Digital Walkabout

The basics of Cloud Networking

As enterprises construct more complicated cloud computing environments, networking also becomes more complicated. The challenges of traditional networking only increase with private cloud, public cloud and hybrid cloud deployments. It's only natural that networking considerations for the cloud can have an IT admin's head spinning.

To get a handle on the obstacles IT admins can face, you must first understand the basics of cloud networking. Where are the pitfalls? How can you tackle network latency? The answers to these frequently asked questions can help IT pros grasp the changing networking issues in a modern IT infrastructure.

What is a cloud computing network, and how does it differ from traditional networking?
In the IT industry, a network is a group of points that connect via communication paths, and networking is the use of that network, whether is the design, construction or everyday use.

When you introduce cloud computing into your IT environment, you're not only adding another point in the network, but you're adding network complications as well. A cloud computing network has to enable the movement of resources across endpoints, but to truly capitalize on the benefits of cloud, you need access to those resources from anywhere. And in a complicated IT environment using multiple cloud providers, the network allows for resource movement across private clouds, public clouds and hybrid clouds, as well as from a cloud installment back to the on-premises data center.

What effects can networking have on a cloud environment? How can I manage these effects?
Network latency could be a cloud killer. Cloud offers enterprises the benefits of scalability and flexibility, but without enough network capacity, these benefits are negligible.

Some companies make the major mistake of not taking into account their networking resources when constructing a cloud presence. When shifting workloads off-premises to take advantage of cloud computing's benefits, you burden the network between the cloud and your data center. And when network capacity between the cloud and the LAN is overburdened, applications and services suffer from poor performance.

To mitigate performance issues, IT admins should plan ahead for their cloud project, making sure network demands are part of the conversation. Once you launch your cloud project or migrate virtual machines to the cloud, network management tools and proper monitoring can help you keep tabs on your networking usage.

What can cause network latency problems in a private cloud?
Avoiding bottlenecks and network latency should be at the forefront of a cloud admin's monitoring plan. To avoid performance issues, look at often-overlooked network latency factors that could hurt a private cloud.

While your Internet connection and usage patterns could be more easily monitored as part of your overall network monitoring strategy, provider-to-provider networking could be a larger hurdle to conquer. Using multiple cloud providers to reduce risk means companies are hoping for full-provider high availability, which could clog traffic and cause networking issues.

How does networking play into migrating an application to the cloud?
There are obstacles in any application migration project, and not to be overlooked are network configuration challenges. After you've moved VMs and migrated data, either from on-premises to the cloud or from one cloud provider to another, you have to reconfigure your network. IT admins can face issues with domain name services (DNS) and internal configurations, such as problems with load balancers and database clients, which need to configure to a new IP address.

With the rise of BYOD and the use of mobile devices in the enterprise, what network changes do I need to make to support these devices?
Modern enterprises have employees using iPads, Android devices, BlackBerrys and other consumer devices. But are modern enterprises prepared to support these devices? Although they remain outside the corporate network and access applications via the cloud, mobile devices can negatively affect network performance and create numerous security issues. In the bring your own device (BYOD) era, companies need to implement mobile device policies to protect network security and maintain bandwidth health.

Primarily, it's important to set up gateways to separate the corporate network from any consumer devices. Other security efforts should include the use of firewalls, anti-malware software, data-removal policies and the implementation of strong passwords. But this is only the bare minimum; network security issues are just beginning to be broached in the BYOD era.

Delivering General Services in the Cloud

The service delivery platform (SDP) was born out of the need to rapidly create advanced multimedia services: IPTV, mobile video, gaming, location-based services. But is there a role for SDPs in the cloud? Potentially, but they can't be service delivery platforms as we know them today. To remain relevant to cloud providers, the service delivery platform must evolve.

Decades ago, the primary service of networks was voice communication, and the features of that service were embedded in the central-office and tandem switches. As network services evolved and became more software-based, this monolithic model was forced to change; even before the advent of the Internet, standards groups were working on how to define the role of computer-hosted intelligence in network services. These circumstances gave rise to the service delivery platform.

