What's the read throughout to get the data back out, and does it scale to what you'd need to have N search indexes building on top of this shared crawl?