LinuxGram

Microsoft filed a formal complaint with the European Commission Thursday charging Google with unfair practices in search, online advertising and smartphone software, a broader raft of charges against the search giant than the EC is believed to be currently investigating.

It is the first time Microsoft has ever complained to the EC about anybody.

Historically the complaints have been made against Microsoft and its chief lawyer Brad Smith took note of that fact saying, “Having spent more than a decade wearing the shoe on the other foot with the European Commission, the filing of a formal antitrust complaint is not something we take lightly.”

However, one of the complaints that got the EC to open a formal antitrust investigation of Google last year was made by a small Microsoft online shopping site property in Germany that charged Google with skewing its search results.

Google has subsequently claimed that Microsoft was behind the whole ruckus. Of course, Google had complained to the EC about Microsoft’s web browser.

Microsoft’s own complaint apparently claims Google prevents Bing, Microsoft’s search engine, from crawling and indexing YouTube, which Google owns, and says Google – how’s this for ironic – is withholding technical information that would let Window Phone 7 software fully display video content from YouTube.

The metadata is supposedly available only to Android and Apple phones under a deal Google CEO Eric Schmidt cut when he was still on Apple’s board. Apple of course does not own a search engine.

Microsoft has also complained about Google’s advertising contracts prohibiting advertisers and agencies from using third-party software to compare results and move their own data from Google’s AdWords to Microsoft’s adCenter.

Smith blogged that “This makes it much more costly for Google’s advertisers to run portions of their campaigns with any competitor, and thus less likely that they will do so. That is a significant problem because most advertisers figure that they have to advertise first with Google. If it’s too expensive to port their advertising campaign data to competing advertising platforms, many won’t do it. Competing search engines are left with less relevant ads, and less revenue. And while this restraint isn’t visible to consumers, its effects are nonetheless felt across the web. Advertising revenue is the economic propellant fueling the billions of dollars needed for ongoing search investments. By reducing competitors’ ability to attract advertising revenue, this restriction strikes at the heart of a competitive market.”

Smith said Microsoft has given the EC a “considerable body of expert analysis” on search engine algorithms to support its case that “Google has engaged in a broadening pattern of walling off access to content and data that competitors need to provide search results to consumers and to attract advertisers.”

Google controls 95% of the search market in Europe and 65%-75% of it in the US.

EC spokeswoman Amelia Torres described the EC’s investigation of Google, which began in November, as still being at the “preliminary stage.” It could be years at this.

See http://blogs.technet.com/b/microsoft_on_the_issues/archive/2011/03/30/adding-our-voice-to-concerns-about-search-in-europe.aspx.

The next best thing to having a monopoly is owning the standard, and if not the de facto standard, well then, by gum, a de iure standard, and there’s not a red-bloodied cloud player out there who wouldn’t thrown his grandmother under an oncoming train to be that standard.

Amazon’s grandmother looks safe for the moment, but Red Hat’s grandmother better start worrying because Red Hat’s sent its Apache Deltacloud API specification down to the Distributed Management Task Force (DMTF) to be prayed over by the DSMTF’s Cloud Management Work Group and resurrected as the arbiter of IaaS cloud portability and interoperability – and we all know how important that’s supposed to be even if early cloud users ignored history largely because, if they waited, they’d never have gotten cloud-borne.

Anyway, the Apache Deltacloud project is an open source implementation of a RESTful Web Service API abstracting common proprietary IaaS cloud management APIs like Amazon’s and Rackspace’s.

It was started by Red Hat last September and moved to the cover of the Apache Incubator earlier this year where – although it’s kinda dependent on Red Hat’s own Cloud Engine widgetry – it’s reportedly gotten “various stages of support and participation” from Cisco, Dell, Cisco, Dell, Gogrid, Goldman Sachs, HP, IBM, Ingres, Intel, Nimsoft, Opsource and Symantec.

Red Hat claims the move make Deltacloud “the only major cloud framework that isn’t tied in some way to a single company’s proprietary code, APIs or other intellectual property.” That doesn’t mean that it isn’t to Red Hat’s advantage, or that Red Hat’s code won’t run best on the thing. Heck, portability in itself is an evangelical advantage.

It plays to the futuristic vision of people moving on-demand – presumably painlessly and regardless of infrastructure or, God forbid, dependencies – between private and public clouds with an SLA, which is what the Hatter says the enterprise and the government want so long as everything can be managed from one place. Hope they’re not holding their breath

Deltacloud is key to Red Hat’s Cloud Foundations, announced in June, which is Red Hat’s everything-we-own-tossed-into-it cloud stack

Red Hat is striking early, probably prematurely, on the standards front ahead of other open source cloud projects like the newfangled OpenStack effort organized recently by Rackspace and still in development.