The roots of the service delivery platform lay in traditional voice services, and today those roots are most evident in the role that SDPs play in mobile voice services, functioning as a host for components of the IP Multimedia Subsystem (IMS). The future of SDPs, however, will likely be tied to two next-generation technologies: over-the-top (OTT) content delivery and cloud computing. The combination of OTT and cloud forms a union that can be called XaaS -- meaning Anything as a Service or Everything as a Service -- reveals the direction service delivery platforms must take.

Why XaaS needs a service delivery platform
The XaaS model presumes that providers will increasingly deliver next-generation services to mobile devices. The data fed to these devices will come from a variety of sources, and the compute resources processing it would likely be distributed as well. Supporting this model with simple, best-effort Internet connectivity is likely to frustrate customers, however, since the failure of any component to play its role will cause the service to fail overall.

Reliability is paramount to supporting XaaS applications -- from the network up to the application components -- as the service must keep users happy and be profitable for the provider. Given these requirements, the service delivery platform may be an ideal platform for XaaS. Service delivery platforms are inherently designed for a higher standard of availability than traditional servers, and they are also built to meet the high-reliability requirements of signaling in telecom services.

Most SDPs use open platforms like Linux, UNIX or a derivative operating system, and in theory, most could be equipped with cloud software and support Web-compatible interfaces. The only thing that today's service delivery platforms lack is an explicit conception of Web services that's on par with their explicit conception of session-based services like Voice over IP (VoIP) and IMS. It seems unlikely that IMS could generate a competitive platform for Web services, and SDPs must significantly enhance basic server or cloud server capabilities if they are to be justified as a competitive option to ordinary servers.

SDPs in XaaS: App stores, certification or orchestration?
To date, the market has identified three applications for which the merits of SDPs are considered valuable: the development of app stores, certification of features and service elements and orchestration of services through composition of elements.

App stores are interesting to service providers because they address a market related to their mobile device deployment and they link to their billing processes. While it would be possible to host app stores on standard servers, as Google and Apple do, the network operator faces a higher standard of reliability, and that is enough to justify a service delivery platform. The issue is that app stores are simply an application -- not the foundation for a service architecture.

Certification is a potentially more compelling step in the evolution of the service delivery platform to an XaaS platform. Certification platforms are responsible for managing the elements of a service network, including the software, servers and databases that fall outside the scope of traditional network management and network operations center tools. Service delivery platforms could be deployed in parallel with traditional servers and could be used to ensure these devices were available and functional, thus overseeing service operation at a high level. But even in this mission, the SDP is outside the service itself -- ensuring resources are available, but not participating.

It's likely orchestration will cement the role of service delivery platforms in XaaS. In online services, orchestration is the composition of a service or experience through an orderly linkage of service components and the movement of requests and responses among those components. Unlike certification, which focuses on verification of availability, orchestration actually creates the service or experience, which means it's critical to building what users see -- and buy.

Orchestration also addresses the underlying challenges of building complex, componentized applications that XaaS creates in the first place. An SDP-orchestrated service could protect against the failure of a single component by simply monitoring it to ensure that requests sent to that component generate responses, or by employing a failover plan if the requests don't respond properly. This means that more traditional Web servers could participate in XaaS without creating the avalanche of failure risk that typically accompanies deployment of multi-part services.

When used in orchestration, service delivery platforms can act as hosts for the service components whose operation is most critical to XaaS and where it's necessary to integrate components with a high-availability network or with OSS/BSS elements. Orchestration is the future of the service delivery platform, and most SDP vendors now acknowledge that in their marketing materials.

SDPs for orchestration: Many obstacles ahead
The challenge that remains is organizing and standardizing orchestration itself -- something that's been attempted for several years without much success. Most SDP vendors offer some orchestration tools, but many of these platforms are still complex and expensive, making them less competitive in the real world of XaaS.

Meanwhile, work on orchestration in the more general cloud and server markets is progressing, particularly through work around DevOps and application integration. Some vendors have also taken the lead on this, including Alcatel-Lucent, Cisco Systems, Dell, Hewlett-Packard, IBM, Microsoft Corp and Oracle. The service delivery platform has to move to the front of these activities if it's to remain relevant.

Service delivery platforms may also have to resolve the conflicts associated with mobile Web services outside of IMS. Many vendors in the IMS world have resisted efforts to define "services" except as a series of sessions -- a model the OTT giants that drive the XaaS market today have rejected. If this debate interferes with the progress of orchestration support on SDPs, it could relegate the service delivery platform to softswitch-like applications and reduce its relevance in the XaaS-driven future.

