Blob cache is to reduce traffic in b/w sql server and front ends. So a copy of frequently accessed files is kept on each WFE. Enabling blob cache also passes information in http headers to client browser to fetch the data from cache rather than round trips.
This needs publishing framework. You might have to increase RAM on each front end on average by 800 bytes per file. Also, server relative urls should be less than 160 characters to be blob cached (with a path in file system as less than 260 characters)
But sometimes large blob cache will reduce your performance rather helping with heavy loads. Blob cache index are written to hard drive periodically and before recycle. When this serialization is progress, it will affect the time taken to serve the client request.
It is even more devastating when app pool recycle corrupt indexes and files are no longer being fetched from blob cache ; this happens generally because – if timeout for recycle is lower than time taken to save the indexes on disk
Best practices for Blob cache:
1. Try to keep the content stable for live environments and blob cache should be planned for content with lesser changes, when blob is invalidated by Framework?
2. Better to add mime types for files being blob cached. Alternate in webconfig [ BlobCache node] use parameter browserFileHandling = “nosniff” [ possible values nosniff|strict ]
3. The content in blobcache folder should be managed by framework. Let it be created and managed by SharePoint.
4. Server relative urls should be less than 160 characters to be blob cached ( with a path in file system as less than 260 characters)
6. For large blob cache increase parameter WriteIndexInterval in webconfig [ BlobCache node]. Default is 60 . You might want to serialize it may after 1 hour [ 86400] for very large blob cache. { this setting is not recommended for small blob cache size }
7. Recommended time limits to shut down and start up web application is 300 seconds. Default is 90 seconds. For large blob cache you might even go up to 600 seconds.
8. Plan for sufficient disk space as per size of content.