The development is called the SeaMicro SM10000-XE. Needless to say, it’s the first fabric-based Xeon microserver ever made.

It’s also supposed to be the most energy-efficient, highest-density, highest-bandwidth Xeon server now available, period.

A single SM10000-XE replaces 32 dual-socket servers, but draws half the power and takes up a third the space without any changes to operating systems, applications or management tools.

It eliminates layers of Ethernet switches, server management devices and expensive load balancers.

Because it’s Xeon-based it can run heavyweight scale out workloads while the Atom-based SM10000-HD is meant for the highly parallel workloads found in the web tier. That means that microservers can capture a bigger piece of the data center. SeaMicro says the microserver will go mainstream running Apache Hadoop, real-time analytics, Java apps, PHP, Memcached and NoSQL.

The widget is built around the Xeon E3-1260L based on Sandy Bridge architecture and Samsung’s new low-power 1.35V Green DDR3 DRAM memory. Since it’s breakthrough, Intel and Samsung were both at its launch Tuesday when SeaMicro CEO Andrew Feldman announced it was shipping in volume.

Microservers are all about density, space and power. The Xeon E3-1260L has a 45W TDP envelope, providing 30% better performance per watt than processors from Intel’s previous generation, and SeaMicro has invented a new technology called TIO, short for Turn It Off, that lets it power-optimize the parts by turning off unnecessary CPU and chipset functions.

The chip’s four cores run at 2.4GHz CPU core. The clock can be throttled to 3.3GHz with Intel’s Turbo Boost Technology 2.0 and it’ll support eight threads with Intel Hyper-Threading Technology.

The SM10000-XE contains 64 CPUs for 256 2.4GHz cores in a 10U or 1,024 cores in a standard rack. It delivers 10 gigabits of bandwidth to each quad-core processor, which SeaMicro says sets a high watermark for bandwidth per unit compute.

It supports up to 32GB of DRAM per socket for a system total of 2.04TB.

The processors are tied together by SeaMicro’s Freedom Supercompute Fabric, which delivers 12 times the bandwidth per unit compute of a traditional server.

The XE also supports up to 16 10-gigabit Ethernet uplinks or 64 one-gigE uplinks. And unlike other microservers, it can support up to 64 SATA hard disks or solid-state disks without reducing computational density.

SeaMicro’s boards are Spartan in their minimalism consisting of only the processor, Samsung’s DDR3 and SeaMicro’s new Fabric ASIC, reportedly the industry’s first second-generation fabric ASIC, capable of supporting both large cores like the Xeon’s and small cores like the Atom’s.

SeaMicro figures eventually it’ll be able to design systems using Atoms and Xeons together in the same enclosure.

The XE’s austerity derives from SeaMicro’s Input/Output Virtualization Technology which reduces the component count, shrinks the motherboard and reduces power, cost and space.

The Freedom Supercompute Fabric is built of multiple Freedom ASICs working together, creating a 1.28 terabits a-second fabric that ties together 64 of the power-optimized mini-motherboards at low latency.

Samsung’s small high-density 30nm Green DDR3 is supposed to achieve more than a 70% power savings over 1Gb 1.5V 50nm class DDR3, improving SeaMicro’s TCO.

The SM10000-XE lists for $138,000 for a base configuration.

CompSec, which works with the US intelligence community, has adopted the widgetry in mission-critical applications to deliver huge amounts of compute power to remote and hard-to-access locations. It says it “greatly reduces response times, provides faster and more in-depth analysis, and helps to advance their mission.”

Mozilla is also on board. Matthew Zeier, director of IT infrastructure and operations, says that between the Atom boxes and the Xeon boxes “SeaMicro is able to meet the computational needs of our entire data center while driving down operating expenses by saving us power and space. We are a very happy customer.”

Intel has previously acknowledged that microservers could represent 10% of the server market by 2016. Its data center unit general manager Jason Waxman explained Tuesday that that forecast included both Atom and Xeon widgetry. Intel may be somewhat conservative considering the Xeon chip SeaMicro is using costs less than $300.

SeaMicro expects the next Atom generation to move chipset functions into the SoC, which will save more power and give it more real estate for processors.

SeaMicro rival Calxeda is building ARM chips into microservers.

It’s the first time Amazon has proposed putting its own software on the ground inside a corporate data center. And the stuff’s targeted at large corporations. Amazon says some customers asked for such a solution. It also expects resellers to offer the service.

It is of course proprietary and a competitive problem for other cloud storage and gateway suppliers. Come to think of it, Amazon as repository of corporate data is a problem for a lot of people.

The Gateway – think of it as an EBS snapshot-taking umbilical cord – will make it easier to use Amazon S3 storage for disaster recovery, business continuity and data mitigation as well as low-latency backup and once a company’s data is mirrored in the cloud – particularly since it’ll probably be stored there as EBS volumes – well, users will be tempted to experiment with other AWS services like EC2. Might as well, the data’s already there.

