William Vambenepe's blog

I have nothing against Animoto. From what I know about them (mostly from John’s podcast with Brad Jefferson) they built their system, using EC2, in a very smart way.

But I do have something against their story being used to set the benchmark for infrastructure flexibility. For those who haven’t heard it five times already, the summary of “their story” is ramping up from 50 to 5000 machines in a week (according to the podcast). Or from 50 to 3500 (according to the this AWS blog entry). Whatever. If I auto-generate my load (which is mostly what they did when they decided to auto-create a custom video for each new user) I too can create the need for a thousands of machines.

This was probably a good business decision for Animoto. They got plenty of visibility at a low cost. Plus the extra publicity from being an EC2 success story (I for one would never have heard of them through their other channels). Good for them. Good for Amazon who made it possible. And who got a poster child out of it. Good for the facebookers who got to waste another 30 seconds of their time straining their eyes. Everyone is happy, no animal got hurt in the process, hurray.

That’s all good but it doesn’t mean that from now on any utility computing solution needs to support ramping up by a factor of 100 in a week. What if Animoto had been STD’ed (slashdoted, technoratied and dugg) at the same time as the Facebook burst, resulting in the need for 50,000 servers? Would 1,000 X be the new benchmark? What if a few of the sites that target the “lonely guy” demographic decided to use Animoto for… ok let’s not got there.

There are three types of user requirements. The Animoto use case is clearly not in the first category but I am not convinced it’s in the third one either.

1. The “pulled out of thin air” requirements that someone makes up on the fly to justify a feature that they’ve already decided needs to be there. Most frequently encountered in standards working groups.
2. The “it happened” requirements that assumes that because something happened sometimes somewhere it needs to be supported all the time everywhere.
3. The “it makes business sense” requirements that include a cost-value analysis. The kind that comes not from asking “would you like this” to a customer but rather “how much more would you pay for this” or “what other feature would you trade for this”.

When cloud computing succeeds (i.e. when you stop hearing about it all the time and, hopefully, we go back to calling it “utility computing”), it will be because the third category of requirements will have been identified and met. Best exemplified by the attitude of Tarus (from OpenNMS) in the latest Redmonk podcast (paraphrased): sure we’ll customize OpenNMS for cloud environments; as soon as someone pays us to do it.

This post started as a comment on the blog of Van Wiles. When it became too long (and turned into a therapeutic rant at the end) I turned it into a blog post of its own. Please, read Van’s post first. Here is my response to him:

Hi Van. Sounds like what you are after is not a mapping of the CIM_Dependency association to a CMDBf record type (anyone can make up such a mapping as you point out), but a generic algorithm to map any CIM association to a corresponding CMDBf relationship record type. Correct? That algorithm needs to handle the fact that the CIM metamodel has the concept of relationship roles while the CMDBf metamodel doesn’t.

Here is a possible such mapping:

1. Take a CIM association (called “myAssociation”) that has two roles (called “thisOne” and “theOtherOne”).
2. Take the item that has role name that comes first alphabetically and make it the source (in this example, it is “theOtherOne”)
3. Take the item that has role name that comes second alphabetically and make it the target (in this example, it is “thisOne”)
4. Generate a CMDBf record type called “{associationName} _from_ {firstRoleNameAlphabetically} _to_ {secondRoleNameAlphabetically}”

You’re done. The new CMDBf record type is “myAssociation_from_theOtherOne_to_thisOne”, the source is the item with the role “theOtherOne” and the target is the item with the role “thisOne”. Everyone who follows this algorithm (of course it needs to be formally defined and evangelized, there is no guarantee here unless we bake CIM-specific concepts in the core CMDBf specification, which would be a mistake) will produce the same CMDBf relationship record type for a given CIM association.

Applied to the CIM_Dependency example, this would generate a “CIM_Dependency_from_Antecedent_to_Dependent” CMDBf record type, in which the source is the CIM Antecedent and the target is the CIM Dependent.

Alternatively, you can have the algorithm generate two CMDBf relationship record types (one going in each direction) for each CIM association. So you don’t have to arbitrarily pick the first one (alphabetically) as the source. But then you need to have model metadata to capture the fact that these relationships are the inverse of one another (and imply one another). As you well know,I have been advocating for the use of RDF/RDFS/OWL in CMDBf for a while. :-)

In the end, there are three potential approaches:

1) Someone (the CMDBf group or someone else) creates an authoritative mapping for all CIM associations (or at least all the useful ones) and we expect anyone who uses the CIM model with CMDBf to use that mapping.

2) Someone (again, the CMDBf group or someone else) defines a normative CIM to CMDBf mapping, e.g. the one above, and we expect anyone who generates a CMDBf relationship record type from a CIM association to use this mapping algorithm. From a pure logical perspective, it is the same as defining a CMDBf record type for each CIM association (approach 1), but it is less work and it doesn’t have to be updated every time a CIM association is created/versioned. At the cost of uglier (more arbitrary) CMDBf record types being defined.

