WebSphere Technology Trends for 2008
gcuomo 060000X9CG Comments (9) Visits (5024)
Is that Agile Jerry doing a jumping back kick? (by SKC 2005)
Each year I outline a handful of technology treads that the WebSphere team is aggressively pursuing. The following article provides a brief outline of seven technology trends that are energizing the WebSphere team in 2008.
There are a few overarching thoughts to these trends – not the least to mention on continued maniacal focus on SOA. However, as I outline these trends a single thought is in the forefront of my mind – that is Agility.
So, here goes...
[+] Click to read the more...
In 2007, we have discussed the compelling relationship between SOA and REST. REST is an architectural style which is best exemplified by HTTP, the backbone of the web. REST embodies the principles of a service-oriented architecture, using the web as the SOA platform. The RESTful approach to SOA inherits the simplicity, power and pervasiveness of the web. The web is like the air; it surrounds us, is abundantly accessible and is intuitive to use. Hence, extending SOA using REST gives enterprises the possibility to unleash their enterprise’s content to the masses, with a very affordable, easy to use, well-understood platform – the web.
In 2008, the WebSphere brand is realizing RESTful SOA by:
A couple more things… the first relates to PHP. I am impressed with the work our team did in 2007 on building a PHP interpreter on our Java virtual machine. Towards the end of the year, we enabled two very popular PHP applications on Zero. (One was phpBBTM, the other I can’t say quite yet). I will elaborate on this in another blog entry… but I am excited about the possibilities that this opens up in 2008. The second is the work Jason McGee is starting on building the Project Zero toolset using Project Zero itself. This includes source edit and debug for Groovy as well as visual feed manipulation tools. This will be fun to watch evolve in 2008. And last, but not least, “BPM via the web” is emerging as a powerful and cost effective form of BPM. As Kyle Brown said to me today, the union of BPM on the Web will “re-emphasize the human in human tasks", which is a vital aspect of BPM. Again, a topic for future blogging.
I was quoted recently in SearchSOA, as saying that CEP is the next “big thing” for SOA and they will come together and be a “beautiful thing” in 2008. Let me explain why I feel this way.
In an Enterprise SOA, services can (and should) be instrumented to trigger “activities of interest” to both business and IT. The WebSphere portfolio has several products that effectively deal with simple event flows through our messaging (e.g., WAS’s platform messaging or MQ) and ESB (DataPower, WMB and WESB) products. I often imagine that the sea of events generated by your enterprise forming a cloud of untapped knowledge above your SOA. While actions can be taken on these individual events, an opportunity exists to look into the “event cloud” and aggregate, correlate and interpret relationships between multiple events and to detect patterns that ultimately can allow you to intelligently react to situations in a highly agile manner. We sometimes refer to this as “reactive SOA”.
Complex Event Processing introduces the dimension of time into the SOA equation. Windows of time can be defined, along with a set of events and event sequences (i.e., patterns). The CEP system observes the events as they enter the cloud and continuously match the events against the set of supplied patterns and rules that have been deployed within the cloud.
IBM’s event processing technology is referred to internally via its code name SCEPTRE (or Scaleable CEP TRansaction Engine), which is being jointly developed with the Tivoli and DB2 teams. SCEPTRE is comprised of several core components –
In 2008, WebSphere will exploit the SCEPTRE technology to focus on business events in the enterprise. This allows a reactive SOA environment to be deployed either stand-alone or more interestingly along-side today’s WebSphere SOA foundation products; including WESB, WebSphere Process Server, DataPower and WebSphere Business Monitor. We are also been experimenting with using Project Zero-based applications to expose activities in the event cloud as REST, ATOM and RSS feeds and services - pretty cool stuff.
Before I leave this topic, a couple of honorable mentions are in order. We continue to make solid progress on low latency and predictable response time processing within the WebSphere portfolio. In 2008 we continue to enhance two standout products in this space; WebSphere Real-time (RT) and MQ Low Latency Messaging (LLM). WebSphere RT will continue JVM improvements in the area of GC, JIT and Ahead-Of-Time compilation. We also have some very useful tools in development that allow engineers to profile their applications and better understand their application's processing time and sources of non-determinism. LLM is a welcomed addition to our already powerful messaging portfolio. LLM uses Reliable Multicast Messaging (RMM) for predictability and reliability delivery. We plan to better leverage LMM across our portfolio, from DataPower to our XTP and CEP technology stacks.
