Comments on: Context Aware Event Processing

By: Marco

Marco — Thu, 05 Feb 2009 09:10:37 +0000

Sounds really interesting. It’s nice to see what kind of different concepts are used to provide a context. I think that there will be many different approaches to context management as it needs to be handled differently depending on the concepts used in various products.

For example in ruleCore I think you could describe the approach as “events that are somehow related” to describe what we mean by our event stream views, which are our concept to create and maintain a context.

By: Louie Lovas

Louie Lovas — Thu, 05 Feb 2009 01:28:03 +0000

Very interesting topic. In Apama, we have the notion of a “sub-monitor”, which can be thought of as a separate “instance” of a set of business logic. Each sub-monitor can load and maintain its own data (e.g. consider a sub-monitor as corresponding to a different situation, with different timeouts, key indicators, client data etc) and can set up its own listeners to key into a (different or overlapping) subset of the event space. This allows us to easily model “factory” like behavior, with each sub-monitor maintaining different state and possibly listening for different (sequences of) events – but applying the same (parameterized) basic logic (including the actions to take when an event condition becomes true). We call this our micro threading or “mThread” architecture.

In the latest Apama release – 4.1 – we actually go further than this and introduce the notion of contexts or silos of execution which take the factory model to the next level. ”context” is a reserved word in Apama 4.1 – it defines a collection of sub-monitors which can inter-operate in a protected address space, with strong semantics for inter-context communication (via event routing). It is also the unit of parallelization – allowing the same engine to spawn multiple “contexts” which key into the event flow but execute in parallel (on multi-core architectures). Contexts in Apama provide many benefits for building complex applications that can scale massively.

By: Marco

Marco — Sun, 01 Feb 2009 18:11:21 +0000

James,

I suppose that reacting only on high value customers could be done in two ways.

In the first approach we could create a semantically rich context which contains only those events which tells us something about the trucks of our most valued customers.

The other approach is to output an event out of our event processing system into a “traditional” system which could do a query or evaluate a rule to find out the value of the customer.

I think both approaches could be used depending on your specific requirement.

In ruleCore we tend to go for the first one. Creating a very expressive way of defining the semantics of the context so that we have the right events to work on to begin with.

Maybe there’s a third or fourth approach too?

By: James Taylor

James Taylor — Sun, 01 Feb 2009 03:51:56 +0000

Of course you would also need to make some business decisions – the example of the truck is great but any action you take is also going to depend on knowing the value of the customers whose deliveries are on the truck – a business decision.
Context+Event is great knowledge but you need to add business decision making too.

By: Marco

Marco — Wed, 21 Jan 2009 10:49:24 +0000

Thanks for the link to the paper, I’ll read it as soon as I have the patience to register on that site and go through the hassle of using my personal card to buy it

Context should indeed be in the glossary. Perhaps also some note about the difference between the windows and context concepts? I tend to spend much time in explaining how a window and a context differs.

By: Opher Etzion

Opher Etzion — Tue, 20 Jan 2009 20:09:01 +0000

Hi Marco. welcome to the context world. I agree that context is a major abstraction in event processing and proposed to add it to the EPTS glossary. We have made context as a major concept in AMiT – an abstract of a relatively old paper (from 2003) can be found here: http://www2.computer.org/portal/web/csdl/abs/proceedings/ams/2003/1983/00/19830160abs.htm

Temporal contexts are much richer than time windows.
Some more discussion in the notion of context from various aspects can be found here: http://epthinking.blogspot.com/search/label/context

Have fun and hope you dress well when going outside.