Hey Christopher: Thanks for the response. Yeah, some performance hit was expected, it was just the magnitude that we weren't expecting. Unfortunately, I don't have a great benchmark to tell you exactly what we saw, though we may have those numbers in the coming weeks. We decided to pursue wire encryption through another means, but I am hoping to review some of the numbers at a later date.
- Logan On Tue, Oct 17, 2023 at 12:42 AM Christopher <ctubb...@apache.org> wrote: > I think a performance hit is expected, due to the expected overhead of the > TLS handshake, and the number of connections Accumulo requires in order to > distribute work across a cluster. I think whether the overhead is tolerable > is a per user decision, and may also be dependent upon the details of your > application, table content, query patterns, hardware, and JVM support. I'm > sure it's not suitable for everybody's use case, but could be a useful > option in some circumstances. It's really hard to make general statements > about whether it's worthwhile, though, because of different people having > different requirements and environments. > > I am curious, though, if you could characterize the overhead you saw, as a > point of comparison. > > On Mon, Oct 16, 2023, 23:18 Logan Jones <lo...@codescratch.com> wrote: > > > Hello: > > > > I know that Accumulo has support for TLS. When turning on TLS support, we > > noticed some pretty serious performance hits as a result of turning this > on > > in 1.10.2. Does anyone actually have TLS turned on for larger clusters? > Are > > there any known performance problems with turning on TLS? > > > > Thanks in advance, > > > > - Logan > > >