Tweet

Follow @petermichaelhan

Software and technologies designed for operating and monitoring the applications, data and services residing in the cloud.

Cloud management tools help ensure a company’s cloud computing-based resources are working optimally and properly interacting with users and other services.

Cloud management strategies typically involve numerous tasks including performance monitoring (response times, latency, uptime, etc.), security and compliance auditing and management, and initiating and overseeing disaster recovery and contingency plans.

With cloud computing growing more complex and a wide variety of private, hybrid, and public cloud-based systems and infrastructure already in use, a company’s collection of cloud management tools needs to be just as flexible and scalable as its cloud computing strategy.

The key to the flexibility of the cloud is closing this feedback loop and making it a seamless process. Cloud infrastructure can easily scale up the number of virtual instances, because they are all independent. Monitoring the performance of VMs is straight forward and standard even for traditional IT.

The novel aspect of cloud solutions is combining these two aspects so that your infrastructure can dynamically provision resources based on real time measurements and monitoring.

At its core, provisioning is, relatively speaking, a simple task. An image is selected from a catalog and instantiated on physical hardware. To build a true cloud infrastructure, there are a number of other capabilities to consider including elasticity, resiliency, cost management, availability, and image management.

A cloud by its very nature does not have ridged boundaries – to an end user it can appear infinite. As an administrator you rely on your provisioning engine to help you elastically scale out your applications running on the cloud. The ability to quickly scale up and scale down your cloud applies just-in-time benefits to IT.

For example, as an online retailer running a Black Friday promotion, it is paramount that you are able to scale out your online storefront to meet the increased customer demand.

In some cases this might a 5x or 10x increase in demand. You need to be able to serve your customers during the spike in demand, but it is not cost effective to permanently scale out your infrastructure. The elasticity of your cloud allows you to perform just in time scaling.

The provisioning engine is the gateway between the world of physical hardware and the world of virtual images. The cloud infrastructure must readily accommodate hardware and software updates. You should be able to add new hardware to your cloud infrastructure as you grow without interruption. Software updates should be handled in a manner that is transparent to the end users. Analysis of outages at major cloud providers like Amazon, Microsoft, and Google has shown that most were caused by software updates.

One of the byproducts of the mass virtualization of resources that has marked cloud entry is image sprawl. For a single business application you could have different image versions for development, test, pre-production, and production. Multiply that by a number of different applications and you can see how this problem gets out of hand quickly.

Image sprawl can lead to overspending by IT. According to Forbes, “in more than 95% of environments we find overspending of some form.” Overspending can come in many forms including operating system license costs, storage costs, and time associated with managing and ever growing inventory of virtual machines.

An image library with the ability to maintain relationships between images, analyze contents of images, and display deltas can really help you as a cloud administrator keep a much better handle on image sprawl. (Does an image library help with overspending on licenses?)

Enterprise monitoring has been around for a long time, so it should be no surprise that it is a core part of the foundation of a cloud infrastructure. When it comes to monitoring in a cloud environment at a minimum you want to focus on availability and capacity.

The key metrics (CPU, storage, memory) will be very familiar to anyone who has implemented enterprise monitoring in the past. The main difference is that in the cloud, these monitoring metrics may apply to a shared pool of resources that may be moved between different applications depending on need.

Monitoring and alerting on the availability of your cloud can greatly reduce down time and speed up problem resolution time. At a high level, this is similar to a traditional data center model. There are a couple key differences. In a cloud environment your application may not always be running on the same physical hardware or with the same number of virtual servers. Your monitoring will also need to be able to dynamically adjust its thresholds and baselines as your cloud application scales up and down.

Additionally, an availability baseline that was established when your application was small may not be relevant any more. For example, if you started running on 10 virtual machines and scaled out to 100 virtual machines, the potential for a problem is exponentially worse – so availability must improve by >10X. It is paramount that your monitoring tools are able to adjust as your cloud environment scales up and down.

Cloud capacity usage is critical for both real time monitoring and historical analysis for future hardware expenditures. In real time, capacity analysis can help you determine the status of your resource pools, and alert you when any of your physical resources are nearing depletion.

This can help you make appropriate placement decisions for new virtual machines. The other area where capacity analysis is invaluable is trending reports. These can help you answer questions such as “What hardware will I need if my business grows by 10% each of the next three years?” or “What happens if our new marketing campaign brings in 30% more orders?” Capacity analysis can really help you make smart financial decisions about your cloud infrastructure.

You can both maximize your current hardware investment as well as effectively plan for future investments.

From an administrator’s point of view, the true power comes from integrating both provisioning and monitoring into a single dynamic cloud management environment. In a cloud environment monitoring data, particularly around capacity, becomes more actionable. Integration of monitoring data and alerts with provisioning actions really helps an administrator effectively manage their cloud infrastructure in response to external or internal events.

