I spent most of my time growing up doing two things, going to school and playing sports. I made many fond memories -- mostly from the latter :) -- and learned more than a few lessons over that time. Of all of those lessons, there was one in particular that stuck out in both the classroom and on the baseball diamond: Sometimes you have to get back to the basics.
In that vein, I think it is time to revisit the basics of WebSphere CloudBurst. In revisiting the basics, I am not talking about the technical basics of the appliance. Rather, I am talking about revisiting exactly why WebSphere CloudBurst exists in the first place. In other words, let's take a look at the problem domains WebSphere CloudBurst addresses, and let's discuss a little bit about how the appliance does so.
One of my favorite things to do with users or potential users of WebSphere CloudBurst is to help them understand how they can construct a custom environment using the appliance. Typically, we take one of their existing application environments and discuss the configuration steps that contribute to its makeup. From there, we map the required configuration actions to different customization capabilities in the appliance. It is one thing to talk about how you can customize every layer of your application stack with WebSphere CloudBurst, it is quite another to talk about it in the context of an existing environment. This exercise usually serves to greatly enhance a user's understanding of how to construct tailored environments with the appliance.
While I cannot take every one of you through this exercise in the context of one of your own application environments, I can propose a scenario that will help to illustrate the WebSphere CloudBurst customization process. Consider that I want to deploy a clustered WebSphere Application Server environment whose application server instances utilize WebSphere DataPower XC10 for HTTP session management. In order to deploy such an environment, I would need to do the following:
Install an OS and WAS
Install the WebSphere eXtreme Scale Client binaries - required for integration
Construct a clustered cell
Augment profiles with WebSphere eXtreme Scale profile templates
Configure the application server instances to use WebSphere DataPower XC10 for session management
So those are the steps, but how do they map to WebSphere CloudBurst? First, I know that the WebSphere Application Server Hypervisor Edition image used by WebSphere CloudBurst encapsulates the installation of the OS and WAS. I also know that WebSphere CloudBurst will automatically construct the clustered cell during the deployment process. That means I need to address the installation of client binaries, augmentation of profiles, and configuration of application server instances. In order to do this, I will use a combination of image extension and custom script packages.
To get started, I extend an existing WebSphere Application Server Hypervisor Edition image and simply install the WebSphere eXtreme Scale Client binaries. I then capture that image and store it as my own unique image in the WebSphere CloudBurst catalog. Now, you may wonder why I did not capture the profile augmentation in the custom image. Remember, you cannot change profile configuration during the extend and capture process as WebSphere CloudBurst resets the profiles as part of capturing the custom image.
My custom image encapsulates the installation of the client binaries, so now I turn to custom script packages. I need two in this case. One script package will augment a profile (either deployment manager or custom node) with the WebSphere eXtreme Scale profile template. The second script package will configure application server instances to use WebSphere DataPower XC10 for HTTP session management. Once done with these script packages, I have all the assets I need to build my target environment.
Using my custom image, I build a pattern that contains the number and kind of WebSphere Application Server nodes that I want. I use the advanced options to define a WebSphere Application Server cluster ensuring its creation happens during deployment. Next, I drag and drop the profile augmentation script onto the deployment manager and custom node parts in my pattern. Finally, I drag and drop the WebSphere DataPower XC10 configuration script onto the deployment manager. The pattern is now ready to deploy!
For those of you that are visual learners like me, this demonstration provides a nice overview of exactly what I wrote about above. Check it out and let me know what you think.
One of my favorite books from childhood is If You Give a Mouse a Cookie. Although targeted at children, the book illustrates a frequently occurring human behavior that is important for all of us understand. That behavior is the tendency for escalating expectations. The book offers this up by starting out with the simple action of giving a mouse a cookie. The mouse in turn asks for a glass of milk, various flavors of cookies, and on and on, until the mouse circles back to asking for another cookie.
Nearly all of us exhibit this same kind of behavior, and it can often produce positive results. In particular, in IT we always push for the next best thing or a slightly better outcome. Personally, I am no stranger to this behavior because I experience it from WebSphere CloudBurst users quite frequently. In these cases, it usually revolves around one particular outcome: speed of deployment.
Bar none, users of WebSphere CloudBurst are experiencing unprecedented deployment times for the environments they dispense through the appliance. The fact that we say you can deploy meaningful enterprise application environments in a matter of minutes is far beyond just marketing literature. Our users prove it everyday. However, just because they are deploying things faster than ever does not mean they are content to rest on those achievements. They want to push the envelope, and I love it.
For our users looking to achieve even speedier deployment times, I offer up one reminder and one tip. First, analyze all of your script packages to ensure you are using the right means of customization. If you have some scripts that run for considerably longer than most other script packages, you may want to at least consider applying that customization by creating a custom image. You still need to adhere to the customization principles outlined here, but you may benefit from applying the customization in an image once and avoiding the penalty for applying it during every deployment. You may also be able to break this customization out with a combination of a custom image and script packages. For instance, instead of having a script that installs and configures monitoring agents, you may install the agents in a custom image and configure them during deployment. Being selective about how and when you apply customizations can go a long way in improving your deployment times.
