Hello,
I've tested with one customer the PDF Indexer with compression (meaning removing of ressources like AFP). And we could get something like 70% compression on our PDF samples.
The customer couldn't accept the solution for one single reason: PDF/A. Once PDF/A gets through the PDF Indexer and the ressources are splitted, then the PDF is a normal PDF without any certificate, all the added feature of PDF were also removed (like custom PDF indexes), etc... etc... So the retrieved PDF was from a visual perspective identical, but the content of the PDF itself was different.
The version was PDF 1.3 (normal for PDF/A) and the retrieved file was a PDF 1.8 or 1.9 don't remember...
So if you are accepting such behavior, then it's interesting, if you cannot accept that the PDF are lighter... then you must look elsewhere.
Concerning the performance, we didn't go that far so cannot say anything... maybe somebody else has something to say on that topic too? :-D
Sincerely yours,
Alessandro