Get the Chaos Monkey on your back

Netflix's "Chaos Monkey," a cloud-testing technology, ensures that the company's service remains up and running on Amazon Web Services -- even during outages that affect parts of the public cloud infrastructure that its service runs on.

 

The company recently released the Chaos Monkey code to the open source community under the Apache 2.0 license via GitHub, a move that's being hailed by cloud experts and other observers -- and one that portends the arrival of a veritable army of testing tools over time based on the concept.

 

"I can only imagine that Chaos Monkey becoming available for a more general public will result in nothing but better results for AWS customers," said Kyle Hilgendorf, principal research analyst at Gartner Inc. in Stamford, Conn.

 

The Netflix Chaos Monkey tool allows cloud customers to proactively launch attack code against their own infrastructure to cause failures. These failures give engineers the ability to fix potential problems before they occur on their own.

 

Some cloud service providers said they have already encountered cases of their customers using Chaos Monkey to help bulletproof their AWS-based infrastructure and test code for hardening their apps.

 

That may be because of the similarities between streaming massive amounts of data interactively in the cloud -- as Netflix does -- and handling other workloads that are appropriate to the same model, such as social computing and collaboration in a corporate cloud environment.

 

"The short form is that tools like this help folks design more robust infrastructure in the public cloud," said Matthew Gerber, CEO of IT-Lifeline Inc., a disaster recovery provider built on AWS based in Liberty Lake, Wash.

 

Planet of the Apps

There is also value in Chaos Monkey for cloud application development.

 

"This will be a great tool for developers to use to aid in the construction of resilient Web applications [because] when a website is down, a company is losing money or trust," said Colin Dean, president of the Pittsburgh LAN Coalition, a group that organizes video gaming events and uses AWS.

 

In fact, Netflix says it has an entire "Simian Army" of testing applications in work, including a "Latency Monkey" that creates latencies that simulate service degradation, a "Conformity Monkey" that shuts down instances that don't conform to best practices, a "Janitor Monkey" that cleans up by releasing unused resources, a "Chaos Gorilla" that simulates outages of entire Amazon Availability Zones, and others.

 

Netflix said that since no single component can guarantee 100% uptime, its goal was to design a cloud architecture where the individual components could fail without making the entire system unavailable.

 

Chaos Monkey has terminated some 65,000 instances running in production and test environments over the past year, according to a Netflix blog post.

 

Additionally, Netflix said it is queuing other monkeys for release, likely beginning with Janitor Monkey.

 

"Any kind of toolset that helps customers build more reliable infrastructure is a good thing," Gerber said.

 

How Chaos Monkey works

The testing methodology is based on the idea that cloud service disruptors create havoc by causing arbitrary outages during business hours while IT staff are available to handle those situations. It proactively breaks running cloud components to expose weak spots so they can be fixed in advance of an actual outage.

 

Hilgendorf said that Chaos Monkey does not attempt to maliciously affect AWS infrastructure. It simply randomly stops and kills customers' own Amazon Machine Images or storage volumes so that customers can be assured that their design can handle random failures. 

 

How do cloud providers feel about customers destructively testing their cloud apps? Neither Amazon nor Microsoft responded to requests for comment about the impact on their respective cloud platforms.

 

Netflix said Chaos Monkey could be made to run on other cloud platforms.

 

RedHat's OpenStack move raises eyebrows

Enterprise IT pros interested in OpenStack got some new options this week as a preview of Red Hat's distribution and a new installation package from Rackspace became available. 

Experts see these releases as steps in the right direction but also say that some basic elements of the software need brushing up for OpenStack cloud products to see widespread adoption in enterprise private clouds. 

Many in the enterprise are waiting and watching OpenStack closely, according to Kyle Basset, principal consultant for Toronto-based independent consulting firm Archestration, but few have adopted it yet. 

Enterprises already familiar with Red Hat Enterprise Linux and Red Hat Enterprise Virtualization are more likely to trust Red Hat's distribution of OpenStack, but "getting a lot of these enterprises to convert over to a fully open source platform that's brand new is going to be a tough challenge," Basset said.

