I believe that the issue is that your analysis does not take into account how many servers of each type there are. To illustrate this point, let's revisit your example. Suppose (for illustration purposes only) that there are 1000 servers with netspeed 1/1000 of the maximum and a single server with the maximum netspeed. Let's denote the total load by X. With NTPv4, each of the "slow" servers should experience (1/2000)X load while the "fast" server should experience X/2 load. As you said, if F=10 (the frequency of Khronos watchdog queries), the increase in load induced by Khronos here is ~0.3X. This increase is distributed equally, and so each server now carries ~0.3X/1000 of the additional load (around 60% increase on the slow servers).
In the NTP pool, the fraction of servers with low netspeed (say <50) constitutes a very considerable fraction of the overall pool, which is why the disastrous scenario you mentioned does not occur (see Figure 12). For such a scenario to occur you would need almost all servers to have the maximum netspeed and for the remaining few to have the minimum netspeed. In addition, Kronos' overhead load can easily be reduced further since even setting F=100 yields considerable security benefits (VI.F) and other simple measures are also possible (e.g., limiting Khronos to sampling servers whose netspeed exceeds a certain threshold).
I hope this clarifies.
On Wed, Jul 12, 2023 at 10:56 AM Miroslav Lichvar <mlichvar@xxxxxxxxxx> wrote:
On Wed, Jul 12, 2023 at 09:35:05AM +0300, Michael Schapira wrote:
> Khronos can certainly be used with normal pool zones without introducing
> meaningful overhead. No fast servers are required. The reason is that (1)
> Khronos queries are much rarer than NTPv4's (see analysis),
The draft says:
In each Khronos poll interval, the Khronos client selects, uniformly
at random, a small subset (e.g., 10-15 servers) of a large server pool
(containing hundreds of servers). While Khronos queries around 3
times more servers per polling interval than NTP, Khronos's polling
interval can be longer (e.g., 10 times longer) than NTPv4, thereby,
minimizing the load on NTP servers and the communication overhead.
So, if all clients used Khronos with the 10x longer interval, they
would add about 30% to the pool load (I'm ignoring the panic mode for
now). But that extra load is spread evenly across all servers,
bypassing the pool weighting. The servers that have their speed set to
the 1/1000th of the fastest servers would not get only 30% extra load,
but something closer to 10000% of extra load.
Where am I making error in my analysis?
--
Miroslav Lichvar
-- last-call mailing list last-call@xxxxxxxx https://www.ietf.org/mailman/listinfo/last-call
