a PowerPC microserver that will help to analyze data gathered by the square kilometre array (SKA), the colossal radio telescope to be built across South Africa and Australia.
Once operational, the SKA is expected to generate around an exabyte – a million terabytes – of data each day. Even sorting this amount of data is going to need plenty of computing power, but also power efficeint.
The microserver is based on a Freescale T4240 chip with 12 phisical cores (24 virtual) and running Fedora 20. The power consumption is around 16W.
The T4240, with 24 virtual cores, is the flagship of the Freescale’s QorIQ T series. It is based on four Power Architecture® e6500 cores and supports:
- frequencies up to 1.8 GHz
- AltiVec technology SIMD engine
- three DDR3/3L RAM controllers
- integrated multiple 1Gbps and 10Gpbs Ethernet connections
- four PCI Express 2.0/3.0 controllers
- hardware acceleration and advanced system peripherals
Read more about the new IBM microserver at The Register and about the T4240.
IBM announced on Tuesday a new high end mainframe – the IBM z13 wich delivers the latest mainframe technology.
The z13 system culminates a $1 billion investment, five years of development, exploits the innovation of more than 500 new patents and represents a collaboration with more than 60 clients.
IBM’s new mainframe provides unprecedented new capabilities, including:
- z13 is the first system able to process 2.5 billion transactions a day – equivalent of 100 Cyber Mondays every day of the year. z13 transactions are persistent, protected and auditable from end-to-end
- z13 is the first system to make practical real-time encryption of all mobile transactions at any scale. z13 speeds real-time encryption of mobile transactions to help protect the transaction data and ensure response times consistent with a positive customer experience
- z13 is the first mainframe system with embedded analytics providing real-time insights on all transactions. This capability can help guarantee the ability of the client to run real-time fraud detection on 100 percent of their business transactions
Impromevents versus its predecessor, zEC12:
- Up to 10% faster uniprocessor performance
- Up to 40% system capacity performance improvement
- 141 cores to configure (versus 101 on zEC12)
- 231 capacity settings (versus 161 on zEC12)
- Up to 10 TB RAIM memory to improve transaction response times, lower CPU costs, simplify capacity planning, and ease deploying memory-intensive workloads
z/Architecture® enhancements designed to enable performance improvements in Linux, Java™, and DB2
- Enhanced cache design which is designed to provide twice as much second level cache and substantially more third and fourth level caches compared to the zEC12, helping to avoid untimely swaps and memory waits while maximizing the throughput of concurrent workloads
- The newest generation of FICON features, FICON Express16S 10KM LX and FICON Express16S SX, are designed to support a link rate of 16 Gbps with reduced latency for large read/write operations and increased bandwidth
Read the press release, product page and product datasheet.
Posted in Servers
Tagged IBM, mainframe, z13
Amazon announced the availability of C4 instances type, announced since last November.
Interestingly, the new C4 instances are based on a Intel Xeon E5-2666 v3 (code name Haswell) processor custom optimized for EC2, that runs at a base speed of 2.9 GHz, and can achieve clock speeds as high as 3.5 GHz with Intel® Turbo Boost.
There are 5 instance types, ranging from 2 virtual cores (on phisycal with HyperThreading enabled) and 3.75 GB of memory to 36 virtual cores and 60 GB of RAM:
Intel® Xeon® Processor (E5-2666 v3) is based on the Intel Haswell micro architecture that is a notable improvement on its predecessors. It is better at predicting branches and more efficient at prefetching instructions and data. It can also do a better job of taking advantage of opportunities to execute multiple instructions in parallel. This improves performance on integer math and on branches. This new processor also incorporates Intel’s Advanced Vector Extensions 2. AVX2 supports 256-bit integer vectors and can process 32 single precision or 16 double precision floating point operations per cycle.
More info at Amazon and The Register.
Mesosphere, a startup building a “data centre operating system” based on Apache Mesos, has landed $36m in funding in a round led by Khosla Ventures with backing from others including Andreessen Horowitz. The cash brings Mesosphere’s total funding to approximately $50m.
Apache Mesos is based on a 2009 idea from UC Berkeley PHD student Benjamin Hindman and that went industrial in 2010, when he was hired to install it at Twitter. Mesosphere has been founded last year to build a commercial version of the Mesos cluster management system.
News of the money came the day Mesosphere released DCOS, its incarnation of its programmable data-centre vision. It’s a set of software and services to co-ordinate resources multiple servers and platforms.
DCOS is a distributed systems kernel using a set of core system services including Marathon, Chronos, DNS and HDFS for set up, discovery and storage.
Platforms supported by DCOS include Red Hat, CentOS and Ubuntu, AWS, Microsoft Azure, Google GCE, VMware and OpenStack and it comes with an API and SDK.
Read more at The Register, Apache Mesos website and Mesosphere.
Cycle Computing which specializes in technical and high-performance cloud computing, and Western Digital’s HGST unit ran the largest Fortune 500 production workload and went from zero to 50,000 cores in 23 minutes.
The cloud computing run utilized various regions on Amazon Web Services (AWS) platform and topped at 70,980 cores. In 60 minutes, the team built a 729 TeraFLOPS cluster, that would be just a little abowe the #71 supercomputer on the November 2014 Top500 list.
HGST’s engineering team was working to identify an optimal advanced drive head design. Essentially, the job involves 1 million simulations for designs based upon 22 different design parameters running on 3 drive media. Running these simulations using an in-house, specially built simulator, the workload takes about 30 days to complete. But using the AWS cloud, it took just 8 hours and $5,593.94.
Read more about at ZDnet and HPCwire.
The new product, launched during SC14, is pitched as double-your-everything: twice the performance and twice the memory bandwidth.
The K80 has 2x Kepler GK210 GPUS that claim 8.74 teraflops single-precision and 2.91 teraflops double-precision. Each board has 24 GB of GDDR5 memory (12 GB for each GPU) and 480 gigabytes/second memory bandwidth. There are 4,992 CUDA cores, and NVIDIA’s GPU booster (automatic overclocking) and dynamic parallelism support.
The product is targeted at HPC market, doubling the number of GPUs that can fit in the same server or workstation form factor as the K40.
Read the product page and a description on nVidia Blogs.
Dell lauched on Wednesday Today Dell announced a new server architecture that combines characteristics of both rack servers and blade servers into a new architecture, known as PowerEdge FX.
The base of the architecture is the 2U PowerEdge FX2 enclosure designed to hold 1 to 8 sleds, depending upon the type of sled used.
- FC630 half-width 2-socket server block with Intel® Xeon® processor E5-2600 v3 product family performance (up to four per FX2 chassis)
- FM120x4 microserver block, each consisting of a half-width “sled” that houses 4 microservers running the low-power Intel® Atom™ C2000 System on a Chip (SoC) processor (up to 4 sleds with 16 total microservers per FX2 chassis)
- FN IO Aggregator that is able to simplify network deployment by aggregating cabling from 8 cables to as little as a single cable; offers cost-effective 10GbE performance; and enables LAN/SAN convergence
Coming soon, new modular building blocks that will continue to accelerate your software-defined data center:
- FD332 half-width storage block with up to 16 small-form-factor (SFF) 2.5″ DAS devices (up to 3 blocks per FX2 chassis – with one FC630 server)
- New Intel® Xeon® powered server blocks will include the FC430 quarter-width 2-socket block (up to 16 per FX2 chassis) and the FC830 full-width 4-socket block (up to 2 per FX2 chassis)
Read the product page and the Press Release.