Cloud computing

What I learned at the Red Hat Summit

To pay homage to the many schoolchildren around the country in the process of relaying what they did on their summer vacation, I thought I’d share what I did last week at the Red Hat Summit with you.

To start with, you can watch most of the Red Hat Summit keynotes and some of the sessions via the video page from Red Hat. This will give you a taste of the large ballroom sessions, and what angle Red Hat’s executives and guest speakers were promoting. Sadly, DreamWork’s Derek Chan’s presentation on how the animation giant is using Linux and Red Hat to render massive 3-D movies isn’t available. The big themes overall were the power of collaboration through open source, interoperability, cloud computing, and meeting business needs faster.

RHEL 5.4 and Satellite 5.3 released, but where is RHEV?
The big news at the Summit was the open availability of Red Hat Enterprise Linux (RHEL) 5.4. But what was missing was the package of virtualization management tools, referred to as Red Hat Enterprise Virtualization (RHEV) for servers and for desktops that are slated to be released “later this year.” The release of Red Hat Satellite 5.3 was also announced, and how the new RHEV tools and Satellite will play together remains to be seen. Brian Stevens, CTO and Vice President of Engineering, said that the two are largely complementary and Red Hat will focus on integrating functionality going forward, providing a seamless experience for the end-user. We’ll look forward to seeing all the features of the RHEV release, which Naveen Thadani, Red Hat’s Senior Director of Virtualization explained would be best suited for those who want turn-key virtualization management.

Red Hat Catalyst Partners
The Red Hat Catalyst partner program was launched on the first day of the event, and was seen by some as an effort to appease the unhappy ISV’s who along with some VARs have complained that Red Hat doesn’t “get” partner programs. But at the show the vendors I talked to relayed their satisfaction and enthusiasm for the new program and the promise it offers for codevelopment and packaging of ISV offerings. Some shared with me that in the meeting with partners, when complaints were aired, Red Hat execs pointed to the Catalyst program in their response. It is clear that the company is hoping this initiative will resolve some of the woes of the past – and the partners can do nothing more than be hopeful as well.

The exhibit hall featured many partners displaying their products and I got one to provide a quick demo of their software. Trusted Computer Solutions has created Security Blanket a Linux security product that locks down the operating system and automatically configures it to meet industry standard and customized security requirements. Red Hat Enterprise Linux 4 and 5 are supported (as well as CentOS 4 and 5, Oracle Enterprise Linux 4 and 5, Fedora 10 and Solaris 10).

Performance tuning and other lessons from the sessions
After a hectic day of press conferences and appointments with various representatives of companies in attendance, I had the chance on Thursday to sit down and listen to some of the session presentations. Red Hat has made some of the session slides available online, including the slides for the back-to-back performance tuning session.

With 131 slides, it is safe the say that Red Hat engineers John Shakshober and Larry Woodman were a bit ambitious, and they didn’t quite get through their deck in the session. But if you are really interested in learning more they provided a lot of good example tools and on slide 128 they provide a list of good resources to check out to learn more about performance tuning. I asked some attendees if they got out of the session what they had wanted. They said that unfortunately, for their high performance computing application that it wasn’t all that relevant, but they had learned some new information and tools that might be helpful. I asked if they couldn’t get Red Hat support to help them with their HPC performance questions and they told me that honestly, they “hadn’t had much luck with that.” I was a bit confused: if you pay for the support, yet you don’t get good help with the support, what’s the point? “Compliance.” So just curious – readers, have you had similar issues? Why do you pay for support licenses on your servers when Fedora is available with no fees?

I also sat in on “Unmatched Security is Manageable” by Spencer Shimko, senior security engineer, Tresys Technology, about using open source system management tools to configure, monitor, and update the security configuration of Linux systems. He covered the open vulnerability assessment language (OVAL), and OVALDI (OVAL + interpreter). I won’t go into too much detail, but we should have a tip on using the language in SearchEnterpriseLinux.com soon.

