Several built-in modes are predefined by the Verity engine, each of which is designed to support a different indexing behavior. For any one collection, the application can implement one or more built-in or custom indexing modes.
Generic Mode (generic)
The generic
mode is the base mode from which all other modes inherit their behaviors. It is optimized to give the average overall performance without assuming anything about the desired indexing rates of documents, how many searches are occurring simultaneously, and so on. generic
mode is not very efficient at performing any particular optimization in a short amount of time. It does not perform advanced search optimizations such as creating spanning word lists or squeezing deleted documents.
generic
mode (named "generic") is the default mode if no default mode is specified in the style.plc
file for all applications or in VdkCollectionOpenArgRec
for custom applications. Also, the generic
mode is the default mode if the style.plc
file does not exist at all.
fastsearch
mode is optimized to index documents so that retrievals happen as quickly as possible. This mode causes the Verity engine to do more work at indexing time.The Verity engine performs the following optimizations for the
fastsearch
mode:
fastsearch
mode is equivalent to setting the following metaparameters:
bulkload
mode is for indexing large numbers of documents in large batches with bulk modify/bulk update mechanism. It is primarily intended to create new collections from a large amount of pre-existing documents. The bulkload
mode inherits most of its settings from the fastsearch
mode.The Verity engine performs the following optimizations for the
bulkload
mode:
bulkload
mode is equivalent to setting the following metaparameter:
newsfeedidx
mode is optimized to accept a moderate number of documents in a short amount of time where the documents arrive in frequent small batches. It is designed keep up with the high arrival rates of news feeds without falling behind in the indexing. Designed to index incoming documents and perform small merges for partitions of up to 100 documents each. These small partitions are not optimized VDBs, since optimization of such small partitions would incur significant overhead.
The
newsfeedidx
mode sets the following metaparameters:If you are developing an indexing application using the VDK API, the following service levels must be set for the session with the
newsfeedidx
mode: VdkServiceLevel_Index
, VdkServiceLevel_Optimize
.
newsfeedopt
mode is designed to perform background work that the newsfeedidx
mode does not. Both modes are designed to be used together.What the
newsfeedopt
mode does is merge partitions (components of collections) that the newsfeedidx
mode creates into large and optimized partitions. This mode ensures fast search performance by:
newsfeedopt
mode:
newsfeedopt
mode: VdkServiceLevel_Optimize
, VdkServiceLevel_DBA
, VdkServiceLevel_Delete
.
readonly
mode is not an indexing mode in the sense that it does not affect how indexing occurs. It disables data writes to the collection. This mode is useful for accessing a collection on a read-only medium such as CD-ROM.