Re: [dns-privacy] Call For Adoption: draft-wing-dprive-dnsodtls

["Jørgen Hovland" <[e-mail]>]

> Any chance you could add in a straight UDP round trip measurement so we have a benchmark to compare like with like?

8,9955 average

8.966 ms
9.020 ms
9.007 ms
8.984 ms
9.011 ms
8.985 ms

using traceroute with also 18 byte data (20 ip + 8 udp + 18 data)

For the crypto/openssl results, the server had CPU AES hardware support (E5-2690 v2) while the client had not (AMD 64 X2 Dual Core 5000+).



At 18:06 27/05/2015 (UTC), Phillip Hallam-Baker wrote:

Any chance you could add in a straight UDP round trip measurement so we have a benchmark to compare like with like?

On Tue, May 26, 2015 at 3:27 PM, Jørgen Hovland <mail?p=compose&to=[e-mail]" rel="nofollow">[e-mail]> wrote:

At 03:04 26/05/2015 (UTC), Paul Hoffman wrote:
> Is the latency for an established TLS connection any worse than for a DTLS connection? It would be good to see numbers if this is the case.

I did a test. The difference seems to be so small that it doesn't matter.

However,
DTLS handshake seems to be a bit slower than TLS handshake (not including TCP connect or DTLS cookie code).
Additionally, if I use the openssl CLI as DTLS server instead of my code, the DTLS handshake increases to around 34ms.
TCP connect takes around 9.1ms between my test-servers.

If you assume that the openssl CLI correctly implements a normal DTLS handshake while my code doesn't (due to no cookie code), TLS is faster than DTLS by around 2ms (8.4%) when a full connect/handshake with one write+read is done.


Server/client 500km apart from each other with simple echo server/round trip test:

TCP data read: 18, time spent write+read: 9.16038ms. Time spent TLS handshake 23.3289ms.
TCP data read: 18, time spent write+read: 9.07797ms. Time spent TLS handshake 21.4217ms.
TCP data read: 18, time spent write+read: 9.10591ms. Time spent TLS handshake 21. 1549ms.
TCP data read: 18, time spent write+read: 9.05283ms. Time spent TLS handshake 21.1289ms.
TCP data read: 18, time spent write+read: 9.02824ms. Time spent TLS handshake 21.2273ms.
TCP data read: 18, time spent write+read: 9.06903ms. Time spent TLS handshake 21.1971ms.
TCP data read: 18, time spent write+read: 9.07378ms. Time spent TLS handshake 21.3058ms.

UDP data read: 18, time spent write+read: 9.02713ms. Time spent DTLS handshake: 23.1482 ms.
UDP data read: 18, time spent write+read: 9.06344ms. Time spent DTLS handshake: 22.3601 ms.
UDP data read: 18, time spent write+read: 8.97628ms. Time spent DTLS handshake: 23.1001 ms.
UDP data read: 18, time spent write+read: 9.14725ms. Time spent DTLS handshake: 22.7562 ms.
UDP data read: 18, time spent write+read: 9.0721ms. Time spent DTLS handshake: 22.0508 ms.
UDP data read: 18, time spent write+read: 9.34728ms. Time spent DTLS handshake: 21.7002 ms.

(disclaimer: I assume my numbers are correct)


At 03:04 26/05/2015 (UTC), Paul Hoffman wrote:

On May 25, 2015, at 6:54 PM, Guangqing Deng <mail?p=compose&to=[e-mail]" rel="nofollow">[e-mail]> wrote:
> Resolution latency is very crucial for DNS system and the latency of DNS-over-DTLS is relatively low compared with DNS-over-TLS.

Is the latency for an established TLS connection any worse than for a DTLS connection? It would be good to see numbers if this is the case.

--Paul Hoffman
_______________________________________________
dns-privacy mailing list
mail?p=compose&to=[e-mail]" rel="nofollow">[e-mail]
https://www.ietf.org/mailman/listinfo/dns-privacy" target="_blank" rel="nofollow">https://www.ietf.org/mailman/listinfo/dns-privacy

_______________________________________________
dns-privacy mailing list
mail?p=compose&to=[e-mail]" rel="nofollow">[e-mail]
https://www.ietf.org/mailman/listinfo/dns-privacy" target="_blank" rel="nofollow">https://www.ietf.org/mailman/listinfo/dns-privacy