IIS vs Kestrel performance comparison

Question

How does the performance of IIS look like comparing to a Kestrel http server?

Seems like Kestrel is significantly inspired by the family of asynchronous and event-driven server frameworks. In the meantime, IIS has a long history of development and obviously more mature in terms of features. Taking all that into account, I'm specifically looking for a comparison of memory utilization, thread utilization, request related comparison (like request-per-second) and streaming capabilities.

That's simply impossible at this stage when Kestrel is not even released. The performance of Kestrel alone is practically not comparable to IIS, as you need to also put its front end nginx into consideration, and that requires lots of tuning before a good result is achieved. — Lex Li, Dec 24 '15 at 02:44
@LexLi not sure that I understand what exactly is impossible :) Kestrel is a part of new generation of asp.net tools and used as standalone server. So it *is* available, since new asp.net is available. Taking about performance, not sure why do I need to put a front-end server for testing - the idea is to test performance impact of event-driven approach introduced by Kestrel (and libuv) and IIS core — madcyree, Dec 28 '15 at 16:58
It is unfair to compare a full feature production web server with a limited feature hosting server. Consider Jetty vs. Kestrel vs. node.js, and they are similar. — Lex Li, Dec 28 '15 at 23:55
It is possible. It runs and so does IIS therefor making them available for benchmarking. — bbqchickenrobot, Feb 02 '16 at 03:04
It should be noted that these are not representative of real world performance. IIS is a reverse proxy and security-hardened, the equivalent would be Nginx + Kestrel, as was pointed out several times by Damien Edwards from the ASP.NET team, they recommend using Nginx as a frontend, as you need one to e.g. restart Kestrel and provide security. So: in reality, noone runs lonely Kestrel in production, I'd like to see how Nginx actually degrades Kestrel perf. We'll see.. — hko, Feb 20 '16 at 01:58

score 20 · Accepted Answer · edited Sep 01 '20 at 19:41

20

Here is the XSLX file from the GitHub repo where you can see the performance comparisson you are looking for.

Open the file on Excel and check the "Latest" tab.

Update: The excel file is no longer available and has been replaced with an interactive Power BI dashboard.

edited Sep 01 '20 at 19:41

numX

830
7
24

answered Feb 18 '16 at 22:06

Matias Quaranta

13,907
1
22
47

1

That benchmark in particular is "plaintext", there are several other scenarios that get benchmarked. They are all scenarios from [TechEmpower benchmarks](https://www.techempower.com/benchmarks/) – Matias Quaranta Apr 02 '16 at 02:01
2

There is no xslx file in the github repo. – Biswajit Panday Jul 07 '20 at 07:06
1

Has the Excel file been moved somewhere else? – Mr_Hmp Aug 23 '20 at 06:34

score 6 · Answer 2 · edited Jul 23 '18 at 11:54

6

As of Nov 2016, straight from the ASP.NET benchmarks.

On Windows Server 2012, at a pipeline depth of 16:

stack        on       RT        requests/sec
ASP.NET 4.6  IIS      CLR           57,792 
ASP.NET 5    Kestrel  CoreCLR    1,188,521

That's a 20x or 2000% speedup. I understand a full blown IIS vs standalong Kestrel but I do hope someone on the ASP.NET team can deep dive on this because the difference is tremendous.

That and why it's so much slower on Linux.

edited Jul 23 '18 at 11:54

surfmuggle

5,527
7
48
77

answered Nov 14 '16 at 06:32

DeepSpace101

13,110
9
77
127

5

Cool. But no one use Kestrel without any http servers as proxy which can handle a strong level of security. So as a result Kestrel is useless in production. – Alexander Pavlenko Mar 02 '17 at 13:56
2

Why Kestrel on Linux has low rps vs on Windows? – Soheil Alizadeh Nov 27 '17 at 18:17
1

I agree. The difference is so large that I'd bet money something was not setup right on IIS. – TheLegendaryCopyCoder Jun 06 '18 at 10:39
2

I find it extremely hard to reach any conclusion from this image. Between both cases marked with arrows, the ASP.NET framework is different, the CRL is different and the webserver is different. One can only guess where the performance difference comes from. :( Also, there are two Kestrel installations with the exact same setup but completely different performance ratings. Looking at this, a huge `WTF` is all that pops in my mind... Maybe I am missing something. Are there any newer statistics for the latest framework? – user2173353 Apr 01 '19 at 07:42

Milan Vidakovic · Answer 3 · 2018-02-13T14:36:01.607

6

I did quite a bit of benchmarking for my current project, hosting both .net core 1.0 and 2.0 apps on IIS and Kestrel. The tests were real-world rest-api CRUDs with authentication/authorization, logging, metrics, rate limiting etc. Also, .net core apps were done completely by the MS book, complying to recommended 1.0/2.0 standards.

With the same hardware setups, hosting behind IIS constantly served around 40% more requests per second. I'm still unable to find an article or a consultant who is able to explain the performance difference.

I also tried to find any sort of optimization tips from the core benchmarking authority https://github.com/aspnet/benchmarks, by rummaging through settings and service initializers, but still, IIS was just faster.

Any pointers?

edited Feb 13 '18 at 14:36

answered Feb 13 '18 at 14:29

Milan Vidakovic

436
5
10

2

so, 2 years later, what's your position on this subject ? – Proviste Mar 04 '20 at 21:13
Switched to Core and never looked back. IIS still smokes kestrel in real-world benchmarks (even more so from v3), but it hasn't really been an issue. Two years later, I have 3 nation-wide public services running in production, and performance-wise they trump everything i've seen so far (using Spring as my reference). A single service instance serves around 200k users (around 1k concurrent) and a response median is less than 80ms. – Milan Vidakovic Mar 06 '20 at 07:58
how do you manage load balancing, bindings, things, restarting the app if it crashes ? (the things I rely on IIS for) edit : thank you for your feedback – Proviste Mar 07 '20 at 18:44
2

Oh, when I said it hasn't been an issue, I meant that solo kestrel was never an option. I had the local Microsoft office code review my stuff, and I got an official confirmation that Kestrel was production ready but not recommended for publicly accessible services. Since my services are all public, reverse proxy was a must (as well as request filtering etc) so initially, IIS was the way to. Last year I switched to linux+nginx with substantial performance boost. So, my current stack is cloudflare -> nginx -> custom rest gateway -> rest api. – Milan Vidakovic Mar 07 '20 at 19:03

score 0 · Answer 4 · answered Feb 02 '16 at 03:10

Here are some preliminary Kestrel benchmarks by the asp.net team. While not as exhaustive as what you were looking for you will find they post some information regarding RPS. I'm sure more will come as they get closer to a release. IIS benchmarks you can find on the internet with a simple google search:

https://github.com/aspnet/benchmarks

score 0 · Answer 5 · answered Jul 15 '19 at 08:16

As for why in-process hosting in IIS is faster (taken straight from the docs):

Using in-process hosting, an ASP.NET Core app runs in the same process as its IIS worker process. In-process hosting provides improved performance over out-of-process hosting because requests aren't proxied over the loopback adapter, a network interface that returns outgoing network traffic back to the same machine.

See Hosting Models.

IIS vs Kestrel performance comparison

5 Answers5

Linked