In addition to the reminder above, I also have a tip. Take a look at all of the script packages you use in pattern deployments and look to see if there are any that you can apply in an asynchronous manner. In other words, identify customizations that need to start, but not necessarily complete as part of the deployment process. Going back to our example of configuring monitoring agents during the deployment process, it may be important to kick off the configuration script during deployment, but is it crucial to wait on the results? Maybe not. If it is not, consider defining the executable argument in your script package in a manner that kicks off the execution and proceeds -- i.e. nohup executable command &. This approach can save deployment time in certain situations.
My advice to users of WebSphere CloudBurst: keep pushing your deployment process! Pare as many minutes off the process as you can. I hope that the tips above help in that regard, and be sure to pass along other techniques that you have found helpful.
Though I feel like we've come a long way in some of the initial confusion surrounding IBM CloudBurst and WebSphere CloudBurst, I still get quite a few basic questions on the solutions. The two most common questions are, 'Are they different products?', and 'Can/should I use them together?'. I put together a really brief overview that answers these questions and talks about the basics of the combined solution. I hope it provides a good introduction!
In previous posts, I have discussed the integration capability between WebSphere CloudBurst and Tivoli Service Automation Manager. Most recently, I discussed this in the context of integrating WebSphere and IBM CloudBurst. Today, I am happy to announce the publication of an article I co-wrote with Marcin Malawski from TSAM development on the subject of this integration.
If you are a WebSphere user interested in a holistic approach in building out a private cloud, I strongly recommend that you check the article out. If you are currently an IBM CloudBurst, IBM Service Delivery Manager, or Tivoli Service Automation Manager user and you provision a significant number of WebSphere environments, I strongly recommend that you check the article out. In fact, regardless of your current situation, do me a favor and check the article out!
As always, I look forward to feedback and comments. Good, bad, or indifferent. You can leave your comments here or on the article page. I look forward to hearing from you!
When I talk with WebSphere CloudBurst users, the topic of custom virtual images comes up frequently. In some cases they simply want to customize a shipped IBM Hypervisor Edition, and in other cases they want to create a completely custom image. Creating a customized version of an IBM Hypervisor Edition is relatively easy since we give you extend & capture in WebSphere CloudBurst. Creating a completely custom image has historically been a bit tougher, mostly owing ot the fact that there was not a standard tool or process for image assembly. I am happy to say that today's publication of the IBM Image Construction and Composition Tool changes all that.
Watch a demo of the IBM Image Construction and Composition Tool
The primary purpose of the Image Construction and Composition Tool is to enable a modular approach to virtual image construction, while taking into account the typical division of responsibilities within an organization. The tool allows the right people within an organization to contribute their specialized knowledge as appropriate to the virtual image creation process. This means OS teams can handle the OS and software teams can handle the appropriate software. A separate image builder can then use both OS and software components to meet the needs of users within the organization. Best of all, the image builder does not need intimate knowledge of how to install or configure any of the components in the image. They simply need to know which OS and software components to use.
When using the Image Construction and Composition Tool, you start by defining the base operating system you wish to use for your images. You can do this by importing an existing virtual image with an OS already installed, providing an ISO for the OS, or pointing to a base OS image on the IBM Cloud. The bottom line is that you have necessary flexibility to start with your certified or ‘golden’ operating system build. Once you have the base OS image defined in the Image Construction and Composition Tool, you can start defining custom software for use in the images you will compose.
In the tool, bundles represent the software you wish to install within a virtual image. The definition of a bundle contains two major parts: Installation and Configuration. The installation component of a bundle tells the Image Construction and Composition Tool how to install your software into the virtual image. You provide a script or set of scripts that install the necessary components into your image, and you direct the tool to call these scripts. These tasks run once during the initial creation of the virtual image, thus allowing you to capture large binaries, long-running installation tasks, or other necessary actions directly into your image.
The configuration section of a bundle defines actions that configure the software installed into the image. Like with the installation tasks, you provide a script or set of scripts for configuration tasks. Unlike installation tasks that run exactly once, configuration scripts become part of the image’s activation framework and as such, run during each image deployment. Using the tool, you can define input parameters for configuration scripts and optionally expose them so that users can provide values for the parameters at image deploy-time. Configuration tasks are important in providing flexibility that allows users to leverage a single virtual image for a number of different deployment scenarios.
Once you have your base OS image and one or more bundles defined in the Image Construction and Composition Tool, you can compose a virtual image. To compose a virtual image, you extend the base OS image and add any number of bundles into the new image. A base OS image plus a set of bundles defines a unique image.