On Wednesday I sat in on a session on using iSNS to simplify iSCSI management presented by Shyam Iyer, a development engineer senior analyst from Dell and Mike Christie a software engineer at Red Hat. In the manage and secure “What’s Next” track, this presentation discussed how storage management can be simplified with a plug-and-play environment for iSCSI SANs. Iyer discussed the new features being built into the open source storage name service. He also answered a couple questions from attendees regarding how iSNS and iSCSI will fit in a virtualized environment and the differences between it and DNS. The iSCSI network will fit on virtualized guests the same as it would on a hardware environment, explained Iyer. And in the same way that you can have a primary and secondary DNS, you can have a primary and secondary iSNS.

Overall, the summit provided a great opportunity to see and hear about what businesses are doing with RHEL, and learn more about what the future will bring. Our news contributor, Pam Derringer, has written specifically about some of the examples and case studies that we hope you find interesting and useful. If you feel like you missed out this year, next year’s summit will be in Boston, Mass., and perhaps it won’t be scheduled opposite VMworld.

JBoss and rPath demonstrate choice and customization leading in Linux

This post was written by News Contributor, Pam Derringer.

Sometimes writing or reading tech stories about entirely different products can uncover new trends and ideas. I was intrigued with the JBoss Open Choice Java Application story last week because Red Hat officials said its new framework would enable customers to add specific functionality like clustering, caching, messaging and security in “microcontainers” or do without it, according to their needs.

In addition, for the first time, JBoss customers would be able to choose between three levels of application complexity, and move from one to another within the same management framework. By offering customers choice, JBoss will create major disruption in the Java application world, according to Aaron Darcy, JBoss product line director. Darcy added that customers are moving away from the bloated, one-size-fits-all applications that must contain code capable of doing everything, and opting for slimmer versions more tailored to their needs.

Darcy’s words struck a chord because last fall and again more recently, I wrote about rPath, a startup birthed by former Red Hat staffers who saw scaling problems with large deployments first-hand, and decided to solve the problem by upending the traditional all-purpose horizontal stack, creating a vertical, app-centric Linux-based stack with only the elements that a customized application needs to run. Obviously, rPath-constructed applications, too, are a lot slimmer than all-purpose counterparts. They are also a lot easier to maintain and update, saving time for IT operations staffs say rPath folks.

Although their approaches are somewhat different, the two companies are both reacting, it seems to me, to similar needs for more choice and customization rather than a hefty one-size-fits-all, “that’s all we offer” approach. Is this a trend that will reshape the software industry as we know it? What do you think?

Ubuntu 9.04 release focus on user experience and data center needs

Canonical, the sponsor of Ubuntu, today announced the simultaneous release of Ubuntu 9.04 Server Edition and the Ubuntu 9.04 Desktop Edition, available for download on Thursday, April 23, 2009, and the Ubuntu 9.04 Netbook Remix, available on Thursday April 30, 2009.

With the new 9.04 server edition, Canonical has worked to extend the range of enabled servers, with 45 of the most popular mid-range servers from IBM, Dell and Sun and HP tested in the Canonical labs.
Ubuntu 9.04 Server edition will preview Ubuntu Enterprise Cloud (UEC). Ubuntu is the first commercially-supported distribution to enable businesses to build cloud environments inside their firewalls. With Ubuntu 9.04 Server Edition, organizations can explore the benefits of cloud computing without the data or security issues associated with moving data to an external cloud provider. Following a successful beta program, Ubuntu Server Edition 9.04 will also be fully available on Amazon Elastic Compute Cloud (EC2).

Mark Shuttleworth, CEO, Canonical shared some of the server improvements in Ubuntu 9.04 Server edition include substantial improvements in some key applications for mail and other common infrastructure requirements.

“There has been an extension of work around suspend and resume of servers,” said Shuttleworth. “Amazon’s EC2 elastic computing meme will penetrate deeply into the enterprise. And organizations will want that same elastic computing internally, along with the power saving capability. The best method is suspending or resuming. Through effective use of elastic computing, we think we can greatly improve the energy savings in the data center.”

Shuttleworth referred to the concept of cloud computing as “the new hotness,” and says that Canonical has chosen to give it a very specific focus in this release. An image of Ubuntu 9.04 is now on EC2, so anyone interested in prototyping on Ubuntu 9.04 can fire it up on EC2. Shuttleworth shared that Canonical has a firm commitment to continue to release updates in the cloud. A description of other Ubuntu virtualization efforts and a more detailed report on the current position of Ubuntu in the data center and enterprise IT environment was published on SearchDataCenter.com at the beginning of April.