3) We let people define the relationships in whatever way they choose and we provide a model metadata framework (aka ontology language) to allow mappings between these approaches. For example, you define, in your namespace, a van:CIM-inspired-dependency CMDBf record type that goes from antecedent to dependent. Separately, I defined, in my namespace, a william:CIM-like-dependency CMDBf record type that carries the same semantics (defined, not so precisely BTW but that’s a different topic, by CIM) except that its source is the dependent and its target is the antecedent. The inverse of yours. A suitable ontology language would allow someone (you, me, or a third party who has to assemble a system that uses both relationship types) to assert that mine is the inverse of yours. Once this assertion is captured, a request for any [A]—(van:CIM-inspired-dependency)—>[B] would also return the instances of [B]—(william:CIM-like-dependency)—>[A] because they are known to be the same. And you know how I am going to conclude, of course: OWL (specifically owl:inverseOf) provides just this.

BTW, approach 3 is not incompatible with 1 or 2. Whether or not we define mappings for CIM relationships and whether or not that mapping gets adopted, there will be plenty of cases in a federated scenario in which you need to reconcile models (CIM-based or not). Model metadata (aka an ontology language) is useful anyway.

Readers who only care about the technical aspects and have little time for rants can stop reading here. But, since I haven’t addressed any constructive criticism to the DMTF in a while, I can’t resist the opportunity to point out that if the mailing list archives for the DMTF working groups were publicly available, we wouldn’t have to have these discussions on our personal blogs. I am very glad that Van posted this on his blog because it is a question that many people will have. Whatever the CMDBf specification ends up doing, developers and architects who make use of it will benefit from having access to the deliberations and considerations that resulted in the specification being what it is. There are many emails in the CMDBf mailing list private archive that I am sure would be useful to future CMDBf implementers, but if they don’t show up on Google they don’t exist for any practical purpose. When grappling with the finer points of some specification or programming language I have often Googled my way into email archives (or old specification drafts) of the working groups that designed them. Sometimes I come out thinking “oh, ok, now I understand why they chose that approach” and other times it’s “ok, that’s what I suspected, these guys were high”. Either way, it’s useful to me as a user of the specification. W3C is the best example (of making working group records available, not of being high): not only is the mailing list available but the phone meetings often have a supporting IRC channel in which key points of the discussion get captured and archived. Here is an example. Making life easier for implementers is probably the single most important thing to make a specification successful. And ultimately, that’s the DMTF’s success too.

And it’s not just for developers and architects. It also impacts industry observers and pundits. Like the IT Skeptic who looked into CMDBf and reported “nothing on the DMTF website but press releases. try to find anything by navigating from the homepage”. And you wonder why his article is titled “the CMDB Federation proceeeds (sic) at its usual glacial pace”. There is good work going on, but there is no way for him to see it. This too is bad for the adoption and credibility of DMTF specifications.

Isn’t it ironic that the DMTF expends resources to sponsor a “hospitality suite” at the Burton Group Catalyst conference (presumably to spread the word about the good work taking place in the organization) but fails to make it easy for the industry to see that same good work taking place? It’s like a main street retail shop that advertises in the newspaper but covers its store window with cardboard, preventing passersby from seeing what’s on offer. I notice that all the other “hospitality suites” seem to be staffed by for-profit vendors (Oracle, IBM, Cisco, Microsoft etc are all there). Somehow W3C and OASIS (whose work is very relevant to some of the conference themes, like identity management and SOA) don’t feel the need to give away pens and key chains at the conference.

Dear DMTF, open source is not just good for code.

Anyone interested in application manageability and/or management integration should read about Jean-Francois Denise’s prototype for RESTful Access to JMX Instrumentation. Not (at least for now) as something to make use of, but to force us to think pragmatically about the pros and cons of the WS-* stack when used for management integration.

The interesting question is: which of these two interfaces (the WS-Management-based interface being standardized or the HTTP-centric interface that Jean-Francois prototyped) makes it easier to write a cross-platform management application such as the poker-cheating demo at JavaOne 2008?

