New-age Transactional Systems - Not Your Grandpa's OLTP
John Hugg discusses high volume transaction processing applications with high and low frequency profiles, and how VoltDB can be used for that purpose.
The content has been bookmarked!
There was an error bookmarking this content! Please retry.
Posted by Miko Matsumura on Aug 16, 2006
Event Driven Architecture (EDA) is a style of software architecture based on real time flows of (you guessed it) events. EDA is a buzzword was being pushed by Gartner as far back as 2003. At the time, Roy Schulte of Gartner went so far as to say that in SOA, connecting services occurs in a linear, predictable sequence, whereas an event-driven architecture allows for multiple, less predictable, asynchronous events to happen in parallel and trigger a single action.
Not that many folks got excited about EDA as a buzzword (most seem to think it's just part of SOA), so Gartner Analyst Yefim Natis teamed up with Oracle to coin the phrase SOA 2.0 (which appears to be another way of reintroducing EDA) another term which got a few folks excited, but not in a good way. Despite the criticism (Mark Little of JBoss posted a notable criticism), EDA enthusiasts continue to develop architecture.
Whatever it is ultimately called, dynamic real-time systems involving the integration of services will continue to be of interest to SOA practitioners. An excellent article about Event-Stream-Processing tools (ESP)is provided here. This article states:
ESP enables an event-driven SOA to decipher causal (if A is followed by B followed by C), temporal (within four seconds), and spatial (within 10 feet) relationships among events - and can do so in real-time. This kind of "enterprise wiretap" lets a business continuously analyse key performance indicators in real-time, identify threat and opportunity in real-time, and act instantly. These capabilities require a new style of computing - stream computing - that can deliver the missing link between an event-driven SOA and real-time business insight.
The introduction to the article is savvy enough to point out that existing terms such as SOA and Enterprise Service Bus (ESB) are capable to answer these needs, but the article goes on to describe a well researched set of concepts that are in application at some large companies. Another article details how to implement EDA using Mule, an open source ESB.
EDA is an implementation style (software architecture) in the same way as REST or RPC are implementation styles (software architecture). However SOA is an architectural style that aims to think of business and technology as services and enable the right selection of the implementation mechanisms. Which could be events, messages, RPC, winged monkeys, shouting or anything you want.
Going beyond EDA to agent and goal driven systems is just another implementation option for SOA, what doesn't change is the Service in its abstract sense or the need for services to communicate or distribute information.
The question is whether SOA = RPC and technology or whether SOA is a bigger shift at the true architectural level (pre-sofware). I'll go for the later. Taking the OASIS SOA Reference Model as a base definition there is nothing that procludes agent, EDA, RPC or winged monkey implementations at the software level.
In my mind services and events are related but not alternatives and I do not think that mentioning one as a better form of the other (which is subliminally implied since it's in the famous upper right corner) is helpful.
A service can have input, output, preconditions and effect and is between clients and a provider. Some services have effects and thus when executed cause a state transition. Messaging is on a logical level between designated parties.
At the moment an event is visible in the integration space it is the effect of a state transition. This notification can be input or output from a service. An event is not related to it's provider or it's handler. Messaging can be multiplied to many handlers.
If a problem space is more about things that happen and taking action the matching architecture can be called EDA, if it's more about business goals and clients and a provider the matching architecture can be called SOA.
EDA is going to become much more prevalent in the coming years, but only for certain types of applications. It is a great way to support state machines, particularly in distributed architectures (SEDA). It is also handy for breaking up problems that have been traditionally impossible to implement transactionally, just due to the overall scope of the transactions (too big, too long). I for one am very excited about the EDA field, and some of the related fields (AI, agent-based programming, etc.)
Peace,
Cameron Purdy
Tangosol Coherence: Clustered Shared Memory for Java
This discussion reminds me of the debate about SOA and BPM. Clearly taking an SOA approach makes it easier/more effective to use BPM to build process automation and composite applications. Equally SOA provides a great platform for EDA or Complex Event Processing. Making functionality availble as services and thinking about how to manage them, control them, version them etc etc will make it much easier to use that functionality to handle events or to be part of a process.
JT
Enterprise Decision Management Blog
However SOA is an architectural style that aims to think of business and technology as services and enable the right selection of the implementation mechanisms. Which could be events, messages, RPC, winged monkeys, shouting or anything you want.
I want shouting winged monkeys, please.
Cameron, would you care to elaborate?
If I understand correctly SEDA is a form of pipes-and-filters architecture with load-balancing routing. I can imagine this is an interesting approach for workitem- or order-oriented functionality. Every stage delivers products of a certain status and it does not matter which unit consumes the 'half-fabricate'. I'm not sure wether this really equates to an event since only one party consumes the item, while event notifications seems to me to be inherently to be noticed by possibly many parties....
You mention EDA as a solution for transactions that are too long-running or too big. What does EDA offer (and how?) that orchestration does not?
I wonder wether EDA and transactions are really related; transactions are about the steps that should take place as parts of a larger step, it's imperative and quite procedural. Events are loosely coupled and the step that causes an event does not care about which step takes place as a result. Aren't events and transactions orthogonal....?
event-driven architecture is broader than its relationship with SOA, including real-time information flow and analysis, and complex event processing. Here's an excerpt of my event-driven architecture overview paper.
I recognize a huge business potential of asynchronous, event oriented design. And I believe that EDA by its nature will be the paradigm to realize the ultimate alignment of business processes and the supporting IT-systems. I think that the ultimate layer between our real world events and artificial application constructs will be an EDA. And I recognize the possibilities offered by the current IT-technology evolutions to support this paradigm.
See my thoughts here...
John Hugg discusses high volume transaction processing applications with high and low frequency profiles, and how VoltDB can be used for that purpose.
Kevlin Henney examines code samples to see what can be learned from them starting from the premise that one won’t write great code unless he knows how to read it.
Jason Ayers share the observations he made watching a team of developers collaborating in real time on the same code base, pushing XP, pair programming and continuous integration to their extremes.
Michael Snoyman presents Yesod, a web framework written in Haskell and containing a web server, templating, ORM, libraries (templating, gravatar, etc.).
Richard Kreuter and Kyle Banker on how to avoid classical RDBMS transactional systems by using compensation mechanisms, transactional messaging or transactional procedures.
Attila Szegedi talks about performance tuning Java and Scala programs at Twitter: how to approach GC problems, the importance of asynchronous I/O, when to use MySQL/Cassandra/Redis, and much more.
One category of risk that project teams need to ensure they address is business value failure – delivering a product that fails to provide value for the business investor.
InfoQ spoke to the authors of Software Systems Architecture on a couple of new topics, the System Context viewpoint and Agile, which have been added to the second edition.
8 comments
Watch Thread Reply