Public cloud service providers also say there's work to be done before they can roll out their cloud services on OpenStack. "OpenStack is still in really heavy development," said Carl Perry, cloud architect for Brea, Calif.-based hosting provider DreamHost.

Improvements ahead with OpenStack Folsom
DreamHost anticipates the next release of OpenStack, dubbed Folsom for its network virtualization layer, Quantum. It also has a storage integration element called Cinder, which allows OpenStack's Nova servers to boot directly from networked storage. 

The Keystone identity access management system and Horizon Web portal, together called Essex, are two other key components of the current release that need improvement, said James Staten, analyst with Forrester Research.

These components caused some friction between the OpenStack community and some of the third parties that deliver OpenStack because they were initially seen as part of the value the third parties could add to their individual distributions, Staten said. But the community eventually won control. 

With the debate settled, the Folsom release will have the first commercially ready versions of Keystone and Horizon.

Red Hat's distribution, based on the current Essex release of OpenStack, is still an unsupported preview. Red Hat also implies that OpenStack will be commercially ready with the Folsom release and pledges a fully supported distribution based on that release in 2013. 

Rackspace advances Nova server
Meanwhile, one of OpenStack's original champions, Rackspace, is working to make the Nova server element of OpenStack easier to install with a new package that combines Opscode's Chef Server configuration automation software with Horizon, Keystone, the Glance Image Library, the Ubuntu operating system and KVM hypervisor.

The installer will set up a relatively modestly sized cluster -- Rackspace has tested the installation package on up to 20 nodes -- automatically, eliminating the headaches commonly associated with setting up Nova.

Today, setup is commonly done through scripts, "which is great if you're a developer, but not good if you're an end user or an administrator," said DreamHost's Perry.

 

 

 

Can vCloud test drive lure enterprises to public cloud?

IT pros interested in cloud computing have long been sold a value proposition for public cloud as an elastic, pay-per-use resource that’s as simple to spin up as entering a credit card number on a website. As it turns out, evaluating the services of VMware’s vCloud partners isn’t quite that simple, prompting the company to disclose this week that it will soon launch its own test-drive service for public cloud. The beta version of VMware’s new Service Evaluation will be launched in the next couple of weeks. It will allow customers to spin up publicly hosted instances of vCloud Director through the vcloud.vmware.com portal, and run them at a rate of four cents an hour per 1 GB RAM VM. Some existing VMware customers balked at the fact that VMware is charging for a beta of an evaluation service, however small the price may be. “This will not persuade me to try VMware's offering,” said Rod Gabriel, IT infrastructure engineer at UFS, a Wisconsin financial services company. Gabriel’s company isn’t planning to use public cloud, so the cost to test VMware’s cloud wouldn’t be approved. “However, seeing what Amazon has to offer for free, if I get some time I may just try out their offering,” he said. Why not make the evaluation service free, especially a beta form? After all, it’s possible to get a year’s worth of a bevy of products on competitor Amazon Web Services (AWS) Free Usage Tier. “There’s a nominal charge associated with it to cover the cost and to ensure that there’s somewhat of a bar for customers, so that there are serious evaluators,” said Joe Andrews, director of product marketing for the vCloud service provider program. Some VMware customers don’t mind the cost. “While the Amazon free service is