Some may say that he cheated in that demo by using the Microsoft-provided WinRM implementation of WS-Management on the VBScript side. Without it, it would have clearly been a lot harder to implement the WS-Management based protocol in VBScript than the REST approach. True, but that’s the exact point of standards, that they allow such libraries to be made available to assist implementers. The question is whether such a library is available for your platform/language, how good and interoperable that library is (it could actually hinder rather than help) and what is the cost to the project of depending on it. Which is why the question is hard to answer in absolute. I suspect that, even with WinRM, the simple use case demonstrated at JavaOne would have been easier to implement using straight HTTP but that things change quickly when you run into more demanding use cases (e.g. event notification with filters, sequencing of large responses into an enumeration…). Which is why I still think that the sweetspot would be a simplified WS-Management specification (freed of the WS-Addressing crud for example) that makes it easy (almost as easy as the HTTP-based interface) to implement simple use cases (like a GET) by hand but is still SOAP-based, which lets it seamlessly enter library-driven territory when more advanced features are added (e.g. WS-Security, WS-Enumeration…). Rather than the current situation in which there is a protocol-level disconnect between the HTTP interface (easy to implement by hand) and the WS-Management interface (for which manually implementation is a cruel - and hopefully unusual - punishment).

So, Jean-Francois, where is this JMX-REST work going now?

While you’re on Jean-Francois’ blog, another must-read is his account of the use of Wiseman and Metro in the WS Connector for JMX Agent RI.

As a side note (that runs all the way to the end of this post), Jean-Francois’ blog is a perfect illustration of the kind of blogs I like to subscribe to. He doesn’t feel the need to post all the time. But when he does (only four entries so far this year, three of them “must read”), he provides a lot of insight on a topic he really understands. That’s the magic of RSS/Atom. There is zero cost to me in keeping his feed in my reader (it doesn’t even appear until he posts something). The opposite of what used to be conventional knowledge (that you need to post often to “keep your readers engaged” as the HP guidelines for bloggers used to say). Leaving the technology aside (there is nothing to RSS/Atom technologically other than the fact that they happen to be agreed upon formats), my biggest hope for these specifications is that they promote that more thoughtful (and occasional) style of web publishing. In my grumpy days (are there others?), a “I can’t believe United lost my luggage again” or “look at the nice flowers in my backyard” post is an almost-automatic cause for unsubscribing (the “no country for old IT guys” series gets a free pass though).

And Jean-Francois even manages to repress his Frenchness enough to not take snipes at people just for the fun of it. Another thing I need to learn from him. For example, look at this paragraph from the post that describes his use of Wiseman and Metro:

“The JAX-WS Endpoint we developed is a Provider<SOAPMessage>. Simply annotating with @WebService was not possible. WS-Addressing makes intensive use of SOAP headers to convey part of the protocol information. To access to such headers, we need full access to the SOAP Message. After some redesigning of the existing code we extracted a WSManAgent Class that is accessible from a JAX-WS Endpoint or a Servlet.”

In one paragraph he describes how to do something that IBM has been claiming for years can’t be done (implement WS-Management on top of JAX-WS). And he doesn’t even rub it in. Is he a saint? Good think I am here to do the dirty work for him.

BTW, did anyone notice the irony that this diatribe (which, by now, is taking as much space as the original topic of the post) is an example of the kind of text that I am glad Jean-Francois doesn’t post? You can take the man out of standards, but you can’t take the double standard out of the man.

[UPDATED 2008/6/3: Jean-Francois now has a second post to continue his exploration of marrying the Zen philosophy with the JMX technology.]

I got a sneak peak at CMDBf v2 today.

I am calling it v2 based on the assumption that the one being currently standardized in DMTF will end up being called 1.0 (because it’s the first one out of DMTF) or 1.1 (to prevent confusion with the submitted version).

At the Semantic Technology Conference, David Booth from HP presented his work (along with his partner, Steve Battle from HP Labs) to provide a SPARQL front-end to HP’s Universal CMDB (the engine under what was the Mercury MAM product). Here are the slides.

The mapping from SPARQL to TQL (the native query interface for UCMDB) was made pretty easy by the fact that TQL is a graph-oriented query language. How much harder would it be to similarly transform a CMDBf (v1) query interface into a SPARQL query interface (and vice-versa)? Not much. The only added difficulty would come from the CMDBf XPath constraints. TQL has a property value mechanism that is very similar to CMDBf’s “propertyValue” constraint and maps well to SPARQL functions. The introduction of XPath as a constraint language in CMDBf makes things harder. It could be handled by adding XPath support to the SPARQL engine using function extensibility. Or by turning the entire XML into RDF and emulating XPath in SPARQL. But in either case, you’ll have impedence mismatch at some point because concepts such as element order that exist in XPath have no native equivalent in RDF.

The use of XPath in selectors on the other hand is not a problem. HP’s prototype uses Gloze (available as a Jena package) to turn the XML returned by UCMDB into RDF. An XSLT transform could turn that same XML into a CMDBf-valid XML response instead and that XSLT could easily handle the XPath selectors from the query request. This is another reason why constraints and selectors should remain separate in CMDBf (fortunately the specification is back to doing this properly).

