I frequently read about how Puma-based DOCSIS modems have severe latency problems. Back in 2016, a user on DSL Reports who states that he is an "EE" that "designs electronics" did an analysis and claimed that he found very frequent and repeatable ~250 ms latency spikes due to his modem's Puma chipset. (It is important to note that this person did not claim to have professional or academic credentials in networking.)
The data he published - a repeatable and consistent 250 ms latency - seems so bizarre that I find it almost impossible to fathom that his findings were typical behavior, or that ISPs would certify such poorly behaving modems. Later in 2016 an article was published that amplified his claims and his story took off as "the experience of everyone with a Puma-based modem".
By April 2017, an article was published in The Register stating that someone was pursuing a class action lawsuit against Arris related to Puma-based modems. [The way I read it, Arris won the essentials of the lawsuit.]
By August 2018, The Register reported that a fix was released, but the documents cited by the "bug fix" article don't appear to discuss the latency issue, but instead a DoS vulnerability. It is unclear how the article's author concluded that the security bug fix was associated with the Puma latency claim.
Now, years later, the "Puma latency claim" is still knocked around as a truth regarding the Puma chipset. But is it true?
I can't repeat it.
"Trust, but verify" is an often repeated proverb. Sadly, many people trust, and many fewer verify.
I can't repeat the findings of the original forum poster's claim. Although the original claim makes it clear that the problem is extraordinarily evident, my deep analysis shows.... nothing.
Since I have a bunch of modems under my control that are Puma-based, specifically Arris SB6190s, I decided to instrument them. I performed a latency test for months on active Internet connections to remote servers through these modems.
 |
| My Puma modem shows very low latency, wildly different than the 2016 claim |
I didn't find any latency issues that look anything even remotely close to resembling the claims of the original forum post. Instead, I see very consistent pings between 10 and 23 ms, with an average around 11.5 ms. And this is over an actively used Internet connection from my home LAN client to my ISP's remote infrastructure.
I conclude that the Puma issue does not exist. The originally reported data may have been due to infrastructure issues at an ISP, data collection issues, a software bug, a hardware failure, a DoS, a saturated network connection, or something else. But whatever it was, it does not manifest itself on any of my Puma configurations of today.
So instead of the original claim, I have a updated claim: Puma: impressively low latency performance.