CimStream / News: Recent posts

There are new unit tests for associations that leverage the sblim client library. Found that much of the functionality was there but I still have a few things to work out. I really like the new testing framework and how it leverages a 3rd party client library. I'm also breaking the tests down into the various functional CIMOM units defined by DMTF. Looking at the specs I might be able to implement pulled enumerations without much trouble.

I'm beginning to port the unit tests to junit4. Being able to run the cimom directly and then attach with the sblim client really makes things easier and the result is far more readable. Its nice to know that the current setup works with an existing implementation. Was just reading a bit on the DMTF standard on generic operations. For some reason they specify that responses have to be all or nothing. I think that is counter to using http chunking. It really seems to be counter to best interest. Now the CIMOMs will have to be allow for possibly huge single chunk data transfers, yet the clients will still have to assume that sometimes the data will be truncated due to errors. All the more reason to use gziped json to get the file sizes as small as possible!

I've implemented experimental support of json and gzip. The results are phenomenal for large data sets. On my mac I produced and consumed over 300k association instances per second. Thats 121 times faster than the uncompressed XML format. The compression ratio is over 200 to 1. Of course compression is really dependent on the actually data. But when there are large data sets, and by that I mean 1000's or 10's of 1000's of items, the items usually are very regular. If not the items themselves then the supporting names for fields and such. Whats amazing is that gzip required only a little configuration. Its already part of the distribution of the app server.