General : Traffic Management Changes - April 2nd

[Chrysalis]

yeah simply downloading at high speeds will use more upload traffic than gaming.

even on 30mbit when I download at full pelt I am using close to a mbit/sec of upstream traffic on acks.

So 100mbit users will be using close to 3meg/sec of upstream traffic just downloading.

Just done a test now, congestion stopping me getting full speed but at 3.5mbyte/sec down I am using 122kbyte/sec up so thats about 29mbit down and 1mbit up in acks.

[qasdfdsaq]

ACKs should only be using about 1.5% of your downstream speed, so less than 1mbps when downloading at 50mbps and less than 2mbs when downloading at 100. Certainly nowhere near the 3% you claim. With SACKs and ACK suppression (which is used on the VM network) actual bandwidth used over DOCSIS will be lower still.

[Chrysalis]

what are VM doing messing with ack's?

[General Maximus]

http://www.cedmagazine.com/articles/...t-optimization

I have sort of got the gist of what is happening in principle but the thing which I dont understand is that the acks should be going to the other end of the connection. The remote host or server will be waiting for certain acks to come back so I dont understand how the cmts can block acks 1 and 2 but allow 3 to go through. It obviously works but I just dont understand how. In principle it is a good idea but from a throughput point of view wouldnt it just make more sense to increase the windows sizes and allow more data to be sent before being acknowledged?

[Chrysalis]

I can only guess that the CMTS still sends normal acks to the other end but its surpressing them on the modem side? According to dumeter I am still sending full acks out (and I do have delayed acks disabled which explains the double ack usage). Although if thie surpression is enabled (we just assuming it is), then I guess the work would be done between the modem and the CMTS, hence dumeter on my pc doesnt see it.

So an average of maybe 1.6mbit/sec used on the 100mbit service whilst downloading then not 3mbit/sec. Which equates to approx 17% of US capacity needed for acks to fully utilise the 4 DS channels. If there is 8 DS channels with 1 US channel then that increases to 34%.

[General Maximus]

if that is the case then it is a good idea on how to go about reducing traffic on the local segements which are the most congested areas of the network.

[Lee]

Did anyone ever confirm whether STM is active between 3-4pm? The charts say it isn't but I know I have been STM'd between those times (I tested it) and also there are posts on the VM Help Forums with other people reporting that too.

My guess is VM just messed up the chart as I cant see any point in there being 1 hr of no STM...

[BenMcr]

It's not active on downstream between 3pm - 4pm but the upstream management starts at 3pm

[Lee]

Originally Posted by BenMcr

It's not active on downstream between 3pm - 4pm but the upstream management starts at 3pm

Thanks Ben, but I'm pretty sure have been STM'd after around 20min of downloading at full speed after 3pm. Others reported this on the VM forums too.

I'll do a test today and check again.

[Chrysalis]

Ben can you get info on the rollout of this stm?

Mark on the VM forums posted the areas for the first weekend and no news since then yet many people seem to be reporting changes.

[BenMcr]

I've not any details myself but I'll see if I can find out (or whether Mark is going to publish anything further)

[data0002]

is the stm time done by the Virgin network time or by BST, as my superhub is still showing GMT time

[Lee]

OK test completed and I didn't get STM'd, so I guess I was imagining it last week

[Sephiroth]

Originally Posted by data0002

is the stm time done by the Virgin network time or by BST, as my superhub is still showing GMT time

STM hours are current UK time.

The modem is time synced to UTC.

[Ken Jude]

I suspect the STM changes may now have been applied in Hatfield. I think I may have just been STM'd (if I haven't the speeds are crap this evening!) but it looks like I've been cut by 50% rather than 75%.