Here is why I call this prototype CMDBf v2: The CMDBf effort (v1 or 1.1), in its current form of re-inventing a graph query, can succeed. Let’s assume the working group strikes a reasonable balance between completeness and complexity, and vendors choose to compete on innovation and execution rather than lock-in (insert cynical comment here). CMDBf may then end up being supported by the main CMDB vendors. It wouldn’t provide federation capabilities, but having a common CMDB query interface supported by the Big Four would help with management integration. And yet, while the value would be real, it would only provide a little help to solve a larger problem:

• As a technology limited to IT systems management, it would be unlikely to see widely available tools (e.g. user consoles and language-specific libraries).
• It wouldn’t get the kind of robustness and interoperability that comes from wide adoption. While pretty similar, there might be some minor differences in the various implementations. Once your implementation has been tweaked to work with the implementations from the Big Four, you’ll call it done. Just like SNMP, another technology that is specific to IT systems management (see it happen here).
• Even if it works perfectly at the query level, it will just hasten the time when developers run into the real problem, model interoperability. CMDBf doesn’t help at all with this. In fact, it makes it harder by hard-coding some dependencies on an XML back-end (the XPath constraints).

In the long run, IT management has to become more automated and integrated. That’s a given. The way it happens may or may not go through CMDB-like configuration stores. But if it does, we’ll have to eventually move beyond CMDBf (v1) towards something that addresses the three requirements above. And federation. I don’t know if it will be called CMDBf v2, and/or if it will come from the DMTF (by then, the CMDBf brand might be an asset or a liability depending on developer experience with the specification). But I strongly suspect (”probability 0.8″ as a Gartner analyst might put it) that it will use semantic technologies. Because the real, hard, underlying problem is a problem of semantic integration. In that sense, David and Steve’s prototype is a sneak peek at what will come after CMDBf v1/1.1.

Pretty much since the beginning of CMDBf I have been pushing for it to ideally embrace SPARQL (with no success) or to at least stay close to it conceptually in order to make the eventual mapping/evolution smooth (with a bit more success). This includes pushing for a topological query language, trying to keep XML idiosyncrasies at bay and keeping constraints and selectors cleanly separated. Rather than working within the CMDBf group, David took the alternative approach of simply doing it. Hopefully this will help convince people of the value of re-using semantic web technology for IT systems management. Yes semantic technologies have been designed for a much more general use case. But the use cases that CMDB systems address are a subset of the use cases addressed by semantic technologies. It’s hard for domain experts to see their domain as just a subset of a larger problem, but this is the case here. Isn’t HTTP serving the IT management community better than a systems management-specific alternative would?

By the way, there is no inferencing taking place in the HP prototype. We are just talking about re-using an existing, well though-through graph query language. Sure OWL inferencing and some rules could be seamless layered on top of this. But this is in no way required to do (better) what CMDBf v1 tries to do.

And then there is the “federation” question. Who do you trust more to deliver this? A bunch of IT system management architects in DMTF or the web and query experts at W3C, HP Labs etc who designed and implemented SPARQL over many years? BTW, it sounds like SPQARL federation was discussed at WWW 2008, based on these meeting notes (search for “federation”).

I have been involved in various capacity with five different specifications that define a GET (or GET-like) operation that takes as input an XPath expression used to pinpoint the subset of the XML document that should be retrieved (here is a quick history as of a couple of years ago, more has happened since). And I must shamefully admit that all but one are simply impossible to implement in an interoperable way.

That’s because they instruct implementers to return an XPath nodeset in the response SOAP message but say nothing about how to serialize the nodeset. While an XPath nodeset contains the kind of things that make up an XML document, it is not an XML document by itself. There is an infinite number of possible ways to serialized an XPath nodeset into XML. To have any hope of interoperability on this, a serialization algorithm has to be clearly described by the specification. Which hasn’t happened.

Let’s start with WS-ResourceProperties (WS-RP). It has a QueryResourceProperties operation that takes an XPath expression as input. The specification says that “the response MUST contain an XML serialization of the results of evaluating the QueryExpression against the resource properties document“. Great, thanks. The example provided happens to return a nodeset with only one node (a boolean), which is implicitly serialized into the text representation of that boolean. What if there is more than one node in the nodeset? What about other types of nodes?

Moving on to WS-Management, which defines a SOAP header that uses XPath to qualify a WS-Transfer GET request such that it only retrieves a subset of the target XML document. While it does a better job than WS-RP at describing the input (e.g. it specifies the context node and what namespace declarations are in scope for the XPath evaluation) it is even more cavalier than WS-RP in describing the output: “the output (lines 53-55) is like that supplied by a typical XPath processor and might or might not contain XML namespace information or attributes“. By “a typical XPath processor” we should understand MSXML I suppose. But as far as I know a “typical XML processor” doesn’t return XML, it returns language-specific data structures (e.g. a C# or Java object, like a nu.xom.Nodes instance). And here too, the examples only use single-node nodesets.

