dslreports logo
 
    All Forums Hot Topics Gallery
spc
Search similar:


uniqs
917
vengance01
join:2009-03-06
Encino, CA

vengance01

Member

Having trouble getting the full 24 down with nothing in use

I the best I can see is 2.8 MB/s which is roughly 22-22.5mbit/s. Whats the overhead for the 24/3? This was directly connected to the router with a Macbook Pro

TestBoy
Premium Member
join:2009-10-13
Irmo, SC

TestBoy

Premium Member

I hardly ever got much more than that when I had 24/3
What's your UV Realtime stats look like?
Okay I take it?
vengance01
join:2009-03-06
Encino, CA

vengance01

Member

Stats are fine. I do get a fair amount of corrected blocks, but nothing else.
WhyMe420
Premium Member
join:2009-04-06

WhyMe420 to vengance01

Premium Member

to vengance01
That's normal with overhead.
vengance01
join:2009-03-06
Encino, CA

vengance01

Member

Wow there is that much overhead? I remember getting more then 2.8 when I used to have u-verse but maybe not..
WhyMe420
Premium Member
join:2009-04-06

WhyMe420

Premium Member

The full 24 Mbps is 3MB/Sec, so you're only missing like 200-300KB/Sec. I typically get 2.7MB/Sec (21.6Mbps) on my 24/3 line, but speedtest.net always says 23Mbps.




That's what I always get. I'm getting the 45/6 installed this Friday. I wonder how close to those numbers I'll get.

TestBoy
Premium Member
join:2009-10-13
Irmo, SC

TestBoy

Premium Member

said by WhyMe420:

That's what I always get. I'm getting the 45/6 installed this Friday. I wonder how close to those numbers I'll get.

Probably pretty close.
This is what I get:
»www.speedtest.net/result ··· 7203.png
WhyMe420
Premium Member
join:2009-04-06

WhyMe420

Premium Member

Sweet. Can't wait!
l36pontiac
join:2012-12-03
77407

l36pontiac to vengance01

Member

to vengance01
Remember the fine print from all ISP, up too.

rolande
Certifiable
MVM,
join:2002-05-24
Dallas, TX
ARRIS BGW210-700
Cisco Meraki MR42

rolande to vengance01

MVM,

to vengance01
Average packet size is usually around 350 bytes. So with a 40 byte IP header and a 5 byte PTM header you'll average around 10-12% overhead. If you test with 1500 byte packets you will see better performance around 23Mbps which is equivalent to ~2.8MB/sec. So that would match up to what you are seeing. With typical application data transfers, you won't necessarily see transfers rates consistently that high.
nephipower
join:2012-02-20
San Antonio, TX

nephipower

Member

So does that mean that AT&T finally did better with the new Power 45 mbit package to account for packet overhead?

It seems like just about everyone's speedtests i have seen for the new package have been over 45 mbits.

But then the upload isn't quiet as good, getting around 5.5 mbits or so

TestBoy
Premium Member
join:2009-10-13
Irmo, SC

TestBoy

Premium Member

said by nephipower:

So does that mean that AT&T finally did better with the new Power 45 mbit package to account for packet overhead?

It seems like just about everyone's speedtests i have seen for the new package have been over 45 mbits.

But then the upload isn't quiet as good, getting around 5.5 mbits or so

For me the upstream rate is only 6080 so 5.9 megabits.
Getting 5.7 is pretty good, IMHO.

rolande
Certifiable
MVM,
join:2002-05-24
Dallas, TX
ARRIS BGW210-700
Cisco Meraki MR42

rolande to nephipower

MVM,

to nephipower
said by nephipower:

So does that mean that AT&T finally did better with the new Power 45 mbit package to account for packet overhead?

It is possible with the much larger bandwidth capacity that they have gotten flexible with the traffic shaping policy for the Internet traffic to provide an additional 1-2Mbps gap for overhead. With 1500 byte packets, accounting for the 45 bytes per packet of overhead, you would need ~46.35Mbps to achieve actual data throughput of 45Mbps or ~5.5MB/sec. That would seem fairly generous of AT&T to give you the extra room for the IP overhead. I guess the 3% difference on someone paying the premium for 45Meg isn't going to break their backs and it plays to more honest statistics on the brochures.
vengance01
join:2009-03-06
Encino, CA

vengance01 to rolande

Member

to rolande
Thank you kind sir for the info.

TestBoy
Premium Member
join:2009-10-13
Irmo, SC

TestBoy to vengance01

Premium Member

to vengance01
It would be a good thing for AT&T to do seeing as TWC has started to do the same.
It may have had to do with reports that ISPs were not giving users what they advertised.

I have no idea what AT&T has pulled into a typical VRAD but it's probably not a small amount of bandwidth.

rolande
Certifiable
MVM,
join:2002-05-24
Dallas, TX
ARRIS BGW210-700
Cisco Meraki MR42

rolande

MVM,

said by TestBoy:

I have no idea what AT&T has pulled into a typical VRAD but it's probably not a small amount of bandwidth.