cool, if I am looking into vCloud, it is for the advanced vCloud and vSphere functionality that Amazon does not offer,” said Bill Hill, a senior engineer working in the public sector. Dodging service provider test-drive runaround In contrast to the new vCloud Service Evaluation, many of VMware’s vCloud platform partners require a sales engagement or a minimum trial period for a test drive. At vCloud partner AT&T, to test out the vCloud service, users must first establish a billing point of contact. From there, this contact must obtain and enter a Company Permission Code before the user can get started. At BlueLock, users must apply for an evaluation program. BitBrains’ website offers access to a test environment for two weeks free of charge only after contact with a sales representative. It’s the same at Carrenza and Earthlink. iLand requires a minimum 30 day trial; NetPlan has the shortest trial period available among vCloud partners at 7 days. ZX Elements comes in second in that category at two weeks. There may be exceptions, as there are 145 vCloud services representing 28 countries around the world, but no vCloud partner that is listed in the “Take a Test Drive” section of the vCloud portal discloses an upfront price per hour or allows access using just a credit card. “Many of these providers have been used to selling in the enterprise sales fashion where it’s an outbound sales call, [and] you’re talking about having to take weeks in some cases to actually get the customer to try out your service and start to realize benefits,” said VMware’s Andrews. Customers “want to go online, with credit card authorization, start to use a service and be up and running with it, and do their proof-of-concept and demo of it that way,” Andrews said. Onerous signs for vCloud adoption Meanwhile, industry observers say VMware’s vCloud products, and especially its public vCloud program, have not seen the kind of adoption the company expected. VMware’s core audience has long been IT pros in large enterprise shops, such as Ed Czerwin, a systems engineer for a large medical devices company near Zurich, Switzerland. He cites compliance and performance reasons for staying away from clouds -- both public and private for the time being. Then there’s also the potential expense of remaining a VMware loyalist in the cloud. "VMware still is currently geared toward the enterprise, and the prices of [its] software to host low-tier types of applications that would go into a cloud maybe aren't just spot on yet," Czerwin said. Analysts say vCloud has had a number of hurdles to negotiate. One being the challenge of getting service providers to offer vCloud with the full vCloud API exposed, said James Staten, analyst for Forrester Research. Of the providers who have exposed the full API to make workloads more portable across clouds, most have hedged their bets by offering some other cloud option, such as Dell’s version of OpenStack or Verizon’s proprietary offering, Staten said. "The other problem [VMware’s] been having has been getting enterprises to actually set up a vCloud,” Staten added. “There hasn’t been a significant pull by enterprises saying ‘give me a vCloud! I want to set up a private cloud based around this.'" [VMware] ended up having vCloud sort of bundled in to a lot of enterprise license agreement renewals, and so it’s sitting on a shelf, and no one’s opening it," he added. No Zephyrs on the horizon yet Despite rumors that VMware plans to offer its own Infrastructure as a Service (IaaS) and confirmation from sources that VMware is indeed building something large in a colocation facility in Las Vegas, it doesn't appear that the vCloud Director evaluation is a trial balloon (or dirigible, as the case may be) for such a project. VMware’s Andrews insisted that the vCloud Service Evaluation is not being hosted on any VMware-owned infrastructure. Offering IaaS "is not the intent here," he said. VMware declined to comment on Project Zephyr rumors.ζ