WS-ResourceTransfer (WS-RT) was supposed to be the convergence of these two efforts, so presumably it would have learned from their mistakes. While it is better written in general than its predecessors, it fails just as badly with regards to specifying the nodeset serialization. And once again, the example provided uses a nodeset with just one node.

And then came the CMDBf query operation which, for some unclear reason, was deemed in need of a built-in XPath transformation of records. As I pointed out in my review of CMDBf 1.0 at the time, this feature was added without taking the pain to define the XML serialization of the resulting nodeset. And there isn’t even an example of the XPath serialization.

It is sad in a way, but the only specification that acknowledges the problem and addresses it came before any of the four above even got started. It is the WSMF (Web Services Management Framework) work that we did at HP, and more specifically the “note on dynamic attributes and meta information” (not available at HP anymore but available from archive.org) . This specification was the first one to define a GET operation that is qualified by an XPath expression. Unlike its successors it also explicitly narrowed down the types of nodes that could be selected (”The manager MUST NOT send as input an XPath statement that returns a nodeset containing nodes other than element, attribute and namespace nodes“). And for those valid types it described how to serialized them in XML (”When a node in the result nodeset is an attribute node, for the sake of the response it is serialized as an element node which has the same name as the name of the original attribute (see example 4 for an illustration). The element is in the same namespace as the namespace the attribute it represents is in. This applies to namespace nodes as well, they are serialized like an attributes in the xmlns namespace“). Turning an attribute into an element of the same QName might not be the smartest thing in retrospect (after all there may be an element by that QName already) but at least we recognized and addressed the problem.

But all is good now, I am told, because XPath 2.0 is here, along with a clean data model and a well-described serialization.

Not so. Anyone wanting to use XPath for a SOAP-based query language still would have to specify a serialization.

The first problem with the W3C serialization is that the XML output method doesn’t work for all nodesets. Try to use it on a nodeset that contains a top-level attribute node and you get error err:SENR0001. And even for the nodesets it accepts, it sometimes returns less-than-useful results. For example, if your XPath is of the form /employee/name/text() and you have four employees, the result will look something like this:

“Joe SmithKathy O’ConnorHelen MartinBrian Jones”

Concatenated text values without separators. I guess W3C is like a department store, they don’t offer complimentary wrapping anymore…

That’s why the nux.xom.xquery.ResultSequenceSerializer class had to define its own wrapping mechanims to produce a useful XML serialization. The API gives you the choice between the W3C_ALGORITHM and the WRAP_ALGORITHM.

Bottom line, and however much some would like to think of it that way, XPath (1 or 2) is not an XML subsetting/transformation mechanism. It could be used to create one (as XSLT does), but you have to do your own plumbing.

In addition to the technical aspects of this discussion, what else can be learned from this sad state of things? The fact that all these specifications define an XPath-driven query mechanism that is simply broken (beyond the simplest use cases) withouth anyone even noticing tells me that there isn’t a real need for full XPath query over SOAP (and I am talking about XPath 1.0, the introduction of XPath 2.0 in CMDBf is even more out there). A way to retrieve individual elements (and maybe text values) is all that is needed for 99% of the use cases addressed by these specifications. Users would be better served (especially in a version 1.0) by specifications that cover the simple case correctly than by overly generic, complex and poorly documented features. There is always time to add features later if the initial specification is successful enough that users encounter its limitations.

I just read the transcript of Jon Udell’s podcast interview of Kingsley Idehen. It’s almost two years old but it contains something that I have tried (and mostly failed) to explain for a while now, so maybe borrowing someone else’s words (and credibility) would help.

Kingsley says:

“A graph model, ideally, will allow you to explore almost all the comprehensible dimensions of the nodes in that network. So you can traverse that network in a myriad of different ways and it will give you much more flexibility than if you’re confined to a tree, in effect, the difference between XQuery and SPARQL. I always see the difference between these two things as this. If you visualize nodes on a network, SPARQL is going to get you to the right node. Your journey to what you want is facilitated by SPARQL, and then XQuery can then take you deeper into this one node, which has specific data that the graph traversal is taking you to.”

Nicely said, especially considering that this is not a prepared statement but a transcript of a (presumably) unscripted interview.

He later provides an example:

“Let’s take a microformat as an example. HCard, or an hCalendar, is a well-formed format. In a sense, it’s XML. You can locate the hCard in question, so if you had a collection of individuals who had full files on the network in the repository, it could be a graph of a social network or a group of people. Now, through that graph you could ultimately locate common interests. And eventually you may want to set up calendars but if the format of the calendar itself is well formed, with XQuery you can search a location, with XPath it’s even more specific. Here you simply want to get to a node in the content and to get a value. Because the content is well formed you can traverse within the content, but XQuery doesn’t help you find that content as effectively because in effect XQuery is really all about a hierarchical model.”