We are in an interesting phase of SOA, it has caught on, and services have been deployed in mass across the gamut of industry sectors. Enterprises are moving beyond the experimental stage of deploying loosely coupled services and the consumption of those services has become something that people rely on to run their business. Now the challenge is how to enhance these service offerings to manage them in real world situations. A new level of rigor is required to apply security, to enforce rights management, and to manage the life-cycle of these services. We need to collect our best practices and establish some “rules” to guide the way we manage our SOA for the long haul. This can be an incremental process, but in order to plan, we need to think about the subtleties surrounding ownership and dependency management.
For example, let’s say I need to add a feature to my service. Can I change it and update the live version? NO. Other organizations may have dependencies on my service. We're getting into the real life challenges around SOA which will require two fundamental things; Policies by which to Govern; and Governance, which is the application of these policies to support the business.
Policy expressions (e.g., operational, architectural and business) provide the language for the discussion between IT and Business. Technologically, we are utilizing industry standards to express policies, wherever and whenever possible because we all know that the SOA can cross technology implementations (JEE/.NET/Etc) and vendor products. XACML and WS-Policy are key standards that can be consumed by policy aware services across IBM and non-IBM components. Service repositories, like our WebSphere Services Repository and Registry (WSRR), are critical as they provide a “point of control” for the business. By decoupling these control points using policy standards, you have the “knobs” you need to agily tune your business and IT governance models.
In 2008, we are focusing on policy-enabling the WebSphere portfolio. We are putting the finishing touches on our WS-Policy and XACML support in DataPower, WebSphere Application Server, WSRR and WebSphere Business Services Fabric (WBSF) – such that they can discover and use common security and operational policies.
After over 10 years of working on WebSphere, I think my colleagues and I have discovered tran
The defining characteristics of Extreme Transaction Processing (XTP) are low/predictable latency, linear scalability and fault-tolerance - supporting high volumes of transactions.
There are four magic elements to WebSphere’s support for XTP – all of which are vividly available across the WebSphere product line. The elements are: Data Cache which is distributed, partition-able and transactional. This is the centerpiece of XTP. To complement the data cache, we have 3 managers that are all savvy to the cache and its partitions - Workload Manager distributes work to the cache partitions. A High Availability Manager ensures that the partitions are always available and an Autonomic Manager to repartition cache to optimize performance and utilization. That’s it!
WebSphere XD, anchored by ObjectGrid (OG), supports these features, and thanks to the handy work of Billy Newport and team, is supporting some ultra-high volume application in production today.
In 2008, we are going to continue our focus on XTP and look into applying more broadly across our portfolio. For example, in the CEP section, I talked about “event clouds”, we are implementing “the cloud” using OG. XTP-enabling our BPM stack, will probably follow soon after. We also hope (time permitting) to continue to collaborate with our friends from DB2 and enhance OG to include SQL capabilities in the form of StreamSQL and are investigating integration with SolidDB. OG is part of our real time story, playing the part of a latency shock absorber to external datasources for low latency and time critical applications. The OG team is also “going RESTful” in 2008. The REST interface will allow OG to interop with non-Java platform like, .NET, C/C++, PHP or any language supporting REST
Honorable mention in the XTP space is distributed Batch – or “XTP for off-line work”. Many of the same principles discussed in XTP apply to batch processing workloads. WebSphere XD already has some compelling support for distributed Batch processing.
The major batch innovation for 2008 in support of XTP patterns is parallel batch. The parallel batch feature (of WAS XD Compute Grid) provides a mechanism to create logical jobs comprised of multiple sub-jobs that execute in parallel. This allows for work partitioning and multi-processor, multi-machine concurrency. This can be combined naturally with OG for very high performance, scalable transaction processing. The XD dynamic WLM facilities ensures smart placement of the sub-jobs on available compute resources. The distributed folks have finally caught onto what the Mainframe folks have known for years… Okay maybe with a bit of help from the Mainframe folks.
Consolidating servers and natural resources (e.g., power and space) are certainly amongst the compelling aspects of virtualization. However, when we look at virtualization, we see this and more. While most of the known world is rightfully buzzing about server virtualization, I am excited about our innovations in the area of application virtualization, which we will expand on in 2008. You see, I think server virtualization, which uses a virtualization container to mask server resources including operating system, is an essential building block, but insufficient to bring our customers the ultimate efficiencies that are possible from virtualization. We believe the middleware and application tier needs to play intimately in the virtualization equation. When it does, we can go beyond consolidation and resource savings and start to improve the quality and robustness of the environment, and really reduce cost of ownership and operation.