You see where this is going, right?

As Amazon put it in its official announcement, “The AWS Storage Gateway also makes it easy to leverage the on-demand compute capacity of Amazon Elastic Compute Cloud (Amazon EC2) for additional capacity during peak periods, for new projects, or as a more cost-effective way to run normal enterprise workloads.”

Cloud backup of course promises the inevitable, tantalizing reduction in on-premise hardware, manpower and fears of tapes getting lost in transit.

The Gateway only needs commodity hardware and works with exiting DAS and SANs and applications. Amazon promises the data will be transported and stored securely using the industry standards required by banks and healthcare.

No applications have to be re-architected to accommodate the Gateway. It uses a standard iSCSI interface, transfers data to S3 over SSL, and AES encrypts it when it gets there. Then it makes it redundant.

Amazon is offering a 60-day free trial after which the service will run $125 a month for each Gateway, with storage set at 14 cents a gigabyte a month.

Each Gateway can handle up to 12TB of storage across 12 volumes. Fear not, accounts can have multiple gateways.

The service is available from Amazon data centers in the US, Ireland, Singapore and Tokyo.

The widgetry is still a beta and relatively rudimentary. Initially – and non-threateningly – all the data will be kept on the appliance – and on-premise – while it’s backed up in the cloud – so-called Gateway-Stored volumes.

Gateway-Cached volumes, where the only full copy of the data is in the cloud, and the data in the appliance is only what’s frequently accessed, won’t be available for a few months. There’s no deduplication or WAN optimization yet either although Amazon likely has plans.

Amazon said the first release of the Gateway takes the form of a VM image for VMware ESXi 4.1, with plans to support other virtual environments in the future. It currently supports mounting of its iSCSI storage volumes using Windows and Red Hat iSCSI Initiators.

That’s the one it said in October would be ready sometime in the first half. Only nobody believed it meant early in the first half. Heck, it’s not even clear anybody thought Oracle could make the first half at all and it probably couldn’t have met so early a date if it hadn’t been secretly closeted for months with Cloudera.

It’s using Cloudera’s version of Hadoop in the thing rather than lose time dicking around rolling its own.

Cloudera is the oldest, most established of the Hadoop start-ups whose ranks now include MapR (tight with EMC, its Greenplum database and the EMC Data Computing Appliance) and Hortonworks (buddies with Microsoft and SQL Server 2012) and it’s assumed to have more customers and more experience than anybody else.

In response John Schroeder, CEO and co-founder of MapR, said, “It’s ironic that the world’s largest database vendor would enter the Big Data market by being a hardware provider and partnering for software. It appears they are taking a page from their Linux playbook where they partner until prepared to support their own distribution. Unlike Linux, Hadoop requires innovations in reliability, performance and ease of use to drive adoption like those brought to market by MapR.”

Observers say Oracle’s use of Cloudera shows it’s serious. Big Data is supposed to be a $70 billion industry growing at maybe 20% a year and Oracle wants more than its fair share so it’s not letting any grass grow under its feet.

To prove it’s serious, Oracle is low-balling the highly engineered system. Rather than charge millions like it does for its Exadata, Exalogic and Exalytics appliances, Oracle’s Big Data Appliance will go for a mere $450,000 a rack with maintenance on both the hardware and software running only 12% a year. The price is a third less than expected.

For the money customers will get a full rack of 18 Sun Fire x86 servers with 216 CPU cores, 864GB main memory, 648TB of raw disk storage, 40 Gb/s InfiniBand internal connectivity and 10 Gb/s Ethernet connectivity, perfectly sized for the greatest number of customers.

Users also get Cloudera’s open source Distribution Including Apache Hadoop (CDH) and Cloudera Manager software, Cloudera’s Google Big Table-ish HBase, an open source distribution of R, the programming language, the Community Edition of Oracle’s NoSQL Database, Oracle’s HotSpot Java Virtual Machine and Oracle Linux, the Oracle fork of Red Hat. The widgetry can be used in multiple ways.

Oracle and Cloudera are going to split support, with Cloudera getting the hard software questions.

Oracle’s also got a bunch of separately priced connectors so users can integrate data stored in the CDH Hadoop Distributed File System (HDFS) or Oracle NoSQL Database with Oracle Database 11g. The four connectors cost $2,000 per server processor.

Betcha Oracle figures it can up-sell Big Data Appliance users on Exadata, Exalogic and Exalytics since everything’s tightly integrated.

It’s also possible that Oracle might want to buy Cloudera eventually depending on how things go and how its vision of itself as a database company morphs. Currently they’re bound together by a non-exclusive multi-year alliance.

A huge win for Cloudera, the start-up is reveling in the validation it’s getting from Oracle and all the feet Oracle can put on the street. It can probably anticipate an uptick in its consulting and training business. It also figures the Oracle ecosystem will produce new tools, applications, systems and services in support the CDH platform.