The Lifecycle of the Cloud Adoption Strategy

This excerpt from Chapter 3 of Is Your Company Ready for Cloud? Choosing the Best Cloud Adoption Strategy for Your Business, by Pamela Isom and Kerrie Holley, examines how an enterprise should strategically plan its cloud adoption, taking into consideration specific business goals, cloud standards, best practices and current and future use. Core areas of focuses in this chapter, "The life cycle of your enterprise cloud adoption strategy," include: Initial cloud planning Enterprise capabilities and cloud vision Target architecture and cloud enablers Gap analysis and transition planning Implementation planning Governance The significance of service-oriented architecture (SOA) Cloud standards Enterprise capabilities and cloud vision It is essential that you understand your enterprise goals and objectives. You need this information so that you establish a realistic cloud adoption strategy, and just as with enterprise architecture (EA) you want alignment between your enterprise vision and your strategic vision for cloud adoption so that activities and outcomes add value to your organization. Specifically, you want to understand your enterprise goals and objectives in the following areas: New capabilities that the business needs to bring to the market in the next 12 months Services or applications currently constrained due to time to market needs Opportunities to preserve capital or avoidance of operating expense Requirements or needs to provide ubiquitous access to consumers, customers, or devices Needs for improved scalability Factoring in your understanding of enterprise goals and needs into the formation of your cloud computing vision provides the business context to make cloud computing transformational for your business. That is, the business use of technology changes as technology related business risk is identified and managed; and IT queues get eliminated as a barrier for creating new business capabilities. All of this requires you to plan for governing your cloud adoption strategy as it ensures appropriate development with proper controls for progression with business continuity, and upon execution governs implementations so that cloud projects are in alignment with your strategy. The introduction of any new technology or technology paradigm shift creates pressure on the organization. Previously, enterprise architecture may have evolved solely to address internal provided technology solutions, and now it extends to address unique aspects of cloud computing in improving agility, enabling new business models or simply preserving capital investments. Questions such as the following must be answered during this phase of the life cycle: Is your business and IT aligned? If not, can cloud computing facilitate this convergence between business and IT? Do your applications and IT architecture support your changing market needs? Is your IT environment inflexible? How does your cloud adoption strategy support strategic imperatives -- that is, your company's vision? How is your company performing today -- that is, what's not working that would make a difference in business performance? Does cloud computing make your processes more effective and more efficient? What cloud business adoption patterns are suitable for your company? Who is accountable for making cloud adoption decisions? What is your organizational readiness capacity for cloud adoption? How will you ensure that your cloud adoption strategy is realized? While you might have started preparation of your business case in the previous step, it is most likely that stakeholders require additional information to sustain initial approvals. In this second phase, you focus on understanding your overarching ability to undertake or extend business operations to support private, public, hybrid and community clouds. Approaches for obtaining information include conducting visioning workshops and capability assessments. Pilots and/or proofs of concepts are optional, but highly recommended. Key activities in the enterprise capabilities and cloud vision include: Develop your cloud decision model: Developing the cloud decision model is described in Chapter 4: "Identifying Cloud Candidates." This model provides the use cases for cloud adoption based on strategic goals of the business. This model is essential to understand the future state made possible for your enterprise with cloud adoption. This decision model is grounded in the reality of what is possible coupled with the optimism of a future state. Develop your cloud business case and ROI inputs: The business case and corresponding return on investment (ROI) model should look at a comprehensive cloud adoption versus ad hoc adoption. It's not easy to do ROI analysis as most companies use intuition or guesswork. You might also easily spend months or a year doing ROI analysis necessary for business case development. Using the following approach can help: Create a benefit value tree where when selecting benefits (such as those described in Chapter 1: "Business Value of a Cloud Adoption Strategy") to distill the value drivers for cloud adoption such as capital preservation, increased time to market, new business opportunities, workforce transformation, accelerated time to market, and so on. Scenarios or use cases should be created for each value proposition. Identify the applicable cost scenario for each use case. Calculate the initial, simple return. Assess and select the cost scenario for the second and subsequent implementations. Keeping the benefits constant, calculate the returns for the second and subsequent implementations. Assess your enterprise's cloud adoption maturity level: Understanding shortfalls between your target state, your cloud vision, and the current state is essential for successful planning and realization of your cloud adoption strategy. Using the traditional CMMI (capability maturity model integration) thinking, organizations can easily determine whether they are in the formative stages or optimized stages of cloud adoption maturity. Figure 3.2 depicts this model. Figure 3.2 This determines what level of assistance is required to move faster or whether the current pace is sufficient. You should plan for improvements and incorporate this input into your roadmap as a natural outcome of this assessment. Conduct vendor selections: A vendor selection process is necessary to select the partners and vendors for enablement of your cloud adoption strategy. A team should be assembled with a vested interest in the cloud selection process. The first task that the vendor selection team needs to accomplish is to define, in writing, the service, capability, and features required from a partner. This gets further elaborated into a defined set of business and technical needs or requirements. Also, define the vendor requirements. Finally, publish your document to the areas relevant to this vendor selection process and seek both internal stakeholders and vendor inputs. Now you are ready to select vendors. Figure 3.3 Determine areas to assess to fully understand your capabilities to succeed with cloud adoption: Figure 3.3 illustrates the cloud capability assessment areas. Customer and cloud market insights provide clarity on the art of what is possible and how other companies are both innovating and differentiating themselves with cloud computing. Areas where business innovation can occur must be assessed and understood. Are your EA and/or SOA capability where it needs to be for the cloud adoption strategy to be successful? Assessing yourself in these areas provides a baseline for improvement and successful realization of your cloud adoption strategy. Outline your enterprise cloud governance strategy: Governance will be mandatory for achieving the goals of your cloud adoption strategy. This includes defining roles and responsibilities for stakeholders essential to cloud adoption. Your EA governance processes should be utilized, as EA will be a critical success factor in accelerating cloud adoption. Your governance should include a governing body that can create and manage the following aspects of a cloud computing environment: Policies Procedures Organizational effectiveness Accountability and reporting Communications Standards adaptation This excerpt is from the book, Is Your Company Ready for Cloud? Choosing the Best Cloud Adoption Strategy for Your Business, by Pamela Isom & Kerrie Holley, published by Pearson/IBM Press, July 2012, ISBN 0132599848. For more information, please visit the publisher site www.ibmpressbooks.com/title/0132599848

