What do you mean by copy number variation data? Do you mean data from which your algorithm will generate CNVs, or do you want already generated CNV data outputs, such as a segmentation analysis which you then process further with your algorithm? Some of the suggestions above point to good resources, but you need to specify what you want more clearly in order to get an answer suited to your particular needs. One can generate CNV data from SNP arrays such as Affymetrix arrays (SNP 6.0, CytoScan, U133, etc). These you can find on GEO. However, you may want to analyze CNVs from whole genome sequencing data, in which case you would get that from different database. Most cancers will have lots of CNVs, so you could get sequencing data from TCGA.