Hello Chris, thanks for your postings, they are a great starting point for me to assume a good performance of this database. However I have some questions left. Lets say I have a node store and a property store. Both of them are individual files. I am going to implement a GIS application which fetches OSM data from the hard disk. When I load a bounding box I want to avoid to many random reads from the disk. More precisely I want to load nodes and properties inside a given bounding box. It would be great if both the nodes and the properties are organized in successive blocks. Is there one id-pool for both nodes and properties, so that I can load for example the nodes with id 1 and 2 and the properties 3, 4 and 5 with one block read? I can be totally wrong because if I save a new node file with id 1, 2 and then save a new property file with id 3, it will start on a new block (windows block size like 4K). When then writing a new node id it would be saved in the first block I guess. What about fragmentation? And is there an improvement when using a Linux system (inodes? I don't know Linux well)? When I am finished with saving the nodes and properties is there some way of reorganization on the hard disk? Lets say I want to enter a new node which is connected to a low id. Will it get the first free id (and it will be saved on the other end of the harddisk perhaps) or does it just get an allready used id and the following records will be reorganized (insert performance)? Maybe I am totally wrong about this, but I would appreciate an efficient way of storage for GIS data.
best regards, Daniel -- View this message in context: http://neo4j-community-discussions.438527.n3.nabble.com/Neo4j-low-level-data-storage-tp3336483p3402827.html Sent from the Neo4j Community Discussions mailing list archive at Nabble.com. _______________________________________________ Neo4j mailing list [email protected] https://lists.neo4j.org/mailman/listinfo/user
