Starting in WebSphere CloudBurst 2.0, there are different levels of elasticity that you can achieve in your WebSphere deployments. As I mentioned in a previous post, the Intelligent Management Pack allows you to define dynamic clusters. This means cluster membership and the number of instances of a given application adjusts on the fly to meet SLAs for your application. This enables a more dynamic environment as opposed to static cluster definitions, but there is a layer of elasticity below this that bears exploring.
Dynamic clusters work with WebSphere nodes that already exists. Users define the nodes available for use by a dynamic cluster, and the runtime uses SLAs and current system state to determine the actual nodes used and application instances started. So, what if you need more nodes than what you currently have in a given WebSphere environment? A dynamic cluster will not create a new node, so you have to define extra nodes. Starting in WebSphere CloudBurst 2.0, this is as easy as pushing a button.
Dynamic virtual machine operations allow you to add and remove nodes on the fly for a given virtual system. For instance, take the pattern in the picture below:
If you were to deploy this pattern, you would end up with a WebSphere Application Server cell with a node makeup similar to the below:
Now that the environment is out there (in mere minutes I should mention), suppose you want to add more nodes? Before WebSphere CloudBurst 2.0, you could have done it, but it would involve creating another pattern with a custom node part and deploying it. This results in two different virtual systems and complicates the maintenance stream. Now, in WebSphere CloudBurst 2.0, you can simply click a button to add a node to the existing virtual system.
From the virtual systems view, if you expand the virtual machines, by each virtual machine you will see an Actions column with a View link. If you want to add a node to the environment shown above, you simply click the View link, and then click the clone icon highlighted in green below:
WebSphere CloudBurst prompts you for the number of nodes to add. You make the selection and then click OK. The appliance creates the new node and federates it into the cell for you. For instance, if you chose to add a single node, at the end of the clone not only would you have another virtual machine in your virtual system, but also another node automatically federated into your WebSphere Application Server cell:
On the flip side, you can remove a node by clicking the delete icon in the same dialog as the clone icon above. This removes the node from the WebSphere Application Server cell and deletes the virtual machine.
The ability to easily add and remove virtual machines from your WebSphere CloudBurst virtual systems enables a very valuable level of elasticity. Now you can very easily add and remove nodes on the fly based on the current demands of your system. As always, let me know if you have any questions or comments.
A while back I had a four part FAQ series inspired by questions arising from customer visits discussing the first release of WebSphere CloudBurst. With the recent release of WebSphere CloudBurst 2.0, I think it is a good time to revisit an FAQ series with an entirely new set of questions.
For the first part of the series, I want to address a question we get all the time now: "What is the difference between WebSphere CloudBurst and WebSphere Virtual Enterprise?" This question was always fairly common, but now even more so because the new Intelligent Management Pack option for WebSphere Application Server Hypervisor Edition allows you to deploy WebSphere Virtual Enterprise cells using WebSphere CloudBurst.
Fundamentally, the difference between the WebSphere CloudBurst Appliance and WebSphere Virtual Enterprise is a complementary one. WebSphere CloudBurst provides a means to create your application environments, deploy them into a shared, cloud environment, and then manage them over time. In this respect, the appliance focuses on bringing cloud-like capabilities to the application infrastructure layer of your application environments. WebSphere CloudBurst does give you management capabilities for your running, virtualized application environments (i.e. applying maintenances and fixes), but for the most part those capabilities do not extend into the application runtime environment.
Now, you may ask why WebSphere CloudBurst does not extend its reach into the application runtime. The answer is simple: We already have a solution that does just that, WebSphere Virtual Enterprise. WebSphere Virtual Enterprise provides capabilities that allow you to dynamically and autonomically manage your application runtime. You can use WebSphere Virtual Enterprise to not only assign performance goals to your applications, but also to declare the importance of a given application meeting its goals relative to other applications in your organization. This enables the dynamic management of your applications and their resources such that your applications perform according to their goals and relative importance to your business. Simply put, you get an elastic runtime at the application layer of your application environments.
As I said, WebSphere CloudBurst and WebSphere Virtual Enterprise are complementary solutions. Both enable notions of cloud computing, but at different layers of your application environments. WebSphere CloudBurst hones in on the application infrastructure components, while WebSphere Virtual Enterprise zeros in on the applications running in those environments. The new Intelligent Management Pack for WebSphere Application Server Hypervisor Edition means that WebSphere CloudBurst can now dispense WebSphere Virtual Enterprise environments into your on-premise cloud. That means you can take advantage of these complementary solutions from a single and simple management plane.
I hope this helps to clear things up. As always, questions and comments are welcome!