After you define the image you want to construct, you initiate a synchronize action in the Image Construction and Composition Tool. When you start the synchronize action, the tool first creates a virtual machine in either a VMware or IBM Cloud environment (based on how you configured the tool). Next, the installation tasks of each bundle you included in the virtual image run to install the required software. Finally, the tool copies the configuration scripts from each bundle into the virtual machine and adds them to the image’s activation framework. This ensures the automatic invocation of all configuration scripts during subsequent image deployments.
Once the image is in the synchronized state, you can capture it. Capturing the image results in the creation of a virtual image based on the state of the synchronized virtual machine. The tool also automates the generation of metadata that becomes part of the virtual image package. When the capture of the virtual image completes, you can export it from the Image Construction and Composition Tool and deploy it using WebSphere CloudBurst, Tivoli Provisioning Manager, or the IBM Cloud.
I am excited for users to get their hands on the Image Construction and Composition Tool. I believe it represents the first big step in helping users to design and construct more sustainable virtual images. Did I mention it is completely free to download and use? Visit the Image Construction and Composition Tool website for more details and a download link. I look forward to your comments and feedback.
One of the key benefits of WebSphere CloudBurst adoption is rapid -- seriously fast -- deployments of middleware application environments. Our users are leveraging the appliance to bring up enterprise-class middleware environments in mere minutes. If you know a little bit about WebSphere CloudBurst, that statistic may be a little surprising considering the appliance dispenses large virtual images from the appliance over the network to a farm of hypervisors. You may ask how the appliance can achieve such rapid deployments in light of the mere physics involved in transferring large amounts of data over a network. The simple answer is caching of course!
WebSphere CloudBurst creates a cache for each unique virtual image on datastores associated with the hypervisors in your cloud. On subsequent deployments of the same virtual image to the same datastore, WebSphere CloudBurst does not need to transfer the image over the wire. It simply uses the virtual disks that are in the cache on the datastore. In the context of the virtual image cache, the deployment process goes something like this:
WebSphere CloudBurst identifies the images necessary to deploy the pattern selected by the user.
WebSphere CloudBurst identifies the hypervisors and associated datastores that will host the virtual machines created during deployment.
WebSphere CloudBurst checks the selected datastores to see if they already have caches for the images it will be deploying. From here, one of two things happens:
WebSphere CloudBurst detects that there is no cache on the datastore and transfers the images over to the hypervisor, thereby creating the cache on the underlying datastore.
WebSphere CloudBurst detects that there is a cache on the selected datastore and uses that cache in lieu of transferring the disk over the wire.
The process may sound complicated, but it is completely hidden from you, the user. You do not need to know how the cache works since WebSphere CloudBurst handles all of these interactions. So, why am I telling you all of this then? As a WebSphere CloudBurst user, it is good to be aware of the cache for two main reasons. First, you need to account for the storage space the cache needs when doing capacity planning for your WebSphere CloudBurst cloud. Second, anytime you upload or create a new image through extend and capture, I would strongly suggest you automatically prime the cache for this new image. You can do this by simply deploying a pattern built on the image to each unique hypervisor/datastore in your environment. This may take a temporary re-arrangement of cloud groups, but it is a simple process, and it guarantees rapid deployments for all users of the new image.
I hope this sheds a little light on a subject we do not discuss too often. As always, if you have any questions, do not hesitate to let me know!
IBM Impact 2011 was a wildly busy week! Customer meetings, entertaining keynotes, informative sessions, and hands-on labs packed the 6 days with more than enough action. I spent a lot of the week presenting sessions and conducting labs for the newly announced IBM Workload Deployer. As one would expect with any new announcement, we got tons of questions about IBM Workload Deployer. While I cannot capture all the questions and their answers here, I will try to cover some of the more prevalent ones below.
Question: What happened to WebSphere CloudBurst?
Answer: The short answer is, it simply went through a rename. WebSphere CloudBurst became IBM Workload Deployer v3.0. The version 3.0 acknowledges this is an evolution of what we started with WebSphere CloudBurst, which was at version 2.0. Why remove WebSphere from the name? The fact that this is now an IBM branded offering is more accurate as it is capable of deploying and managing more than just WebSphere software.
Question: What is new in IBM Workload Deployer?
Answer: While there are many new features that I will be talking about over the coming months, the most prominent new facet is the introduction of workload patterns (also referred to as virtual application patterns). As opposed to topology patterns (traditionally referred to as simply patterns in the WebSphere CloudBurst product), workload patterns raise the level of abstraction to the application level. Instead of focusing on application infrastructure and its configuration as you do with topology patterns, workload patterns allow you to focus on the application and its requirements. When using workload patterns, you provide the application, attach policies that specify functional and non-functional requirements, and deploy. IBM Workload Deployer handles deploying and integration the middleware infrastructure necessary to support the application, and it automatically deploys your application on top of that middleware. In addition, IBM Workload Deployer manages the application runtime in accordance with the policies that you specify in order to provide capabilities such as runtime elasticity.