Here is one way to translate this to the IT management domain. Replace hCard with an XML-formated configuration record. Replace the graph of social relationships with a graph of IT-relevant relationships (dependency, ownership, connections, containment…). Rather than attempt to XQuery across an entire CMDB (or, even worse, an entire CMDB federation), use a graph query (ideally SPARQL) to find the items of interest and then use XPath/XQuery to drill into the content of the resulting records. The graph query language in CMDBf is an attempt to do that, but it has to constantly battle attempts to impose a tree-based view of the world.

This also helps illustrate why SPARQL is superior to the CMDBf query language. It’s not just that it’s a better graph query language, one that has received much more review and validation by people more experienced in graph theory and queries, and one that is already widely implemented. It also does something that CMDBf doesn’t attempt to do: it lets you navigate the graph based on the semantics appropriate for the task at hand (dependency relationships, governance rules, distributed performance management…), something that CMDBf cannot do. There is more to classification than simply class inheritance. I think this is what Kingsley refers to when he says “in a myriad of different ways” in the quote above.

Here is a way to summarize the larger point (that tree and graph views are complementary):

Me Tarzan, you Jena

Where Tarzan (appropriately) represents the ability to navigate trees and Jane/Jena represents the ability to navigate graphs (Jena, from HP Labs, is the leading open source RDF/OWL/SPARQL framework). As in the movie, they complement each other (to the point of saving one another’s life and falling in love, but I don’t ask quite that much of SPARQL and XQuery).

On a related topic, I recently saw some interesting news from TopQuadrant. Based on explicit requests from the majority of their customers, they have added capabilities to their TopBraid Composer product to better make use of the RDF/OWL support in the Oracle database. TopQuadrant is at the forefront of many semantic web applications and the fact that they see Oracle being heavily used by their customers is an interesting external validation.

[UPDATED 2008/03/05: more related news! The W3C RDB2RDF incubator group has started is life at W3C, chaired by my colleague Ashok Malhotra, to work on mappings between RDF/OWL and relational data.]

I just noticed this result from the 2007 DMTF member survey (taken a year ago, but as far as I can tell just released now). When asked what their “most important interoperability priority” is, members made it pretty clear that they want the current CIM/WBEM infrastructure fixed and polished. They seem a lot less interested in these fancy new SOAP-based protocols and even less in using any other model than CIM.

It will be interesting to see what this means for new DMTF activities, such as CMDBf or WS-RC, that are supposed to be model-neutral. A few possibilities:

• the priorities of the members change over time to make room for these considerations
• turn-over (or increase) in membership brings in members with a different perspective
• the model-neutral activities slowly get more and more CIM-influenced
• rejection by the DMTF auto-immune system

My guess is that the DMTF leadership is hoping for #1 and/or #2 while the current “base” (to borrow from the US election-season language) wouldn’t mind #3 or #4. I am expecting some mix of #2 and #3.

Pushing the analogy with current US political events further than is reasonable, one can see a correspondence with the Republican primary:

• CIM/WBEM is Huckabe, favored by the base
• CMDBf/WS-RC/WS-Management etc is Romney, the choice of the party leadership
• At the end, some RDF and HTTP-based integration-friendly approach comes from behind and takes the prize (McCain)

Then you still have to win the general election (i.e. industry adoption of whatever the DMTF cooks up).

[UPDATED 2008/2/7: the day after I write this entry, Romney quits the race. Bad omen for CMDBf and WS-RC? ;-) ]

While doing some research on the different ways to probe and squeeze business process definitions to extract insight relevant for IT management I ran into this very interesting paper: Querying Business Processes. It defines a query language (called BP-QL) to query process definitions. Not much in common with CMDB Federation at first sight, and CMDBf was not on my mind at the time. Until I looked at the description of the query language that the researchers came up with. It is strikingly similar to the CMDBf query language. This is not very surprising since both are graph-based query languages that rely on patterns (where the patterns mix topological aspects with constraints on node/link properties).

CMDBf is more complete in some respects. It supports properties on the relationships, not just the items. The “depthLimit” element provide more control than BP-QL’s double-headed edges. BP-QL has its own extra features, including support for joins (something we discussed in CMDBf and that could be added to the specification) and negation at the graph level (e.g. A and B are not connected by any relationship of type “foo”, which may be useful but one should remember that CMDB discovery is rarely guaranteed to be comprehensive so an open-world approach is often preferable).