Among the major features of the new virtual application pattern in IBM Workload Deployer is the notion of elasticity. That is, as your application needs more resources, it gets them. When your application can meet its SLAs with fewer resources, the environment shrinks. With this kind of pattern, you enable elasticity by specifying a policy and defining the scaling trigger (i.e. CPU usage, application response times, database response times, etc.). What may have been a bit lost in some of these new announcements regarding IBM Workload Deployer is the fact that you can now leverage this core feature of cloud, elasticity, in your virtual system patterns.
If you have read this blog in the past, you probably already know that the Intelligent Management Pack is an option for virtual system patterns built using WebSphere Application Server Hypervisor Edition. When you enable the Intelligent Management Pack option, you are essentially building and deploying WebSphere Virtual Enterprise (WVE) environments. For those of you not familiar with WVE, the best way to describe it is that it provides you with application and application infrastructure virtualization capabilities. Of its many capabilities, one most germane to our discussion today is the ability for users to attach SLAs to applications and then have WVE automatically prioritize requests and manage resources in order to meet those SLAs. Inherent in this capability is the ability to dynamically start and stop application server processes (JVMs) as required. In other words, WVE provides JVM elasticity.
The fact that WVE provides JVM elasticity is nothing new. Further, IBM Workload Deployer started providing virtual machine (VM) elasticity in previous versions (when it was WebSphere CloudBurst). With this feature, you could add or remove VMs to an already deployed virtual system using dynamic virtual machine operations provided by the appliance. The catch was that the VM elasticity was a manual action and you could not link this elasticity to the same SLAs tied to your applications. Well, thanks to a new feature in WebSphere Virtual Enterprise and easy integration provided by the Intelligent Management Pack, this is no longer the case.
Starting in IBM Workload Deployer 3.0, you can take advantage of a new WVE feature called Elasticity Mode when using the Intelligent Management Pack. Elasticity mode is not unique to IBM Workload Deployer, but a concept new to the base WVE product. It allows one to define actions for how WVE should grow and shrink the set of nodes used by application server resources. Like the basic JVM elasticity capability in WVE, these node elasticity actions trigger based on SLAs tied to your applications. Consider the case that you are using elasticity mode and your application is not currently meetings its SLA. If WVE does not think it can start any more application server instances on the current set of nodes, it will grow the set of nodes per your elasticity configuration. Conversely, if WVE detects that it can meet SLAs with fewer nodes, it will shrink the resources per your elasticity configuration.
In IBM Workload Deployer, using elasticity mode becomes even easier. All you need to do is use the Intelligent Management Pack and enable the elasticity mode option in your virtual system patterns. When you do this, you get automatic integration between IBM Workload Deployer and the deployed WVE environment. What does that mean? It means that if WVE detects it needs more nodes, it will automatically call back into IBM Workload Deployer and request that the appliance provision a new VM that will serve as a node for application server processes. It also means that if WVE detects it could meet SLAs with fewer resources, it will call into IBM Workload Deployer and ask it to remove a node. All of this happens without any user scripting. All you have to do is enable this option in your patterns and configure SLAs appropriate for your applications.
To me, this exciting new feature brings out the best of elasticity capabilities in both IBM Workload Deployer and WebSphere Virtual Enterprise. The result is a single management plane that gives you both VM and JVM elasticity for your cloud-based application environments. Best of all, elasticity actions map directly to SLAs for your applications. After all, when it comes to cloud, it's the application that really matters!
In a previous post, entitled Layers of Elasticity, I talked about the new dynamic virtual machine operations in WebSphere CloudBurst. Specifically, I showed you how to use the WebSphere CloudBurst web console to add more virtual machines (nodes) to an existing virtual system. Well, you can do this with the WebSphere CloudBurst command line interface as well.
First, let's assume I start off with a basic WAS ND environment represented by the pattern below:
When I deploy this pattern in WebSphere CloudBurst, I end up with two virtual machines: one for the deployment manager with an embedded IHS instance, one for my custom node federated into the cell. After deployment, suppose I want to use the CLI to interact with this virtual system. Assuming the name of my virtual system is Cluster, I can view my custom node virtual machine with the following CLI code:
The call to the clone function above takes care of creating a new profile and federating the new node into the cell. In addition, WebSphere CloudBurst automatically invokes any script packages from the source virtual machine marked to run at virtual system creation. All because of this single line of code!
The WebSphere CloudBurst CLI is a powerful interface that enables you to automate the function of the appliance. Check it out, become familiar with it, and make WebSphere CloudBurst processes a seamless part of your overall data center management approach.