Question: If I am a current WebSphere CloudBurst user, what does this mean for me?
Answer: Not to worry. You will be able to use all of your WebSphere CloudBurst assets (patterns, scripts, images) in the new IBM Workload Deployer. All of the capabilities previously in WebSphere CloudBurst are present in IBM Workload Deployer (terminology may vary slightly -- topology pattern instead of just pattern for instance). Additionally, we continue to expand on the functionality that you are familiar with from WebSphere CloudBurst. This includes updates for Environment Profiles, new IBM Hypervisor Edition images, new pattern building capabilities, and more. Stay tuned for more information about these new features and for information on how you can move your WebSphere CloudBurst resources to the new IBM Workload Deployer.
Question: How do I choose between using workload and topology patterns?
Answer: There are a number of factors that will lead you to using either workload patterns, topology patterns, or both. The primary decision point will be how much control you really need (not want). When using workload patterns, you sacrifice some customization control over the configuration, integration, and administration of the middleware application environment since the workload pattern and management model abstracts away the 'guts' of the system. Everything about the workload pattern is application-centric. On the other hand, topology patterns give you intimate control over the configuration, integration, and administration of the middleware application environment. As a general rule of thumb, if your application requirements match the capabilities of a workload pattern, that is the way to go as it can greatly reduce complexity and cost associated with deployment and management. If a workload pattern does not meet the needs of your application, topology patterns can still greatly reduce cost and complexity and you can tailor them to fit almost any need. Beyond generalities, there is no hard and fast rule for choosing one over the other. It comes down to understanding your application environment and its needs.
Question: Is IBM Workload Deployer an appliance like WebSphere CloudBurst?
Answer: Yes, it is still an appliance, but an updated one! The new appliance is 2U, and it provides more storage, processing power, and memory. It is still just as easy to setup, but just slightly bigger.
Well, that is all for now, but I will be back many times over the coming months with more information. In the meantime, if you have any questions, please leave them in a comment below.
Jason McGee will be leading the second GWC Lab Chat this week on Wednesday, 4/20. The very timely topic is related to recent announcements from IBM regarding the IBM Workload Deployer (see previous posts). Entitled "Application-Centric Cloud Computing" the discussion will focus on the concept of deploying and managing your application workloads in a shared, self-managed environment rather than manually creating and managing the application middleware topologies. It places the focus on the application rather than the infrastructure. This concept promises to deliver greater simplicity, elasticity, and
density among other things. It can position your business to react more
quickly and efficiently to the increasing demands of your customers and
free you from the managing all of the details.
Many of you may have already heard Jason speak last week at IMPACT 2011 in the cloud mini main tent or perhaps at any number of other sessions that Jason was involved in. Jason is the key architect behind IBM's WebSphere cloud activities. Obviously, Jason understands the cloud space very well and has a clear view of the evolution into Application-Centric Cloud Computing. This GWC Lab Chat will provide the opportunity to get your questions answered and share your perspective on this technology.
Jason will provide a brief introduction to the concepts and ideas and then lead an open discussion. Put it on your calendar and plan to attend - and please plan to bring your questions and comments to help foster a rich discussion. We want to hear from you.
If you haven't registered yet it is not too late - learn more and register here. It is easy to register and there is no cost. This is a very timely event and a great way to dig a little more deeply into concepts you first heard at IMPACT or perhaps hear them for the first time. Don't miss it!
When one uses IBM Workload Deployer (previously WebSphere CloudBurst) to deploy a virtual system pattern, they benefit from a completely automated deployment process. The automation includes the creation and placement of virtual machines, injection of IP addresses, initiation of internal processes, and invocation of included scripts. Most of these processes are straightforward and require little more than a brief overview. However, the placement of virtual machines stands out, and it's inner workings are the subject of quite a few questions when I discuss the appliance. With that in mind, I thought I would provide a little more information on how the placement algorithm in IBM Workload Deployer works.
The placement subsystem in IBM Workload Deployer considers three primary elements: compute resource, availability, and license optimization. Compute resource availability is the gating factor for placement. That means that IBM Workload Deployer first looks at the available CPU, memory, and storage resource in the collection of hypervisors making up the cloud group(s) you are targeting for deployment. If a particular hypervisor cannot provide enough resource based on the amount you requested for your deployment, then it is automatically taken out of the eligible hosts pool. It is important to note that IBM Workload Deployer will overcommit CPU, and it will overcommit storage if you direct it to do so. It will not overcommit memory because that could severely degrade the performance of the application(s) running in the virtual machines.