Cloud breathes new life into SOA

Enterprises with expanding plans for cloud computing, whether public, private or hybrid cloud, are increasingly interested in the intersection of cloud and SOA. In order for software to be useful in the cloud, the cloud and SOA, or service-oriented architecture, need to be compatible. For a growing number of businesses, though, the cloud is being recognized as a proactive driver to SOA when, in fact, SOA is a critical point of support for expanding the use of cloud in the enterprise.

SOA has two goals: componentization and consistency of exposure. SOA builds functional elements that are exposed as "services" through application programming interfaces (APIs). These elements then compose applications, which create SOA's dual benefits of reusing software components to improve development efficiency.

To create an application, a set of components is "threaded" into a workflow, often using a workflow "engine" or service bus software element. The workflow for a given application is directed to the correct component through a directory function, often called the Universal Description, Discovery and Integration (UDDI) in most SOA standards. When an application component is installed, a UDDI entry allows the application workflow to locate the component. This is where the cloud and SOA intersect.

Any time an application or application component is assigned to a flexible resource in any resource pool -- including the cloud -- it's associated with an address. And that address must be published to other components for the software to integrate into a company's overall IT process. Because SOA provides a way to locate components, that mechanism can be used to record what happens when an application runs in the cloud. In most cases, this mechanism will allow a company to deploy an application in the cloud and register its location, making the app accessible to the user. Resolving other addressing issues that involve URLs also requires DNS updating.

Short-term hybrid cloud and SOA concerns

The relationship between SOA and a hybrid cloud environment has its benefits, but it also has downfalls. One issue is the potential for performance problems when the application workflow moves across the public-private cloud boundary. In normal SOA applications that run within the data center, the data center network can maintain fairly efficient workflows across component boundaries. Having those workflows move data over a WAN to the cloud, however, could introduce delays, packet loss and, in some cases, exposure to security problems.

The component registration process for SOA apps in a hybrid cloud also has its pros and cons. On the plus side, you can use the public cloud to host a component that can no longer run on-premises because of a system failure. This creates a failover option for the application. If the application and the workflow or system bus processes support the use of multiple component instances, you can also manage that through SOA registration.

However, hosting a component on the public cloud is transparent to the user and IT -- unless the UDDI is examined. But doing so could expose the end user to public cloud usage charges if the component or application isn't brought back on-premises when the system is restored. Part of any SOA management procedure for hybrid cloud applications should include steps to ensure public hosting is used only when necessary.

Additionally, because of the "service" nature of SOA software, applications are accessed either through a graphical user interface (GUI) or through APIs and a third-party GUI tool. When using SOA with the cloud, it's important that the GUI supports the mechanism used to address the application. In most cases, this would be the UDDI, the DNS or both. It's the cloud user's responsibility to ensure associated directories are properly updated, which means the directories must be accessible from both the data center and the public cloud.

In the long run, cloud and SOA are a fit

Highly componentized applications with elements that register automatically upon loading fit perfectly into the users' vision of an elastic cloud resource pool. They also could facilitate load balancing and failover between private cloud elements or between a private cloud and the standard data center. In fact, many argue that in order to achieve all the benefits of cloud architecture -- a plug-and-play, fully elastic, self-provisioning, application execution framework -- you need SOA software.

With the industry trending toward the adoption of SOA for complex software products, future applications will likely become more SOA-compliant. And this makes them perfect candidates for a flexible, elastic hybrid cloud.

Service providers already see the value of the cloud and SOA link. One major European carrier that offers cloud services listed SOA expertise as a requirement for chief technology officer (CTO) status. Enterprises agree, as they begin to embrace a private cloud model that focuses more on creating a flexible framework that allows you to blend private IT with hosted public cloud services. Ten years after its launch, SOA may be destined to succeed -- in the cloud.

 

 

 

 

 

Binary.js

BinaryJS is bidrectional realtime binary data with binary websockets

A year ago if someone asked you "how do I stream binary data / audio / video / files to Javascript?" the answer would have been "Flash" or "no" (or "Java applets")

The BinaryJS framework employs websockets to send, stream, and pipe binary data bidirectionally between browser javascript and Node.js.