Oracle in its conference call never once said it was, claiming instead that Q2 was a one-off event and that “short of a global meltdown” Q3 “won’t be a repeat.”

Still, the news ricocheted through high-tech leaving the Nasdaq a blood-red mess. Oracle itself is now down about 13%.

Watchers are still trying to figure out what happened. Obviously there was a struggle to close deals in Q2, which ended with the hateful November. Ellison, Catz and Hurd are three pretty slick articles, still it’s hard to believe that the company didn’t already have the management controls in place to monitor last-minute approvals but that’s what Catz said. Approvals came through – lurchingly and apparently with more, higher-up signatures than they used to need – and Oracle has had to figure out how to deal with that wrinkle going forward. Apparently any deals missed are expected to close this quarter.

Both its Q2 revenue and earnings were short of consensus. It reported earning 54 cents a share, up 6%, on revenues of $8.79 billion, up 2% year-over-year, when the Street thought it would do 57 cents on $9.23 billion. GAAP income was up 17% to $2.2 billion or 43 cents a share. Unfortunately Oracle had guided to revenue of $8.99 billion-$9.34 billion.

It said new licenses were up 2% to $2 billion when it had guided to an increase of 6%-16% and updates and product support was up 9% to $4 billion.

It’s supposedly not seeing a slowdown in Europe. It even said the US public sector was pretty good. CRM is supposed to be up close to 20%. Most of everything else, however, looks off. Currency also turned into a headwind.

Hardware was down 14% to $953 million – when it was supposed to be flat to down 5% – and that’s despite what Oracle CEO Larry Ellison said was accelerated sales of engineered systems.

According to him “Exadata growth was well over 100% compared to last year, and Exalogic grew more than 100% on a sequential basis. We shipped our first SPARC SuperCluster in Q2 and expect to begin deliveries of our Exalytics system and the Oracle Big Data Appliance in Q3.”

Evidently the product transition from the T3 to T4 Sparc chip was a hold-up since it requires a brand new system and people aren’t going to buy outdated equipment so late in the cycle. In addition, Oracle could only deliver a few SuperClusters since they only became available at the end of the quarter. But Hurd maintains that Oracle pipeline is as full as it’s ever been, with Exalogic ramping faster than the Exadata machine, something he’s said before.

Larry said 200 Exadata/logic machines were sold in Q2 and prophesized that 300 would be sold this quarter and 400 in Q4. By then they’ll be a billion-dollar business that he said will double next fiscal year. Bernstein ace analyst Toni Sacconaghi noticed that wasn’t quite as many as previously forecast and Ellison had to admit he was right. Oracle might not triple its installed base, maybe it’ll only be up 2.5x, Ellison said, but “It’s still spectacular.” Oracle cut the appliance projections from 2,000 to 1,000.

Oracle’s GAAP operating margin was 35%; its non-GAAP margin was 45%, which it’s pleased enough with, and it claims it should return to pre-Sun margins soon.

Oracle’s workforce was up 1,700 salesmen in the first half.

Co-president Mark Hurd said in a statement, “We believe that this increase in our field organization combined with innovative new products like Fusion Cloud ERP and Cloud CRM will enable solid organic growth in the second half of this year.”

Oracle exceeded estimates in the four previous quarters and it’s going to have tough compare this quarter because last year was sensational. It’s forecasting total revenues will be up 3%-7% this quarter. That would work out to somewhere between $9 billion and $9.42 billion. It said new software licenses would be flat to up 10% and hardware down 5% to 15%. Adjusted earning should be 56 cents-59 cents. The Street had it down for 58 cents on $9.46 billion.

To console investors Oracle said it could buy back another $5 billion worth of stock at some point.

It’s got $31 billion in the bank.

The unit combines four existing divisions: netbooks and tablets, ultra mobility (smartphone processors), mobile communications (baseband) and mobile wireless (Wi-Fi).

The idea, it said, is to “speed up and improve the development process.”

At least there’ll be less duplicated effort, but what Intel needs and quickly is a winning energy-efficient design.

It’s supposed to have a new Atom mobile chip code named Medfield early next year. The widget and Intel’s alliance with Google to optimize Android for the Atom may or may not improve its traction (while its old buddy Microsoft cozies up to ARM and ports Windows 8 to its architecture).

The new group will be run two-in-box by Hermann Eul, once a top Infineon executive, and Mike Bell, who used to work at Apple on the iPhone and joined Intel last year from Palm.

Bell has co-managed the old ultra mobility unit since March and after Intel acquired Infineon and before his latest apotheosis, Eul was running Intel’s Mobile Communications Division, created out of Infineon.

Fortune broke the reorg story Wednesday morning when Intel sent out an internal memo and Intel confirmed it.

Earlier this week, Intel cut its Q4 guidance on reduced orders caused by the hard drive shortage created by the crippling floods in Thailand.