Suppose your company posts a video of your latest product announcement on the corporate website which is hosted on your private cloud.

The website is handling requests without issue until a prominent blogger recommends your video and it goes viral. With an integrated management environment your monitoring will detect the increase in volume and automatically request that the provisioning engine provision more capacity. As the hype dies down the extra capacity can be automatically de-provisioned.

A fairly common internal event is a hardware failure. The cloud monitoring detects a failure in one of the blades in your data center. The alert automatically requests that the provisioning engine re-deploy the workload that was running on the failed blade to a different resource. The failed blade can be removed and it will automatically be removed from the available resource pool. Once it has been repaired it can be replaced and automatically added back to the resource pool.

Traditionally IT scalability was something that required lots of planning and foresight. Looking back at the “Black Friday” example, traditionally a business could plan months in advance and build out an infrastructure to support the increase in demand. In a cloud environment the combination of monitoring and provisioning can handle this on demand. This can lead to significant cost savings, not only in hardware, but additionally in the reduced labor costs associated with manually planning and scaling out the infrastructure.

To expand on the example even further, in a cloud environment with the appropriate monitoring in place you can react with the same speed to an unpredicted spike in traffic just as easily as one that you know about a year in advance. The combination of provisioning and monitoring allows for the elasticity of a cloud environment to be truly dynamic.

Management in the cloud starts with a solid foundation. At IBM we offer products to help build this foundation. IBM SmartCloud Provisioning and IBM SmartCloud Monitoring are both part of our recently announced IBM SmartCloud Foundation offerings. For more information on IBM’s cloud solutions please visit IBM.

Many IT shops moving from a virtualized infrastructure to a cloud-based environment initially fall back on their hypervisor provider for management capabilities. However, as they expand into public, private, and hybrid deployments, the baseline cloud management tools fall short.

IT staffs aren't able to successfully integrate all of the various types of clouds, or deal with the complexities of moving between environments, according to various cloud experts.

"The 'aha' moment typically comes when companies want to start moving activities between clouds, offering self-service requests for internal resources, or deploying workloads, [and] thus need tools to connect private and public clouds," said Dave Bartoletti, senior analyst of infrastructure and operations at Forrester Research. "It's not really a function of size--it's a function of when you want to start kicking in cloud economics. You start to realize you're spending too much time creating, deploying, and recovering cloud resources and that you're going to run out of steam really quickly if you are still manually provisioning resources.

That's not really behaving like a cloud."

Once an IT organization becomes serious about cloud management, keep three danger zones in mind, experts say:

-- 1. Comparing tools won't be easy. Cloud management services are still fairly embryonic, and the platforms are evolving so it's hard to make apples-to-apples comparisons. A cloud management platform that has legs to meet future requirements must address a broad range of capabilities, not just performance monitoring. It also should provide automation, self-service provisioning, chargebacks, security and compliance audits, and governance. The ability to support a multi-cloud environment is perhaps the most critical differentiator.

Even organizations that are just starting out should take a long-term view and define a cloud roadmap so it can best match its requirements with what a particular cloud management provider delivers today and what it's committed to deliver tomorrow.

"All of the major players have management tools around their platforms, but they're all different from one provider to another, each with their own nuances," said Rick Blaisdell, CTO at ConnectEDU, who is directing the buildout of the college and career planning site's cloud architecture.

ConnectEDU is currently using proprietary tools to manage its private cloud, hosted by a managed services provider, but Blaisdell is fully aware that longer term, this strategy will need some fine tuning. "Eventually we'll want one pane of glass so we can see all the SaaS applications and the infrastructure-as-a-service stuff tied all together on one management system with dashboards so we can look back at our SLAs and evaluate uptime. We can use third-party tools to get some of this information, but we're not there yet."

2. Don't ignore internal process revamps. The cloud management platform needs to be not only multi-cloud aware, but also integrated with existing internal processes. Given that most firms don't do a wholesale migration to the cloud, the cloud management platform needs to be linked to the existing infrastructure management tools, which is a rebuilding and integration effort many companies don't anticipate.

Further, companies need to think through what they want to offer from the standpoint of a standard services catalog--another hurdle related to integration because that image library needs to be pushed out to span multiple clouds. "You have to spend time thinking about what services you want to provide to end users from either a public or private cloud," said Forrester's Bartoletti. "How many deployment options do you want to provide? Three sizes of a Windows development environment, or six flavors of Linux?

You have to take the time upfront to figure that out." Bartoletti's recommendation: "Offer a limited number of standardized services upfront and only expand when you're forced to. That's how you drive cloud economics."

