ycsb sdk 3 ssl 15% throughput reduction

Avg Throughput (ops/sec), Workload A, 3 nodes, 12 vCPU, SSL

http://showfast.sc.couchbase.com/#/timeline/Linux/kv/ycsb/ycsb3

1027 - http://perf.jenkins.couchbase.com/job/hercules/10358/console - 557768

1038 - http://perf.jenkins.couchbase.com/job/hercules/10556/console - 459803

http://cbmonitor.sc.couchbase.com/reports/html/?snapshot=hercules_710-1027_access_daf7&snapshot=hercules_710-1038_access_ab99

http://changelog.build.couchbase.com/?product=couchbase-server&fromVersion=7.1.0&fromBuild=1027&toVersion=7.1.0&toBuild=1038&f_asterixdb=off&f_backup=off&f_build=off&f_cbas=off&f_cbas-core=off&f_cbauth=off&f_cbbs=off&f_cbgt=off&f_couchbase-cli=off&f_eventing=off&f_eventing-ee=off&f_forestdb=off&f_go-couchbase=off&f_goxdcr=off&f_indexing=off&f_kv_engine=on&f_magma=off&f_n1fty=off&f_ns_server=off&f_phosphor=off&f_plasma=off&f_platform=off&f_product-metadata=off&f_query=off&f_query-ee=off&f_testrunner=off&f_tlm=off

Three things I see from the graphs - connections in timewait state are much lower for 1038, and ep_dcp_replica_items_remaining in 1038 is much more gradual in the beginning of the test, and free memory remains higher and pagecache memory remains lower for 1038. I see commits related to tls configuration as well as memory allocation managment. Could be a combined result but I am running more tests to get exact build of regression. Will post when I have the data.