• BinaryPack serialization format is fast, binary, and JSON-type compatible. Data stays binary end to end
• Automatically chunks large binary data buffers for streaming performance
• Send multiple streams of data concurrently over multiplexed websocket connection
• API implements Node.js read/write Streams. You can pipe any stream into BinaryJS streams (and vice-versa)
• "pause," "resume," and "end" as in the Streams API

BinaryJS presently supports Chrome 15+, Firefox 11+, Internet Explorer 10, and Safari nightly builds.

View the Hello World example.

 

 

Rackspace releases OpenStack for production ... finally

OpenStack made headlines this week after one of its co-creators, Rackspace, has begun offering public cloud services based on the open source cloud software stack, though it’s unsure how far into the market OpenStack’s “DIY” appeal will penetrate.

So far, public cloud service providers with the development time and brainpower to tinker with software have praised OpenStack for its unique customizability.

“We threw our weight behind OpenStack partially because it was open and so it gave us the ability to go in and drive some changes,” said Carl Perry, cloud architect for Brea, Calif. based DreamHost, which is putting together a public cloud storage and cloud compute offering based on OpenStack and slated to open for business by the end of the year.

One change DreamHost became involved in was the development of OpenStack’s “Quantum” virtual network service, which led DreamHost to network virtualization software maker Nicira (now a part of VMware Inc.).

Prior to Quantum, OpenStack offered two choices for creating cloud networks: VLANs, which weren’t scalable beyond about 4000 tenants, or a flat network in which tenants shared one VLAN broken up by IP space, which wasn’t totally secure.

“Quantum abstracted all the networking functionality in OpenStack away from how it is physically implemented, so … when you say, ‘allocate me a virtual network,’ the plugins that are configured by the operator of the OpenStack implementation will choose the correct technology,” Perry said.

DreamHost needed a customizable stack for storage, as it prefers to use the Ceph system it has been developing over the last seven years over OpenStack’s Swift object storage repository.

OpenStack’s flexibility also extends to the compute layer, Perry said.

“One of the things we’re considering down the road is possibly offering something other than x86 VMs, something like ARM.”

The ability to customize network services outside of Quantum is among its perks, according to Argentina-based MercadoLibre, a staunch OpenStack supporter.

“At network creation time, we need a custom add-on to send API calls to our custom in-house API to create VLANs on physical switches that have no drivers to integrate them with OpenStack or Quantum,” said Leandro Reox, senior infrastructure engineer and cloud architect at MercadoLibre. “So we made an in-house API to interact with our switches.”

Admittedly, this requires a certain level of skill with coding and command-line interfaces.

“We don’t care about pretty dashboards,” Reox said. “We are API- and CLI- oriented guys.”

Even for experienced developers, there are some kinks to be worked out with OpenStack, according to DreamHost’s Perry, which include the formulation of standardized operational best practices for deploying it. This, in turn, makes interoperability between OpenStack clouds or hybrid OpenStack clouds a tricky proposition for the time being.

“A lot of people have deployed OpenStack but not a lot of them are talking about how they’ve done it,” he said. “That’s something we’re trying to address.”

So many open source cloud competitors, so little time
Meanwhile, as OpenStack has developed, competitors such as Citrix Systems Inc.’s CloudStack have also gained a foothold in the market for cloud stack alternatives, which is a busy one.

The good news for OpenStack is the market remains up for grabs, but that won’t be true forever.

“CloudStack is definitely a more complete, more mature and polished cloud platform than OpenStack, that's just a fact,” said Carl Brooks, an analyst with 451 Research based in Boston. “[CloudStack] might have as many as sixty to seventy production deployments; OpenStack might have two dozen or so.”

“To put that in perspective, there are hundreds of vCloud deployments and many, many thousands of VMware environments that might legitimately be called ‘private cloud’ right now,” Brooks added.

Then, there is a panoply of cloud stacks being offered for hosting providers, including OnApp, CA Applogic, ElasticStack, Platform Computing (now owned by IBM), and Virtustream, among others.

“I think it’s fair to say that OpenStack has the greatest share of the attention and the least share of the business at this point,” according to Brooks.