So far what makes it different, Red Hat says, is that it’s conceived as a Web Service and that means:

- The API can either be offered directly by the cloud provider or by individual users running their own server;
- Client libraries can be written in any number of computer languages, and are already available for popular ones;
- The core API logic resides on the API server, enabling consistent behavior across all client libraries; and
- Support for new clouds can be added to the API without changes to clients.

Drivers for Amazon EC2, GoGrid, OpenNebula, Rackspace, Red Hat Enterprise Virtualization Management, RimuHosting, Terremark and vCloud are either written or in-process as is a Microsoft Azure cloud storage driver.

Of course if you can’t hoist apps up into the cloud the exercise is kinda useless so, aside from wanting to be the cloud’s interoperability bridge, Red Hat’s got an application infrastructure or platform-as-a-service (PaaS) play afoot too à la Azure, Google App Engine and VMforce that’s based around its JBoss Enterprise Middleware also part of its Cloud Foundations.

Red Hat claims to be the only company other than Microsoft capable of delivering a cloud stack consisting of an operating system, middleware and virtualization that’ll run a hybrid cloud environment.

It sees enterprises, cloud service providers, ISVs and Software-as-a-Service (SaaS) providers using its JBoss PaaS widgetry to take existing assets and develop new applications and deploy them to a range of public and private clouds, its interoperability vision again.

The company said Wednesday that its next-generation PaaS solution would simplify the development of simple new web applications as well as complex transactional enterprise applications and integrate them into an enterprise.

To advance the scheme it’ll have a reference architecture so existing apps can be re-purposed in the cloud.

The PaaS plan involves Red Hat’s prospective cloud engine for application lifecycle management and promises to let developers use their favorite frameworks and languages, stuff like Java EE, POJO, Spring, Seam, Struts, GWT, Groovy and Ruby. JBoss Developer Studio is supposed to have a bunch of Eclipse plug-ins to deploy applications into a JBoss platform instance in a cloud.

Red Hat expects JBoss cloud images to be available through a variety of public and private clouds including Red Hat Enterprise Linux, Red Hat Enterprise Virtualization, Amazon EC2 and Windows Hyper-V as well as others through its cloud engine.

Let’s remember too that VMware stole a lot of JBoss’ thunder with its acquisition of SpringSource. It converted the Spring Framework into its own cloudified PaaS solution for Java applications.

By the way, Dreamworks (Shrek) Animation is using Red Hat Cloud Foundations to build a “one of the world’s largest” private clouds for future films.

The easiest way to build a bridge between private and public clouds is to own both ends of the bridge, which is exactly what VMware is proposing to do with a scheme it calls vCloud Express.

VMware figures it’s got the internal private cloud covered with its vSphere widgetry.

Now it’s started building from the other side and has recruited a bunch of services providers to build external or public clouds that instantly connect with VMware-based widgetry, rapidly provisioning the stuff on-demand and asking users only to pay for what they use by the hour with a credit card à la Amazon, all managed by the same tools like, say, RightScale’s migration skills.

Poof! It’s nominally solved the interoperability and portable issues associated with clouds – though not the proprietary lock-in – and dodged the problem of making it necessary for an application to be written specifically for a particular cloud.

The external cloud widgetry is also supposed to be cheaper than Amazon – like cheaper by half – and support more operating systems though no more hypervisors.

The company has Terremark, the managed hoster where VMware has some money parked, BlueLock and Hosting.com in the US, Logica in EMEA and Melbourne IT in APAC beta testing its vCloud Express public cloud widgetry and their own services. Some will have production-grade SLAs; initially most are thinking prototyping and development.

VMware claims to have amassed a following of a thousand service providers in the last year, which would appear to put it ahead of the eight ball, but most of them don’t have an infrastructure-as-a-service platform, according to Forrester.

vCloud Express depends on a REST-based API for application vendors, service providers and enterprise IT that VMware means to have standardized.

Despite crippling losses, filing for Chapter 11, massive job cuts, the abrupt departure of its CEO and CFO, a search for somebody to buy it or merge with it, incapacitating debt, delisting threats and a patent suit, Spansion, the spun-off former AMD-Toshiba joint venture, has finally reached the jumping-off point it’s been aiming to get to for a couple of years now. It’s going to try to disrupt the industry by substituting NOR Flash for conventional server main memory.

This is not the NOR that goes into cell phones.

