Akka.Serialization.TestKit 1.4.25-beta

• .NETCoreApp 3.1

  • Akka.Tests.Shared.Internals (>= 1.4.25)

• .NETFramework 4.7.1

  • Akka.Tests.Shared.Internals (>= 1.4.25)

• net5.0

  • Akka.Tests.Shared.Internals (>= 1.4.25)

Maintenance Release for Akka.NET 1.4**
Akka.NET v1.4.25 includes some _significant_ performance improvements for Akka.Remote and a number of important bug fixes and improvements.
Bug Fixes and Improvements**
[Akka.IO.Tcp: connecting to an unreachable DnsEndpoint never times out](https://github.com/akkadotnet/akka.net/issues/5154)
[Akka.Actor: need to enforce `stdout-loglevel = off` all the way through ActorSystem lifecycle](https://github.com/akkadotnet/akka.net/issues/5246)
[Akka.Actor: `Ask` should push unhandled answers into deadletter](https://github.com/akkadotnet/akka.net/pull/5259)
[Akka.Routing: Make Router.Route` virtual](https://github.com/akkadotnet/akka.net/pull/5238)
[Akka.Actor: Improve performance on `IActorRef.Child` API](https://github.com/akkadotnet/akka.net/pull/5242) - _signficantly_ improves performance of many Akka.NET functions, but includes a public API change on `IActorRef` that is source compatible but not necessarily binary-compatible. `IActorRef GetChild(System.Collections.Generic.IEnumerable<string> name)` is now `IActorRef GetChild(System.Collections.Generic.IReadOnlyList<string> name)`. This API is almost never called directly by user code (it's almost always called via the internals of the `ActorSystem` when resolving `ActorSelection`s or remote messages) so this change should be safe.
[Akka.Actor: `IsNobody` throws NRE](https://github.com/akkadotnet/akka.net/issues/5213)
[Akka.Cluster.Tools: singleton fix cleanup of overdue _removed members](https://github.com/akkadotnet/akka.net/pull/5229)
[Akka.DistributedData: ddata exclude `Exiting` members in Read/Write `MajorityPlus`](https://github.com/akkadotnet/akka.net/pull/5227)
Performance Improvements**
Using our standard `RemotePingPong` benchmark, the difference between v1.4.24 and v1.4.24 is significant:
_v1.4.24_
```
OSVersion:                         Microsoft Windows NT 6.2.9200.0
ProcessorCount:                    16
ClockSpeed:                        0 MHZ
Actor Count:                       32
Messages sent/received per client: 200000  (2e5)
Is Server GC:                      True
Thread count:                      111
Num clients, Total [msg], Msgs/sec, Total [ms]
1,  200000,     96994,    2062.08
5, 1000000,    194818,    5133.93
10, 2000000,    198966,   10052.93
15, 3000000,    199455,   15041.56
20, 4000000,    198177,   20184.53
25, 5000000,    197613,   25302.80
30, 6000000,    197349,   30403.82
```
_v1.4.25_
```
OSVersion:                         Microsoft Windows NT 6.2.9200.0
ProcessorCount:                    16
ClockSpeed:                        0 MHZ
Actor Count:                       32
Messages sent/received per client: 200000  (2e5)
Is Server GC:                      True
Thread count:                      111
Num clients, Total [msg], Msgs/sec, Total [ms]
1,  200000,    130634,    1531.54
5, 1000000,    246975,    4049.20
10, 2000000,    244499,    8180.16
15, 3000000,    244978,   12246.39
20, 4000000,    245159,   16316.37
25, 5000000,    243333,   20548.09
30, 6000000,    241644,   24830.55
```
This represents a 24% overall throughput improvement in Akka.Remote across the board. We have additional PRs staged that should get aggregate performance improvements above 40% for Akka.Remote over v1.4.24 but they didn't make it into the Akka.NET v1.4.25 release.
You can [see the full set of changes introduced in Akka.NET v1.4.25 here](https://github.com/akkadotnet/akka.net/milestone/56?closed=1)
| COMMITS | LOC+ | LOC- | AUTHOR |
| --- | --- | --- | --- |
| 32 | 1301 | 400 | Aaron Stannard |
| 4 | 358 | 184 | Andreas Dirnberger |
| 3 | 414 | 149 | Gregorius Soedharmo |
| 3 | 3 | 3 | dependabot[bot] |
| 2 | 43 | 10 | zbynek001 |
| 1 | 14 | 13 | tometchy |
| 1 | 139 | 3 | carlcamilleri |