After choosing the pool of hypervisors that are capable of hosting the virtual machines in your deployment from a compute resource perspective, the appliance then considers high availability. To better understand this particular placement stage, let's consider an example. Consider you are deploying a pattern based on WebSphere Application Server Hypervisor Edition and it contains two custom node parts. It is conceivable, and in fact likely, that these two custom node parts will host members of the same cluster, and thus the two nodes will support the same applications. As such, IBM Workload Deployer will attempt to place the two custom nodes on different physical machines in order to prevent a single point of failure. Of course, this depends on having two hypervisors with enough resource (CPU, memory, storage) to host the virtual machines, but the appliance makes that decision in the first placement stage.
After considering compute resource and high availability, IBM Workload Deployer moves to the last stage of placement: license optimization. In this stage, the placement subsystem attempts to place the virtual machines on hypervisors in a way that minimizes the licensing cost to you. The appliance can do this because it is aware of IBM virtualization licensing rules and takes those into account during this stage (if you aren't familiar with virtualization licensing rules and you are curious, ask you're sales representative to explain some time). During this stage, it will not violate any resource overcommit directives or rules in place, nor will it compromise system availability, but it will seek to minimize costs within these parameters.
At this point, I should make something clear that may already have occurred to you. You can override most of these placement rules by creating a cloud group containing only one hypervisor. In this case, IBM Workload Deployer will put all virtual machines on the single hypervisor until it runs out of compute resource (memory is likely to be the constraining factor). I would not suggest that you do this unless you have a good reason or you are in a simple pilot phase, but I do like to point out the art of the possible!
While not incredibly deep from a technical perspective, I do hope that this provided a few helpful details on what goes on during the placement stages of deployment. If you have any questions, do let me know.
Application-centric cloud computing is the main thrust behind the new capabilities of IBM Workload Deployer v3.0. But what does that really mean? After all, application-centricity is really just a concept. Granted, it is an important concept, but it is fairly meaningless until it is put into action or implemented. IBM Workload Deployer does just that with its new Virtual Application Patterns (VAPs).
VAPs are the embodiment of the workload pattern approach I briefly discussed in an overview post a few weeks back. The idea with a VAP is to give the user an interface through which they can provide their application, specify dependencies, declare functional and non-functional requirements and then deploy. Of course application middleware is a part of the overall solution, but IBM Workload Deployer has the smarts to build, configure, and integrate the necessary infrastructure in order to support the user's application. This is completely hidden from the user, so they are liberated to focus on the application and its requirements.
If we scratch a bit further beneath the surface of a VAP, we see that these patterns contain three primary pieces. These primary pieces are components, links, and policies, and they are fundamental to understanding how virtual application patterns work. Let's start with the building blocks of VAPs, components. Put simply, components represent different resources and functionality profiles that make up your application environment. As an example, the IBM Workload Deployer Pattern for Web Applications is a VAP that contains components for an EAR file, WAR file, message queue, and any number of other components that are typical requirements for a web application. The components will certainly vary based on the workload type (i.e. the components included in a web application VAP would be different than those included in a batch application VAP), but they are the foundation of any VAP.
From the ground up, the next logical element we come to in the VAP is a link. A link is a way to declare a dependency or integration point between two components. As an example, consider a VAP with a WAR file component and a database component. You might draw a link between the WAR component and the database component to indicate that your web application uses or otherwise depends on the database. IBM Workload Deployer interprets this link, and takes it as a directive to configure the integration between the two components as a part of deployment. In this case, that may mean configuring a data source in the application's container. This is just a simple example, and an application may have any number of links between components.
Finally, we come to the policy element within the VAP. A policy is a way for a user to specify functional and non-functional requirements for their application environment. Users attach policies to the VAP, or to components in their VAP, and IBM Workload Deployer interprets and enforces those policies. In the context of a web application, one example of a policy could be a scaling policy. The scaling policy might indicate scaling requirements for the application that included minimum application instances, maximum application instances, and conditions that triggered scaling activities. IBM Workload Deployer would use the information in a scaling policy within a VAP to appropriately manage the deployed, running environment. Other examples of a scaling policy may include a JVM policy that provides configuration directives for the java virtual machines in your application environment or a logging policy that defines logging configuration options. In any case, the policy element allows VAP builders to influence the configuration and management of the application environment.
In the example VAP below you can see the use of components (Enterprise Application, Database, User Registry, Messaging Service), links (blue lines between components), and policies (Scaling Policy, JVM Policy):
In total, when I look at a VAP a particular word sticks out to me: declarative. VAPs really enable declarative, application-centric cloud computing. What do I mean? By declarative, I mean you are telling IBM Workload Deployer what you want, but not necessarily how you want it done. It is the job of IBM Workload Deployer to take care of the how. This shift in approach to application environments enables the potential for significant savings, and more importantly to me, lays the foundation for a more agile, flexible approach to deploying and managing application environments.