In 2008, we have 3 activities well underway that I will outline.
Ruth Willenborg has been doing some very interesting work around building WebSphere-oriented virtual images. While I’m not crazy about the term “virtual appliance” (seems like a bit of an oxymoron and I like to save to word “appliance” for real hardware) many people refer to it as that. Ruth has been publishing recipes for how to virtualize WebSphere Application Server and stacked products (Portal, Process Server, etc.). This important building block will be used as we move closer towards appl
WebSphere XD is our solution to virtualizing a set of applications in a cluster. XD manages applications at a cluster-level using operational policies – hence, you need not worry about which machines your application is running on, or how many running instances are currently active. XD is focused on optimizations using application level metrics; for example, my payroll application is getting the required 1/2 second response time, rather then OS metrics (cpu or memory utilization) like the server virtualizaton managers on the market do today. Brian Martin and the XD team are hard at work adding virtual image capability, which will allow an XD cluster to not only contain physical machines, but also virtual machine images, which will further increase the overall resource utilization and robustness of your system.
We are putting it all together in an incubator project that we started late last year called IronTX. I will blog about this more in 2008. To give you a quick preview - we are putting the best of what XD, Application Virtualization, and our Power6/AIX systems have to offer - creating a WebSphere Data Center in a Rack.
WebSphere DataPower SOA Appliances have collectively emerged as one of the most effective and efficient means of building out a SOA. DataPower’s quick time to value, overall ease of use, low cost of operation and green energy footprint makes it a model product, representing the values in agility and simplification that is at the heart of our 2008 strategy.
With those accolades, there is only one thing left to do…. Build more appliances!
The first thing we are doing in 2008 is bringing the appliance up to date with the latest specs in CPU, memory and industrial engineering (IBM owns many patents in industrial engineering that greatly benefit the new DataPower design). The new boxes are tamper-proof powerhouses and will “gobble up” even the most stringent connectivity challenges.
Functionally, in 2008, we will continue our focus on integration with the System z environment; including security integration (RACF) and optimized integration with IMS, CICS and DB2.
DataPower is “front and center” for SOA management and policy enforcement. We will continue to work with our colleagues in Tivoli on integration with security and management tools using the latest industry standards like XACML and WS-Policy.
We are looking at the first new form factor for the appliance; the DataPower Blade. The DataPower blade will allow DataPower capability to be plugged directly into a IBM BladeCenter.
We are also looking to introduce two appliance functions – B2B Gateway appliance, which excels in industry transformations (e.g., SWIFT) and a low latency messaging appliance that excels at market data feed integration.
Honorable mentions – I am hoping in 2008 that we will have time to work on deeper integration with our WebSphere XD; specifically integrating feature from XD’s On Demand Router (ODR), we will see.
We have been saying that the key to a healthy SOA is collaboration between business and IT. We have a pretty complete set of SOA tools for IT. Over the past couple of years we have also been focusing on providing tools suitable for business users. Our best examples of such tools today are in our WebSphere Business Modeler and Monitor products. In 2008, we will continue this trend.
If business agility is a prime motivation for SOA, then business tools are the way this will be realized. How does a business analyst program his or her SOA? There are several ways of doing this for example; drawing and simulating business models in Modeler, adjusting Key Performance Indicators in Monitor. It seems that most business analysts that I know prefer programming using business rules – which is a form that is closer to how we behave in the non-electronic world. Separating business rules from your core services implementations is one of the most effective ways of reacting to changes in your business, and to foster more reuse across your SOA.
Today, WebSphere BPM products support sequential rules. This is a good place to start, but we are looking to do more, especially as we move into the areas of complex event processing. The addition of AptSoft tools start to propel our SOA tooling for business users beyond sequential rules.
I am also increasingly intrigued with inference-based technologies, having studied this for a year in graduate school. It seems to me that inferencing is closer to the way that people think in real-word situations, especially when these situations become complex – which is the case for most real world business problems.
We are enhancing our WebSphere BPM portfolio, AptSoft for example, with enhanced business rules technologies – which will more precisely help our business analysts participate in SOA and allow them to program their SOA, without needing to program in Eclipse.