I would start by collecting sequences coming from Bos taurus (e.g. RefSeq, UniProt in case of proteins) and use the 'formatdb' tool that comes with blast to create a blast db out of it.

The Ensembl genome builds are more logically organized by species. I recommend to look here: ftp://ftp.ensembl.org/pub/release-64/fasta/bos_taurus/ Most likely you'll want the file at cdna/Bos_taurus.UMD3.1.64.cdna.all.fa.gz

These are ordinary fasta files that should be suitable for local blast.

None of the resources on the page to which you link are specific for Cow.

You should look at the Cow Genome Resources and grab those resources that most closely match your needs, then you can build the BLAST databases you need using BLAST+'s makeblastdb.

None of these.

Bos Taurus reference sequences:

Transcriptome: ftp://ftp.ncbi.nih.gov/refseq/B_taurus/mRNA_Prot/cow.rna.fna.gz

Proteome: ftp://ftp.ncbi.nih.gov/refseq/B_taurus/mRNA_Prot/cow.protein.faa.gz

Bos Taurus genomic assemblies:

btau4.2: http://www.hgsc.bcm.tmc.edu/ftp-archive/Btaurus/fasta/Btau20080815/LinearizedChromosomes/

UMD3.1: ftp://ftp.cbcb.umd.edu/pub/data/assembly/Bos_taurus/Bos_taurus_UMD_3.1/ *Get chromosomes individually

Additionally, the adjusted UMD3.1 transcriptome/proteome are available at ftp://ftp.cbcb.umd.edu/pub/data/assembly/Bos_taurus/Bos_taurus_UMD_3.1/annotation/

Alternatively you could extract all Bos taurus sequences from NCBI databases NR/NT using the taxid 9913. You can do that online at the public instance but it can also easily be done locally (See the blast documentation on how to easily do that on local DBs (not at hand, sorry)).

NCBI taxonomy browser will give you what you want, for instance all proteins.

Write a filter program that detects "Bos" "taurus" "cow" and the like in the FASTA headers and pass the NCBI sequence collection through it. Then just run BLAST locally against the filtered set. Most of NCBI databases are not species specific.