SQL Server FILESTREAM has some known limitations.
1) Database mirroring does not support FILESTREAM.
2) For clustering, FILESTREAM filegroups must be put on a shared disk which defeats the purpose of creating the clusters.
Given these limitations in FILESTREAM is it advisable to build a FILESTREAM solution. Im looking to save and retrieve .5 million files in a FILESTREAM database (approx 1TB of disk size) which would be accessed simultaneously by approx 2000 users. Given the fact that the FILESTREAM cannot be clustered or mirrored how does one devise a scalable solution.
If I live with a non scalable solution what would be the performance of such a system. Can I serve up say 100 users with 100 1 MB files within a 5 second window?
Reality check: Issue 2 is a non-issue. In a cluster ALL data must be on shared discs, otherwise the cluster failover can not access the data. If that defeats the purpose fof a cluster you are invited to install a SQL Server cluster without shared discs. ALL data storage on clusters must be on shared discs. Has been like this since cluster service was first created for windows.
Which basically makes your conclusions already quite – hm – wrong.
You also need to make some mathmatics. 5th grade style.
Ignore SQL Server for a moment. Depending on how I read this thi is either 100mb or 10.000mb. Anyhow, 100mb in 5 seconds = 20mb per second, which runs around 200mbit. This is serious traffic. We taalkg of minimum 250 to 300mbit needed external bandwidth.