Assuming a suitable CMDB model for business processes, a CMDBf-compliant CMDB should cover many of the simpler use cases addressed by BP-QL. And reciprocally, the more advanced features in BP-QL are not really specific to business process definitions (even though that’s the scope of the paper) and could well be applied to CMDBf. I was also very interested by the BP-QL “compact representation” and the implementation choices. I hadn’t heard of Active XML before, something to look into especially if, as the paper hints, it does a better job than XQuery at dealing with idrefs. And Active XML introduces some interesting federation (or at least distribution) capabilities that are not currently exploited by BP-QL but which I find intriguing and which reinforce the parallel with the declared goal of CMDBf.

Is this similarity between the query languages just an interesting pattern to notice? Or is there more to it? The parallel between BP-QL and CMDBf invites the question of whether one should model business processes in a CMDB. And if so, is a business process represented by just one CI or do you break it down into a model similar to the one the BP-QL query language works on? You would need to go that far if you wanted to use queries to the CMDB to answer questions such as those handled by the BP-QL engine. And by doing this in the context of a CMDB that contains a lot more than just process definitions, you’d be able to enrich the queries with considerations from other domains, such as application or host topology. Modeling business process steps/activities may seem like very fine-grained modeling for a CMDB, but isn’t this part of the sales pitch for federated CMDBs, that participants in the federation can provide different levels of granularity? Of course, CMDB federation might never work out. If it does work and if we use it that way, we are not talking about just supporting change management processes (which are more likely to take place at the level of the overall process definition than the individual step) but rather about management integration for a wide variety of use cases. If that means we need to drop the term CMDB along the way (and leave it for the sole usage of the IT process people), I am more than happy to oblige.

[UPDATE on 2008/01/11: Prof. Milo pointed me to this follow-up paper that proposes a similar looking query language except that this time it is targeted at monitoring process instances rather than analyzing process definitions. And the monitoring runs as a set of BPEL processes within the monitored BPEL engine. Her group is doing some very interesting work.]

I like XML. Call me crazy but I find it relatively easy to work with. Whether it is hand-editing an XML document in a text editor, manipulating it programmatically (as long as you pick a reasonable API, e.g. XOM in Java), transforming it (e.g. XSLT) or querying an XML back-end through XPath/XQuery. Sure it carries useless features that betray its roots in the publishing world (processing instructions anyone?), sure the whole attribute/element overlap doesn’t have much value for systems modeling, but overall it hits a good compromise between human readability and machine processing and it has a pretty solid extensibility story with namespaces.

In addition, the XML toolbox of specifications is very large and offers standard-based answers to many XML-related tasks. That’s good, but when composing a solution it also means that one needs to keep two things in mind:

• not all these XML specifications are technically sound (even if they carry a W3C stamp of approval), and
• just because XML’s inherent flexibility lets one stretch a round hole, it doesn’t mean it’s a good idea to jam a square peg into it.

The domain of IT management provides examples for both of these risks. These examples constitute some of the technical deficiencies of management-related XML specifications that I mentioned in the previous post. More specifically, let’s look at three instances of XML mis-use that relate to management-related specifications. We will see:

• a terrible XML specification that infects any solution it touches (WS-Addressing, used in WS-Management),
• a mediocre XML specification that has plenty of warts but can be useful for a class of problems, except in this case it isn’t (XSD, used in SML), and
• a very good XML specification except it is used in the wrong place (XPath, used in CMDBf).

Let’s go through them one by one.

WS-Addressing in WS-Management

The main defect of WS-Management (and of WSDM before it) is probably its use of WS-Addressing. SOAP needs WS-Addressing like a migraine patient needs a bullet in the head (actually, four bullets in the head since we got to deal with four successive versions). SOAP didn’t need a new addressing model, it already had URIs. It just needed a message correlation mechanism. But what we got is many useless headers (like wsa:Action) and the awful EPR construct which solves a problem that didn’t exist and creates many very real new ones. One can imagine nifty hacks that would be enabled by a templating mechanism for SOAP (I indulged myself and sketched one to facilicate mash-up style integrations with SOAP) but if that’s what we’re after then there is no reason to limit it to headers.

XSD in SML

The words “Microsoft” and “bully” often appear in the same sentence, but invariably “Microsoft” is the subject not the object of the bullying. Well, to some extent we have a reverse example here, as unlikely as it may seem. Microsoft created an XML-based meta-model called SDM that included capabilities that looked like parts of XSD. When they opened it up to the industry and floated the idea of standardizing it, they heard back pretty loudly that it would have to re-use XSD rather than “re-invent” it. So they did and that ended up as SML. Except it was the wrong choice and in retrospect I think it would have been better to improve on the original SDM to create a management-specific meta-model than swallow XSD (SML does profile out a few of the more obscure features of XSD, like xs:redefine, but that’s marginal). Syntactic validation of documents is very different from validation of IT models. Of course this may all be irrelevant anyway if SML doesn’t get adopted, which at this point still looks like the most likely outcome (due to things like the failure of CML to produce any model element so far, the ever-changing technical strategy for DSI and of course the XSD-induced complexity of SML).