Ubuntu 9.04 Desktop Edition improves user experience
Mark Zimmerman, CTO, Canonical explained some of the new features in Ubuntu 9.04 Desktop Edition include a reduction in start-up time from 45 seconds to 25 seconds. According to Zimmerman, the release also includes an improved notification subsystem, which is the first in a series of design-led improvements.

“We are really working on improving the intrinsic experience of using Ubuntu on the desktop,” explained Zimmerman. “The notification subsystem has a standardized way of displaying [notices], that adds to the polished feel of the desktop.”

In addition, the desktop version of Ubuntu 9.04 features OpenOffice.org 3.0. This release of OpenOffice includes a lot of compatibility between Microsoft Office suite products that can make the user experience more seamless and easy, and wasn’t available at the last Ubuntu Desktop Edition release.

Open source to shine despite dim economy

In the final hours of 2008, Roger Burkhardt, president and CEO of Ingres, shared his predictions about the future of open source.

Here are summaries of his predictions:

1. As the economy worsens, adoption of open source software will increase .
2. Open source adoption will accelerate across the full infrastructure software stack and into applications.
3. SaaS and cloud computing solutions will grow and pull open source with it.
4. Mergers and acquisitions of open source vendors will continue.
5. Competition from open source will drive proprietary software vendors to take the first steps in changing their business models.
6. Open source stacks will emerge from strong market partnerships.
7. Systems integrators will guarantee first-year cost savings for migrations to open source.

For the most part, I’m with him. As budgets tighten, it only makes sense that “free” will flourish. We hear that consumers are only looking for sales these days, so why wouldn’t companies? With this increase generally, and as more acceptance and trust of open source occurs, it seems natural that it will extend through the stack and through more applications (e.g., Vyatta has been busy releasing three open source networking appliances in 2008). And perhaps the upside of a down job market is that some programmers, with free time on their hands, can finish projects they had on the back burner. In 2008, we heard a lot of buzz about the cloud, and this cost-conscious computing model pairs well with open source. If you buy the cloud hype, you can easily see the increased opportunity for open source adoption that it provides.

We know that open source gained ground in the enterprise, and we even saw major proprietary software vendors change their tune (if not their business models) — ceding some ground to open source devotees.

What are your predictions for open source? Share them below or make your case against Burkhardt’s divinations.

LAMP stack story overlooks impact of cloud, reader says

My recent story on the dimming of the LAMP stack sparked a thoughtful reader response from John Locke, the manager of Seattle-based Freelock Computing. The story concluded that while an all-open source stack is still a valid concept, there are many more open source options that LAMP (Linux, Apache, MySQL and Perl, Python, PHP) is largely irrelevant.  I made a single exception for Apache, the popular Web server.

Locke argued, however, that even Apache has a growing array of alternatives such as the Lighttpd Web server, the Apache FastCGI Web interface,  the Nginx proxy server and others.

But what undercuts the LAMP stack more than the advent of additional open source options is the emergence of cloud frameworks, Locke said.

Initially, cloud computing meant renting compute power on demand from the likes of Amazon Elastic Compute Cloud (EC2). This meant renting a host virtual machine, programming the top layer, adding libraries and then when it was all done, managing the host and the virtual application, Locke said.

The problem with this model is that data centers are responsible for scaling the application up or down in response to changing volume requirements, he said. To solve this problem, Google, as well as Microsoft’s recently announced Azure platform, go beyond computing-on demand and manage the entire process with frameworks. All you do is write the application code (yes, you still need the P in LAMP), put it atop an application framework, and the framework will scale the application up and down as needed. No further involvement required. No LAMP stack required either.

Two successful examples of cloud frameworks are Salesforce.com and Facebook, he said.

The downside of frameworks, however, are loss of control and potential vendor lock-in, Locke said. The risk is less with Amazon EC2 since its controls are far more limited, he said. When writing an application for a specific vendor’s framework, however, a customer can lose portability because the provisioning and scaling mechanisms are behind-the-scenes and the source code and licensing are not necessarily readily available, he said.

