Usually first question for DLP customers who want to used DLP mail prevent is “What will be impact on mail delivery?” Yes, DLP will introduce a latency in mail delivery but I will try to show you with below tests that it will not be noticeable for most end users.
Testing system
All test were performed on a Windows 2012 multi-tier environment and using SYMANTEC DLP v14.5MP1 solution. Mail were generated with a homemade system which has multithreaded mail generator in order to simulate set of mail servers and a “latency-meter” which will receive email after DLP analysis and it will compute global latency introduce by DLP.
System was configured to get a smooth email traffic of 10k email generated in over 40 minutes. Different policies were active on mail prevent using most of DLP detection techniques (DCM, IDM, EDM).
Results
Graphics below shows that traffic was quite flat and around 4 email/s. A single DLP mail prevent server is able to process this traffic without specific issue.
Graph below shows latency measured for all messages generated by the system. For most emails latency is lower than 1s.
Higher latency is observed for email with few Mb attachment size. But even there, most of them are processed in less than 5 seconds.
Apart of latency measurement, we have also checked server resources usage when traffic becomes higher. We observed that CPU is most impacted resource when traffic reach high level (during our test we have reached over 40 email per seconds for our single mail prevent server). Of course, network could also become a bottleneck if traffic became higher than available bandwidth. We did not reach limit in memory usage with our system, but as all software if you reach it you may start to use virtual memory on disk and performance will decrease.
All these tests may not be exactly in same configuration as your environment but it shows that mail prevent server will not induce a latency in your messaging system which could be noticeable by end users.