XPath in CMDBf

I have already covered this in my review of CMDBf 1.0. The main problem is that while XML is a fine interchange format for the CMDBf specification, one should not assume that it is the native format of the data stores that get connected. Using XPath as a selector language makes life difficult for those who don’t use XML as their backend format. Especially when it is not just XPath 1.0 but also the much more complex XPath 2.0. To make matters worse, there is no interoperable serialization format for XPath 1.0 nodesets, which will prevent any kind of interoperability on this. That omission can be easily fixed (and I am sure it will be fixed in DMTF) but that won’t address the primary concern. In the context of CMDBf, XPath/XQuery is an excellent implementation choice for some situations, but not something that should be pushed at the level of the protocol. For example, because XPath is based on the XML model, it has clear notions of order of elements. But what if I have an OO or an RDF-based backend? What am I to make of a selector that says that the “foo” element has to come after the “bar” element? There is no notion of order in Java attributes and/or RDF properties.

Revisionism?

My name (in the context of my previous job at HP) appears in all three management specifications listed above (in increasing level of involvement as contributor for WS-Management, co-author for SML and co-editor for CMDBf) so I am not a neutral observer on these questions. My goal here is not to de-associate myself from these specifications or pick and choose the sections I want to be associated with (we can have this discussion over drinks if anyone is interested). Some of these concerns I had at the time the specifications were being written and I was overruled by the majority. Other weren’t as clear to me then as they are now (my view of WS-Addressing has moved over time from “mostly harmless” to “toxic”). I am sure all other authors have a list of things they wished had come out differently. And while this article lists deficiencies of these specifications, I am not throwing the baby with the bathwater. I wrote recently about WS-Management’s potential for providing consistency for resource manageability. I have good hopes for CMDBf, now in the DTMF, not necessarily as a federation technology but as a useful basis for increased interoperability between configuration repositories. SML has the most dubious fate at this time because, unlike the other two, it hasn’t (yet?) transcended its original supporter to become something that many companies clearly see fitting in their plans.

[UPDATED 2008/3/27: For an extreme example of purposely abusing XML technologies (namely XPath in that case) in a scenario in which it is not the right tool for the job (graph queries), check out this XPath brain teasers article.]

It’s now official, the CMDBf specification has been submitted to the DMTF and will be standardized there. Here is the press release and here is the specification (unchanged) republished on the DMTF site. The CMDBf working group was created a while ago at the DMTF but I didn’t report it since it wasn’t clear to me whether that was public information or not. The press release makes this clear now.

As a side note, this is one of my ongoing frustrations with the DMTF. Almost everything happens in private with no publicly-accessible URL until a press release comes out and of course lots of interesting things happen that don’t get a press release. I have heard many times that the DMTF is working on opening up the process, but I still haven’t seen much change. If this had been OASIS or W3C, the call for formation of the new working group would have been publicly accessible even before the group was created. OK, end of ranting.

As always, there isn’t much useful information to be gleaned from the text of the press release. Only that, as expected, the authors addressed the question of how this relates to CIM, since for many DMTF=CIM. So the press release proactively declares that the CMDBf work will not be limited to CIM-modeled configuration data. What this means in practice will be seen later (e.g. will there be CIM-specific extensions?).

Having seen how executive quotes for press releases get generated I hate to read too much into them, but another thing I can’t help noticing in the press release is that none of the quotes from the companies submitting the specification tout federation, but simply “integration” or “sharing”. For example: “integration and interoperability” (BMC), “share data” (CA), “sharing of information” (HP), “view, track and change information” (IBM), “exchange data” (Microsoft). This more realistic assessment of what the specification does stands in contrast to the way the DMTF presents it in the press release : “this specification provides a standard way to federate management data stored in multiple different data models”. At this point, it doesn’t really provide federation and especially not across different models.

All in all, it’s as good thing for this work to be moved to a standards organization. I may join the CMDBf group at the DMTF to track it, but I don’t plan to engage very much as this area isn’t my focus anymore now that I am at Oracle. But of course everything is linked at some level in the management field.

[UPDATE  on 2007/11/30: two days after posting this message I got the monthly DMTF newsletter which touches on points I raise here. So here are the relevant links. First, Mike Baskey, DMTF Chairman, shares his view on what CMDBf means for DMTF. Second, as if to respond to my rant on the opacity of the DMTF, Josh Cohen, DMTF Vice-chairman, gives an update on process improvements. Some progress indeed, but still a far cry from opening up mailing list archives so that observers can see in real time what issues are addressed and can go back in time to understand how a specific technical decision was made and what were the considerations.]