There will be more in the weeks and months to come on IBM Workload Deployer, so stay tuned. I also want to put a plug in for a new blog from Jason McGee. For those that do not know Jason, he is an IBM Distinguished Engineer, and the lead architect behind IBM Workload Deployer. Be sure to check out his blog for insights on this new offering, as well as for all things cloud.
One of the fundamental tenants of IBM Workload Deployer is a choice of cloud deployment models. Starting in v3.0, users will be able to deploy to the cloud using virtual appliances (OVA files), virtual system patterns, or virtual application patterns. The ability to provision plain virtual appliances is a way to rapidly bring your own images, as they currently exist, into the provisioning realm of the appliance. As such, I think the use cases and basis for deciding to use this deployment model are fairly evident. However, when comparing the two patterns-based approaches, virtual system patterns and virtual application patterns, the decision requires a bit more scrutiny.
Our pattern approach is a good thing for you, the user. Basically, when we refer to patterns in the context of cloud, we are referring to the encapsulation of installation, configuration, and integration activities that make deploying and managing environments in a cloud much easier. Regardless of what kind of pattern you end up using, you benefit from treating a potentially complex middleware infrastructure environment or middleware application as a single atomic unit throughout its lifecycle (creation, deployment, and management). In turn, you benefit from decreased costs (administrative and operational) and increased agility via rapid, meaningful deployments of your environments. That said, it is imperative to understand the differences between virtual system and virtual application patterns, and more importantly, it is important to understand what those differences mean to you. Let's start by considering the admittedly simple 'Cloud Tradeoff' continuum below.
In the above graph, the X-axis represents the degree to which you have customization control over the resultant environment. The degree of control gets lower as we move from left to right. The left Y-axis represents total cost of ownership (TCO), which decreases as we move up the axis. The right Y-axis represents time to value, which similarly decreases as we go up the axis. Naturally, enterprises want to move up the Y-axis, but, and it can be quite a big but, they are sometimes hesitant to relinquish much control (move to the right on the X-axis) in order to do so. In that light, I think it helps to explore our two patterns-based approaches a bit more.
The most important thing to understand about this continuum is that the X-axis really represents the customization control ability from the point of view of the deployer and consumer of the environment. An example is probably the best way to explain. Let's consider a fairly simple web service application that we want to deploy to the cloud. If we were to use a virtual system pattern to achieve this, we would probably start by using parts from the WebSphere Application Server Hypervisor Edition image to layout our topology. We may have a deployment manager, two custom nodes, and a web server. After establishing the topology, we would add custom script packages to install the web service application and then configure any resources the application depended on. Users that wanted to deploy the virtual system pattern would access it, provide configuration details such as the WAS cell name, node names, virtual resource allocation, and custom script parameters, and then deploy. Once deployed, users could access the environment and middleware infrastructure as they always have. That means they could run administrative scripts, access the administrative console provided by the deployed middleware software, and any other thing one would normally do. The difference in using virtual system patterns is not necessarily the operational model for deployed environments (though IBM Workload Deployer makes some things, like patching environments, much easier). Instead, the difference is primarily in the delivery model for these environments.
Using a virtual application pattern to support the same web service application results in a markedly different experience from both a deployment and management standpoint. In using this approach, a user would start by selecting a suitable virtual application pattern based on the application type. This may be one shipped by IBM, such as the IBM Workload Deployer Pattern for Web Applications, or it may be one created by the user through the extensibility mechanisms built into the appliance. After selecting the appropriate pattern, a user would supply the web service application, define functional and non-functional requirements for the application via policies, and then deploy. The virtual application pattern and IBM Workload Deployer provide the knowledge necessary to install, configure, and integrate the middleware infrastructure and the application itself. Once deployed, a user manages the resultant application environment through a radically simplified lens provided by IBM Workload Deployer. It provides monitoring and ongoing management of the environment in a context appropriate for the application. This means that there are typically no administrative consoles (as in the case of the virtual application pattern IBM ships), and users can only alter well-defined facets of the environment. It is a substantial shift in the mindset of deploying and managing middleware applications.
Okay, with that explanation in the bag, let's revisit the diagram I inserted above. I hope it's clear that, all things being equal, virtual application patterns indeed provide the lowest TCO and shortest TTV because of the degree to which they encapsulate the steps involved in setting up complex middleware application environments. So, let's get back to my assertion that the customization control continuum really applies to the deployer and consumer. Why do I say that? It's simple. In the case of either the virtual system pattern or the virtual application pattern, the pattern composer has quite a bit of liberty in how they construct things. Sure, we enable you right out of the chute by shipping pre-built, pre-configured IBM Hypervisor Edition images, as well as pre-built virtual system and virtual application patterns. The key is though, that the IBM Workload Deployer's design and architecture also enables you to build your own patterns -- be they the virtual system or virtual application type. With anywhere from a little to a lot of work, you can build virtual system and virtual application patterns tailored to your use cases and needs.
