dslreports logo
 
    All Forums Hot Topics Gallery
spc
uniqs
51

NetFixer
From My Cold Dead Hands
Premium Member
join:2004-06-24
The Boro
Netgear CM500
Pace 5268AC
TRENDnet TEW-829DRU

1 edit

NetFixer to JohnInSJ

Premium Member

to JohnInSJ

Re: [Speed] Comcast Business Class Speed Upgraded in California

said by JohnInSJ:

rebooted this morning in san jose
12/2 is now

Upstream: No shaper detected.
Median received rate: 3397 Kbps.

Downstream: Burst size: 9843-9962 KB; Shaping rate: 17024 Kbps.

17/3 - nice.

Not sure why it didn't see a speedboost on the uplink.

I randomly see no PowerBoost reported on both the upstream and the downstream when testing with ShaperProbe.

But I would actually like to see Comcast offer a config option to disable PowerBoost anyway. In fact, I usually run through a VLAN segment of my Netgear GS108e switch just behind my cable modem so that I can use its rate limiting QoS capability to compensate for the sometimes extreme buffer bloat effect that the combination of Comcast's PowerBoost and channel bonding sometimes exacerbate. Since the "leak" of Comcast's plan to increase the bandwidth for the two lower business class tiers, I have removed that VLAN connection until I could verify exactly what the new speed tier delivered. Now that I know, I will probably put that rate limiting VLAN segment back into operation.

EDIT:
I just put my rate limiting VLAN segment back into service, and I thought I would post some ICSI Netalyzr results to illustrate why I did it:


Netalyzr results without rate limiting


The 690 ms upstream buffer value above is actually an improvement over the ~1000 ms value I was seeing prior to the rate increase from 2mbps to 3mbps (which makes sense because the improved excess buffering ratio tracks the real bandwidth increase ratio).


Netalyzr results with rate limiting


Here is a link to an ICSI FAQ that explains the effects of excessive buffering (AKA buffer bloat): »n3.netalyzr.icsi.berkele ··· fer.html

Here is a brief extract from that FAQ:
said by ICSI :
Often, your computer wants to send or receive data faster than some point in the network allows. Thus that point needs a buffer to store this data until it is able to send it. The problem arises when the buffer is too large or too small. If the buffer is too small, network protocols such as TCP are unable to send as fast as the network allows. If the buffer is too large, a single transfer will fill up the buffer, delaying all other traffic.

Currently, a lot of users have networks which are significantly over-buffered. As a result, if the network is running at capacity, packets may be substantially delayed. This is why, for example, BitTorrent may slow down a user's web surfing, even though TCP is able to share all the bandwidth, the file transfer fills up the buffer which now slows down all other traffic.

In other words, those of you who are bemoaning an apparent loss of PowerBoost with the new speed increase, might want to be thankful instead of angry.