3. Cloud tools must be a new kind of dynamic. The dynamic nature of the cloud demands a different type of management tool, one that understands the difference between an outage and something that is shut off purposely, said Dave Roberts, senior VP of business development and platform ecosystem for ServiceMesh, which provides cloud management products and services.

In the cloud, something could theoretically spring into existence in the morning and be taken down every night. "From a monitoring tool perspective, it's hard to tell the difference if the tool isn't set up to understand the characteristics of the cloud," said Roberts. A cloud management tool is set up in the path of provisioning so it can monitor the infrastructure and tell if someone is simply flipping a switch to shut down a cloud or if there a bigger problem, Roberts said. A traditional management tool can't make that distinction..........

CLOUD MANAGEMENT TOOLS AND PROPER USE

Cloud management is a hot topic, so hot that every startup and established vendor has some form of tool for managing cloud computing environments. There are tools that monitor, tools that provision, and tools that cross the divide between both. Then there's just vaporware, and sorting through that can be a challenge.

Understanding cloud computing:

If you are still iffy on the concept of cloud computing, this definition provides illumination on the distinct characteristics of cloud.

If your cloud deployment is fairly static or not mission-critical, then you may not need a dynamic provisioning system. In that case, the standard tools for resource adds/changes/removals included with the product may suffice.

Several providers have products designed for cloud computing management (VMware, OpenQRM, CloudKick, and Managed Methods), along with the big players like BMC, HP, IBM Tivoli and CA. Each uses a variety of methods to warn of impending problems or send up the red flag when a sudden problem occurs. Each also tracks performance trends.

While they all have features that differentiate them from each other, they're also focused on one key concept: providing information about cloud computing systems. If your needs run into provisioning, the choices become more distinct than choosing "agent vs. agentless" or "SNMP vs. WBEM."

The main cloud infrastructure management products offer similar core features:

• Most support different cloud types (often referred to as hybrid clouds).
• Most support the on-the-fly creation and provisioning of new objects and the destruction of unnecessary objects, like servers, storage, and/or apps.
• Most provide the usual suite of reports on status (uptime, response time, quota use, etc.) and have a dashboard that can be drilled into.

When it comes to meeting those three criteria, there are a few vendors that offer pervasive approaches in handling provisioning and managing metrics in hybrid environments: RightScale, Kaavo, Zeus, Scalr and Morph. There are also options offered by cloud vendors themselves that meet the second and third criteria, such as CloudWatch from Amazon Web Services.

The large companies known for their traditional data center monitoring applications have been slow to hit the cloud market, and what products they do have are rehashes of existing applications that do little in the way of providing more than reporting and alerting tools. CA is on an acquisition spree to fix this and just acquired 3Tera, a cloud provisioning player.

An example of the confusion in the industry is IBM's Tivoli product page for cloud computing. You'll notice that clicking the Getting Started tab results in a 404 error. Nice work, IBM.

Meanwhile, HP's OpenView (now called Operations Manager) can manage cloud-based servers, but only insofar as it can manage any other server. BMC is working on a cloud management tool, but doesn't have anything beyond its normal products out at the moment.

More on cloud management:

Cloud computing management overview

In place of these behemoths, secondary players making splashes on the market are offering monitoring-focused applications from companies like Scout, UpTime Systems, Cloudkick, NetIQ and ScienceLogic. There is also the "app formerly known as" Hyperic, now owned by VMware through the acquisition of SpringSource.

In truth, we could rival John Steinbeck and Robert Jordan in word count when it comes to writing about all the products in this field, though within a year or two it should be a much smaller space as acquisitions occur, companies fail and the market sorts itself out. T

There's a lot on the way in cloud computing, not the least of which is specifications. Right now the cloud is the Wild West: vast, underpopulated, and lacking order except for a few spots of light.

These are the best infrastructure management and provisioning options available today:

RightScale
RightScale is the big boy on the block right now. Like many vendors in the nascent market, they offer a free edition with limitations on features and capacity, designed to introduce you to the product (and maybe get you hooked, ala K.C. Gillette's famous business model at the turn of the 20th century). RightScale's product is broken down into four components:

• Cloud Management Environment
• Cloud-Ready ServerTemplate and Best Practice Deployment Library
• Adaptable Automation Engine
• Multi-Cloud Engine

A fifth feature states that the "Readily Extensible Platform supports programmatic access to the functionality of the RightScale Platform." In looking at the product, these features aren't really separate from one another, but make a nice, integrated offering.

Helpful cloud tips from RightScale:

RightScale's management environment is the main interface users will have with the software.

It is designed to walk a user through the initial process of migrating to the cloud using their templates and library. The management environment is then used for (surprise!) managing that environment, namely continuing builds and ensuring resource availability.