At this point, you may be saying, "Well now you have really confused things! How am I supposed to decide what kind of patterns-based approach fits my needs?" I have some advice in that regard. First, map your needs to things that we enable with the assets you get right out of the box with IBM Workload Deployer. If your application fits into the functional scope of one of the virtual application patterns that we ship, use it. If you can support the application by using IBM Hypervisor Edition images, virtual system patterns, and custom scripts, do it. In this way, you benefit most from the value offered by IBM Workload Deployer. However, if you find that you cannot use any of the assets we provide right out of the box (e.g. you want to deploy your environment on software not offered in IBM Hypervisor Edition form or in a virtual application pattern), then ask yourself one simple question: "What do I want my user's experience to be?"
In this sense, I primarily mean a user to be a deployer or consumer of your patterns. You need to decide whether you favor the middleware infrastructure centric approach afforded by virtual system patterns, or if you prefer the application centric approach proffered by virtual application patterns. There is no way to answer this generically for all potential IBM Workload Deployer users. Instead, you have to look at your use case, understand what's available to help you accomplish that use case, and finally, decide on what you want your user's experience to be. I hope this helps!
As I have mentioned before, IBM Workload Deployer v3.0 introduces choices in pattern-based deployment models. One of those models, virtual system patterns, is a carry over from the WebSphere CloudBurst Appliance. When you use virtual system patterns in IBM Workload Deployer, you can take advantage of all of the techniques you put to use in WebSphere CloudBurst. This is certainly good news for current WebSphere CloudBurst users, but it goes a bit further. Instead of simply maintaining the status quo with virtual system patterns, which would have been reasonable considering the introduction of virtual application patterns, we chose to continue to expand on your customization options for this pattern deployment model. In particular, I want to discuss three new features in IBM Workload Deployer that may help you to better construct and manage virtual system patterns.
The first new feature is one that I have been eagerly awaiting. In the new version of the appliance, we provide you with the ability to specify part and script package ordering in your pattern. This means that, within the virtual system pattern editor, you can tell IBM Workload Deployer in which order to start the virtual machines in your pattern, and you can specify in which order to invoke the script packages within the pattern during deployment. This eliminates the need for special script invocation orchestration logic in your pattern (I had customers resorting to a semaphore like approach using a shared file system), and it allows you to be more declarative about the virtual machine bring-up process. There are constraints, specifically with the part ordering. Some images will impose an implied part start-up order that you cannot change. For instance, deployment manager parts in the WebSphere Application Server Hypervisor Edition image must start before custom node parts. The good news is the pattern editor will not allow you to specify a part start-up order that violates these constraints. The image below shows an example of the ordering view in the virtual system pattern editor.
Another new feature that may influence the way you build virtual system patterns is the introduction of Add-Ons. You can think of Add-Ons as special script packages that you can include in your virtual system pattern that perform system-level configuration actions. Specifically, you can include add-ons in your virtual system pattern to add an operating system user, add a virtual disk, or add a NIC during the deployment process. You include Add-Ons in your pattern by simply dragging and dropping them onto a part in your pattern, just as you do with script packages today. The difference between script packages and Add-Ons is that IBM Workload Deployer will ensure the invocation of all Add-Ons before any other scripts run during deployment. We include default Add-On implementations for adding a user, disk, and NIC.
The last new feature I want to talk about today has more to do with how you manage or govern the deployment of virtual system patterns. In WebSphere CloudBurst 2.0, we introduced the idea of Environment Profiles as a way to extend your customization reach into the deployment process. Initially, these profiles gave you the ability to directly assign IP addresses to virtual machines in your deployment, declaratively specify virtual machine naming formats, and easily split a single pattern deployment across multiple cloud groups. In IBM Workload Deployer, you will be able to use these same profiles to set resource consumption limits for pattern deployments. In particular, you will be able to set cumulative limits for virtual CPU, memory, storage, and software licenses used by deployments tied to a specific profile, thereby giving you finer-grained control over cloud resource consumption. The picture below shows the new resource limit aspects of environment profiles.
Virtual system patterns are key in the deployment model choices for IBM Workload Deployer. Not only did we carry the concept over from WebSphere CloudBurst to IBM Workload Deployer, but we made it even better. Expect this trend to continue!
In my opinion, declarative deployment models are key to the entire notion of Platform as a Service (PaaS). That is, users should concern themselves with what they want, but not necessarily how to get it. The PaaS system should be able to interpret imperatives from the user and automatically convert that to a running system. In this respect, I think the new virtual application pattern, and more specifically policies, in IBM Workload Deployer takes a giant leap toward a more declarative deployment model.