The biggest challenge to LAMP as well as the Java and .NET stacks, therefore, is not the growth of additional choices but the cloud frameworks which may make all the stacks irrelevant. While handing over management and control is convenient, it also has its downside: you have to live by someone else’s rules, Locke said. Just  like a condo or regulated housing community, you’ve delegated the work, but you’ve also lost your freedom. Time will tell if you’ve made a good bet.

Don’t want to go to the cloud? Cassatt says to build your own

You may be one of the data center administrators who’s heard the buzz about the cloud and has decided that it’s just hype. But the advantages of the cloud infrastructure are clear:

• Extremely low operating costs (as low as 1/10 traditional IT costs)
• Extremely high energy efficiency
• Extremely low levels of complexity
• High economies-of-scale
• Metered billing (based on use); transparent costs
• Other benefits of large data centers without owning capital

There are also a bundle of fears or disadvantages that are enough to keep many from jumping on board. Here are a few of Cassatt’s observations on what these are:

• Security: infrastructure sits outside the enterprise’s walls
• Service levels: Nascent model; agreements and levels unproven
• Performance
• Auditability and logging/traceability issues
• Potential for cloud platform architecture “lock in”
• Does not lower existing cost of installed capital or operations

In preparation for Cassatt’s internal cloud release this week, I talked to Ken Oestreich, director of product strategy, Steve Oberlin, chief scientist, and Jay Fry, VP of marketing at Cassatt. Oestreich explained how a product that the company had been developing for five years (initially referred to as its “utility computing product”) aligns with the cloud computing model. Essentially, Cassatt’s Active Response 5.2 allows you to turn your entire data center into a “cloud,” but without the disadvantages above.

“People who haven’t outsourced because of regulatory or security are not going to change,” said Oestreich. “They’re not going to a cloud.”

Seeing this opportunity, and realizing that many CIOs would love to take advantage of the efficiencies of the cloud but can’t afford the risk, Cassatt’s product allows users to get pretty close. Additionally, there are some advantages to the “internal cloud” model that include no platform-dependency issues and no “lock in” to an external cloud provider. Active Response 5.2 provides multi-platform support for Linux, Solaris, Windows and AIX; virtual machine support for VMware, Citrix (Xen) and Microsoft Hyper-V as adoption warrants; and networking support for Cisco, F5 and Force10.

In addition to Active Response 5.2, Cassatt has introduced its Active Profiling Service.

“Before you can embark on creating an internal cloud and merging application groups into pools that can share these resources, you have to know what you’ve got,” said Oberlin. “That’s what enables you to create a management strategy.”

Some consolidation planning software exists on the market, but Cassatt’s team thinks that it misses the mark and doesn’t provide users with all of the information they need. Cassatt points out that existing inventory tools don’t look at usage patterns, application dependencies or workload dynamics, and consolidation tools don’t consider workload management and server repurposing.

“A lot of companies today buy consolidation planning software if they’re doing virtualization. What this software doesn’t do is what all of our customers ask us about — provide a picture of the dynamics of the data center,” explained Oestreich. In order to manage a virtualized environment, it’s helpful to have an idea of “… which apps are quiescent and when, where the orphan servers are, where is virtualization appropriate and not, where is power management appropriate, and where is the internal cloud computing appropriate and not.”

Oberlin shared that the actual setup and implementation of the software can be rather rapid (a day or two). “The longest period of time is recording performance and utilization data to capture a reasonable business cycle to get a decent utilization profile of the applications over time,” said Oberlin.

The team envisions its internal cloud offering as something that users can gradually work into. I imagine dipping a toe in and then easing into the hot-tub and relaxing while the data center is efficiently managed.

1. Analyze infrastructure and opportunities using Active Profiling Service
2. Get started using policy management
3. Take advantage of the power-management infrastructure and achieve increased energy efficiency
4. Manage virtualization across multiple vendors, simplifying and automating virtual infrastructure
5. Implement application availability across platforms and achieve greater operational efficiencies
6. Implement resource repurposing across physical and virtual platforms and achieve greater capital efficiencies
7. Meter infrastructure use, regardless of physical or virtual

In a time of budget cuts and reduced staffing in the data center, there’s no question that improved efficiency in physical and virtual machine management is beneficial. This type of move can help any data center prepare for future increased utility costs and trim down on new equipment provisioning. And who knows, maybe one day you’ll consider joining a community cloud.