This is where the automation engine comes into play: being able to quickly provision and put into operation additional capacity, or remove that excess capacity, as needed. Lastly, there is the Multi-Cloud Engine, supporting Amazon, GoGrid, Eucalyptus and Rackspace.

RightScale is also working on supporting the Chef open-source systems integration specifications, as well. Chef is designed from the ground up for the cloud.

Kaavo
Kaavo plays in a very similar space to RightScale. The product is typically used for:

• Single-click deployment of complex multi-tier applications in the cloud (Dev, QA, Prod)
• Handling demand bursts/variations by automatically adding/removing resources
• Run-time management of application infrastructure in the cloud
• Encryption of persisted data in the cloud
• Automation of workflows to handle run-time production exceptions without human intervention

The core of Kaavo's product is called IMOD. IMOD handles configuration, provisioning and changes (adjustments in their terminology) to the cloud environment, and across multiple vendors in a hybrid model. Like all major CIM players, Kaavo's IMOD sits at the "top" of the stack, managing the infrastructure and application layers.

More on working with Kaavo:

One great feature in IMOD is its multi-cloud, single system tool. For instance, you can create a database backend in Rackspace while putting your presentation servers on Amazon. Supporting Amazon and Rackspace in the public space and Eucalyptus in the private space is a strong selling point, though it should be noted that most cloud management can support Eucalyptus if it can also support Amazon, as Eucalyptus mimics Amazon EC2 very closely.

Both Kaavo and RightScale offer scheduled "ramp-ups" or "ramp-downs" (dynamic allocation based on demand) and monitoring tools to ensure that information and internal metrics (like SLAs) are transparently available. The dynamic allocation even helps meet the demands of those SLAs. Both offer the ability to keep templates as well to ease the deployment of multi-tier systems.

Zeus
Zeus was famous for its rock-solid Web server, one that didn't have a lot of market share but did have a lot of fanatical fans and top-tier customers. With Apache, and to a lesser-extent, IIS, dominating that market, not to mention the glut of load balancers out there, Zeus took its expertise in the application server space and came up with the Application Delivery Controller piece of the Zeus Traffic Controller. It uses traditional load balancing tools to test availability and then spontaneously generate or destroy additional instances in the cloud, providing on-the-fly provisioning. Zeus currently supports this on the Rackspace and, to a lesser extent, Amazon platforms.

Scalr
Scalr is a young project hosted on Google Code and Scalr.net that creates dynamic clusters, similar to Kaavo and RightScale, on the Amazon platform. It supports triggered upsizing and downsizing based on traffic demands, snapshots (which can be shared, incidentally, a very cool feature), and the custom building of images for each server or server-type, also similar to RightScale. Being a new release, Scalr does not support the wide number of platforms, operating systems, applications, and databases that the largest competitors do, sticking to the traditional expanded-LAMP architecture (LAMP plus Ruby, Tomcat, etc.) that comprises many content systems.

Morph
While not a true management platform, the MSP-minded Morph products offers similar functionality in its own private space. Morph CloudServer is a newer product on the market, filling the management and provisioning space as an appliance. It is aimed at the enterprise seeking to deploy a private cloud. Its top-tier product, the Morph CloudServer is based on the IBM BladeCenter, and supports hundreds of virtual machines.

Under the core is an Ubuntu Linux operating system and the Eucalyptus cloud computing platform. Aimed at the managed service provider market, Morph allows for the creation of private clouds and the dynamic provisioning within those closed clouds. While still up-and-coming, Morph has made quite a splash and bears watching, particularly because of its open-source roots and participation in open-cloud organizations.

CloudWatch
Amazon's CloudWatch works on Amazon's platform only, which limits its overall usefulness as it cannot be a hybrid cloud management tool. Since Amazon's Elastic Compute Cloud (EC2) is the biggest platform out there (though Rackspace claims it is closing that gap quickly), it still bears mentioning.

CloudWatch for EC2 supports dynamic provisioning (called auto-scaling), monitoring, and load-balancing, all managed through a central management console -- the same central management console used by Amazon Web Services. Its biggest advantage is that it requires no additional software to install and no additional website to access applications through. While the product is clearly not for enterprises that need hybrid support, those that exclusively use Amazon should know that it is as robust and functional as the other market players.

Conclusion
With the plethora of tools on the market, it is important that data center managers begin their assessments early and keep an eye on how the market progresses. This change in the very nature of IT infrastructure is both vast and fast, necessitating the constant revision of plans and review of products, as well as observation of what is happening to each of the companies behind these products.

It is clear that RightScale has the early lead, but the other vendors are catching up quickly and bear close watching...................