This is enterprise-grade NOR in a DIMM form factor, stuff that Google, Facebook and all the top sites are reportedly looking at.

It’s meant to solve the problem of today’s typical x86 servers not being able to support the giant datasets people are currently trying to force down their throats because of their limited DRAM capacity, a solution other than simply throwing more poorly utilized servers at the problem.

Spansion’s widgetry is called EcoRAM and it’s supposed to be way faster, cheaper and cooler than DRAM in read-intensive environments.

Forget write-intensive situations, they’re not NOR’s strong suite but Spansion claims an outrageously high 60x performance improvement – sweet Lord, 60x – on oil and gas modeling and visualization applications, for instance.

The Stanford Exploration Project, the industry-funded research consortium, used it to reduce the processing time on multi-terabyte datasets and was reportedly able to visualize half-terabyte datasets on typical x86 servers for the first time. Stanford said the data reorganization step in its processing flow was cut from 22 hours to 22 minutes.

EcoRAM is supposed to be the first technology to make near real-time analysis possible.

It’s also supposed to improve the performance of certain read-intensive Internet applications – like search and social networking – up to a whopping great 50x – to repeat 50x, numbers nobody’s used to – over tradition DRAM + hard drive server platforms.

EcoRAM expands main system memory eight-16 times the capacity of a typical x86 server and with as much as half-a-terabyte – that’s 512GB – of EcoRAM data doesn’t constantly have to shuffle back and forth between the hard disk and the DRAM. Instead large datasets are stored directly in the Flash memory, speeding up performance because the data is hugging the CPU.

And with maybe as much as an 8:1 consolidation rate, EcoRAM releases a considerable amount of data center real estate and saves an awful lot of TCO.

Then figure a 4GM DRAM DIMM needs 10W of power. Spansion says its 32GB EcoRAM DIMM only uses 10W of power.

The company claims a 75% reduction in energy costs over four years, a 75% reduction in footprint and a 45% reduction in CAPEX for an overall TCO reduction of 65%.

It figures 1,250 of its servers each fitted with 128GB of its widgetry can replace 5,000 traditional servers each fitted with 32GB of DRAM.

Spansion currently has two server OEMs, Virident Systems and Appro, exploiting the widgetry. Virident, a start-up that Spansion owns a piece of, is optimizing those two web favorites, MySQL and Memcached, with the widgetry (see separate story). And Appro is going after the HPC and oil and gas markets with two-socket and four-socket systems. Figure up to 256GB of EcoRAM in the first and 526GB in the second.

Spansion is also supposed to be working with other server OEMs applying EcoRAM to Hadoop clusters. Hadoop is the increasing popular open source framework used in Internet server clusters to reorganize and analyze huge and fast-changing datasets.

EcoRAM is now available in capacities of 32GB per DIMM. The widgetry currently works only on AMD boxes.

Jan Silverman, the VP of Spansion’s Server & Storage Business Unit, said that when Spansion started on its adventure “Intel was not a good choice.” Intel still had that old-fashioned bus and Opteron’s HyperTransport widgetry brought something to the party. Since Intel is currently on top that’s going to have to change for Spansion to make a real impression.

EcoRAM is compatible with a number of 64-bit Linux applications. To support Windows, Microsoft is going to have to be persuaded to make a few changes in its operating system, Silverman said.

Since it’s best at read-intensive situations, Spansion figures that makes it ripe for analytics, bioinformatics, business intelligence and the government besides the Internet and oil and gas.

The widgetry supports read latencies in the hundreds of nanoseconds, which makes it competitive with DRAM latencies, 10,000 times faster than hard drives and maybe as much as 100x times faster that state-of-the-art solid state drives.

Spansion says read bandwidth reaches up to 2.2 GB/s, which is 20x your typical 100 MB/s enterprise-class hard disks. Write performance is supposed to be comparable to high-speed enterprise-class hard disks, which are in the hundreds of MB/s.

The widgetry includes an accelerator – which lets servers address EcoRAM like it was DRAM – and Linux driver software. Spansion is currently using CentOS. It’s waiting for certification on Red Hat and SUSE.

The company says its EcoRAM architecture – which still calls for DRAM DIMMs because the server’s operating system and apps run faster out of DRAM – was designed to leverage upcoming high-speed connectivity solutions from both Intel and AMD.

Silverman said OEMs should figure the price of a 32GB of EcoRAM coming in under the going rate for an 8GB quad-ranked ECC DRAM on a per gigabyte basis. Spansion’s goal is to get it below 4GB.

According to IDC, the Internet and analytics market for x86 servers will reach $5 billion this year.