Comment
Author: Admin | 2025-04-27
To use the “workstation GC” or “server GC”.By default, an application uses the workstation GC, though some environments (like ASP.NET) opt-in to using server GC automatically. You can explicitly opt-in in a variety of ways, including by adding true into your project file (as we did in the Benchmarking Setup section of this post). Workstation GC optimizes for reduced memory consumption, while server GC optimizes for maximum throughput. Historically, workstation employs a single heap, whereas server employs a heap per core. That typically represents a tradeoff between amount of memory consumed and the overhead of accessing a heap, such as the cost of allocating. If a bunch of threads are all trying to allocate at the same time, with server GC they’re very likely to all be accessing different heaps, thereby reducing contention, whereas with workstation GC, they’re all going to be fighting for access. Conversely, more heaps generally means more memory consumed (even though each heap could be smaller than the single one), especially in lull periods where the system might not be fully loaded, yet is paying in working set for those extra heaps.The decision for which to use isn’t always so clear. Especially in the presence of containers, you frequently still care about really good throughput, but also don’t want to be spending memory uselessly. Enter “DATAS,” or “Dynamically Adapting To Application Sizes”. DATAS was introduced in .NET 8 and serves to narrow the gap between workstation and server GC, bringing server GC closer to workstation memory
Add Comment