In IBM Workload Deployer, policies allow you to 'decorate' your virtual application pattern with functional and non-functional requirements. In other words, they provide a vehicle for you to tell the system what qualities of service you expect for your application environment. To put a little context around this discussion, let's examine the policies available in the virtual application pattern for web applications. Specifically, let's look at the four policy types you can attach to Enterprise Application, Web Application, and OSGI Application components in this pattern:
Scaling policy: When it comes to cloud, the first thing many folks think about is autonomic elasticity. Applications should scale up and down based on criteria defined by the user. Well, that is exactly what the scaling policy lets you do. You simply attach this policy to your application component, and then specify properties that define when to scale. First, you choose a scaling trigger from a list that includes application response time, CPU usage, JDBC connection wait time, and JDBC connection pool usage. After choosing your trigger, you decide the minimum and maximum number of application instances for your deployment, and then you choose the minimum number of seconds to wait for an add or remove action. At this point, you can deploy your application and IBM Workload Deployer will monitor the environment, automatically triggering scaling actions as needed.
JVM policy: I would be willing to bet that nearly all of you tune the JVM environment into which you deploy your applications. The JVM policy allows you to take two common tuning actions, setting the JVM heap sizes and passing in JVM arguments, as well as attach a debugger to the Java process (especially useful in development and test phases). You can also use the policy to enable verbose garbage collection (invaluable to understanding heap usage patterns for your application) and select the bit level (from 32 or 64) for your application. Again, all you have to do is attach the policy and specify the properties. IBM Workload Deployer will take care of the required configuration updates.
Routing policy: The routing policy provides a simple way to specify virtual hostnames and allowable protocols (HTTP or HTTPS) for your application. Attach the policy, provide the virtual hostname you want to use, select the desired protocols, and that's it! Remember, once you set the virtual hostname you will need to update your name server to map the hostname to the appropriate IP address.
Log policy: During the development and test phase, it is likely that you will want to enable certain trace strings in the application runtime. The log policy allows you to provide trace strings for your application, and it makes sure that the appropriate configuration updates occur in the deployed environment.
While this is not an exhaustive explanation of each of the policies above, I hope it gives you a basic idea of what they are and how to use them. To me, declarative deployment models are going to be a crucial part of making PaaS successful, so I am really excited about the notion of policies in IBM Workload Deployer. What do you think?
A few weeks ago, I had a conversation with a current WebSphere customer about the potential value they could derive from the use of IBM Workload Deployer. Right away, this customer saw value in the consistency that a patterns-based approach could afford them. It was clear that patterns eliminate the uncertainty that can make its way into even the best-planned deployment processes. Initially though, the customer questioned the value of being able to do fast deployments because, in their words, "We don't deploy WebSphere environments that often." So, we continued our discussion, and then they asked an important question that I encourage all of our users to ask: "Why don't we deploy our WebSphere environments more frequently?"
It is interesting to talk with our WebSphere users that have a long history with our products. Often times, they have been taking a shared approach to WebSphere installations for many, many years. They develop innovative approaches and isolation schemes that allow them to carve up a single WebSphere installation (cell) amongst multiple different application teams. This allows them to avoid having to setup a cell for each application deployment and saves them the associated time. However, having talked to many different users taking this approach, it is not without its challenges.
As was the case in the customer I mention above, users typically made trade-offs when electing for larger, shared cells. As an example, if you have multiple different application teams with different types of applications using a single cell, applying fixes and upgrades to that cell can be a lot more complex. After all, you now have to coordinate plans across a number of different teams and find a window that fits all of their needs. For the same reason, trying incremental function via our feature packs is much more arduous in these types of cells. Additionally, administrative controls become more complex since teams with varying needs all require administrative access. Admittedly, this gets simpler with newer fine-grained security models in WebSphere Application Server v7 and v8, but it still requires organizational discipline and process.
At this point I should be clear that I am not denigrating the shared cell approach. It can work well, and we have many facilities built into the WebSphere Application Server product to support that model. However, if you are using this approach and you find yourself stumbling too much for your own liking, then I would strongly suggest that you explore the patterns-based approach of IBM Workload Deployer. By deploying patterns that represent your WebSphere cells using IBM Workload Deployer, you can quickly and consistently setup multiple WebSphere Application Server cells to support the varying needs of your application teams. You will still avoid spending an inordinate amount of time installing and configuring cells as that is an automated part of pattern deployment, and your application teams will still get the resources they need. Further, this can liberate your application teams in terms of how they apply maintenance, install upgrades, and absorb new function in the form of feature packs.
I am not suggesting a complete pendulum swing in your approach to how you manage multiple application environments. There is definitely a happy medium in terms of how many cells you end up with. After all, you do not want to trade in one set of problems for the problem of managing way too many different cells. However, I do think that decomposing monolithic, multi-purpose cells into smaller, more purposeful cells can be beneficial. In the course of thinking about this different approach, you may come to the same conclusion that the customer I mention above did. IBM Workload Deployer's rapid deployment capabilities are indeed valuable if you take a slightly different view of current processes.