It used to be a pair of OC-3 circuits @155Mbps each to the old legacy RTs running the old ADSL service. I have not heard any details on the VRADs uplink connectivity, but I would guess it is at least a pair of OC-12's @622Mbps each.
vengance01
join:2009-03-06
Encino, CA

vengance01

Member

thats just a sick amount of bandwidth
Matt7
join:2001-01-02
Columbus, OH

Matt7 to rolande

Member

to rolande
said by rolande:

It used to be a pair of OC-3 circuits @155Mbps each to the old legacy RTs running the old ADSL service. I have not heard any details on the VRADs uplink connectivity, but I would guess it is at least a pair of OC-12's @622Mbps each.

It's crazy to think that there is OC-12s running within 1000 ft of me.. yet I am limited to 24/3

my thoughts to rolande

Anon

to rolande
If i follow correctly 2 bpon connections at 622M equals 1244M
Max of 32 connection per joined bpon equals 38.88 per connection
Att distribution is 32/5 profile total 37M per connection
This shared bandwidth is prioritized for 1) voip, 2) iptv, 3) hsia

Att bases internet in steps of 6, cable seems to based on steps of 5
If T based on cable standard then
Elite 6M is 5M, cx speedtest of 5.5 means getting 10% overage instead of 90% upto.
Max 12M is 10M, speedtest of 11.2 is 10% overage
Max Plus 18M is 15M, speedtest of 16.5 is 10% overage
Max Turbo 24M is 20M, speedtest of 22 is 10% overage.

If a simple change of how advertised cx would be saying getting speeds +10% over
Instead of consistent not getting what i am paying for and complaining about fine print up to.

But just my thoughts

weaseled386
join:2008-04-13
Edgewater, FL

weaseled386 to rolande

Member

to rolande
said by rolande:

said by TestBoy:

I have no idea what AT&T has pulled into a typical VRAD but it's probably not a small amount of bandwidth.

It used to be a pair of OC-3 circuits @155Mbps each to the old legacy RTs running the old ADSL service. I have not heard any details on the VRADs uplink connectivity, but I would guess it is at least a pair of OC-12's @622Mbps each.

The VRAD's (7330's) have never had any type of OCxx link. They've strictly used 1GB links back to the 7450's in the CO.

rolande
Certifiable
MVM,
join:2002-05-24
Dallas, TX
ARRIS BGW210-700
Cisco Meraki MR42

rolande

MVM,

said by weaseled386:

The VRAD's (7330's) have never had any type of OCxx link. They've strictly used 1GB links back to the 7450's in the CO.

That would make sense, since it would probably be cheaper to deliver and provide greater capacity. I assume it is Long Range Ethernet over dark fiber.

TestBoy
Premium Member
join:2009-10-13
Irmo, SC

TestBoy to vengance01

Premium Member

to vengance01
So we have multiple gig links to a VRAD.
Do the VRADs operate like a HFC cable network does?
I forget the term but basically if no subscribers on a node are watching a particular channel then the system does not send it.
When a customer requests something a channel is allocated and the receiver is told where to get it. When another customer requests the same one the system tells that box to tun to the previous allocation.

I guess my question is... does a VRAD get sent all of the streams and from there they are allocated or do the VRADs requests streams be sent to them from the head end when a customer wants it?

Not sure how to articulate this question.... so forgive me if it's botched

rolande
Certifiable
MVM,
join:2002-05-24
Dallas, TX
ARRIS BGW210-700
Cisco Meraki MR42

rolande to my thoughts

MVM,

to my thoughts
said by my thoughts :

If i follow correctly 2 bpon connections at 622M equals 1244M

It is definitely not BPON or GPON or even SONET. As weaseled386 See Profile corrected me below they use Gig ethernet over fiber links from the 7330's back to the 7450's in the CO. That is a Gig of bandwidth to share across all VRAD subscribers with likely a live and protect uplink. It would be really interesting to know what the subscriber/bandwidth ratios are and what kind of utilization is typically seen on an average VRAD.

I believe the 7330's are sized to handle 192 physical VDSL connections. With pair bonding that would equate to 96 total customers. On the 32M profile they hit oversubscription after the 31st customer. Even with the TV service using multicast, if there were enough customers to stream all channels simultaneously, they would need over a Gig of bandwidth to do so, let alone deliver Internet bandwidth. I'm sure that AT&T has modeled it extensively and run the numbers over and over and they can predict exactly when they would need to add more bandwidth to a VRAD. It would be cool to see what those models look like.

StillLearn
Premium Member
join:2002-03-21
Streamwood, IL

StillLearn to vengance01

Premium Member

to vengance01
said by vengance01:

I the best I can see is 2.8 MB/s which is roughly 22-22.5mbit/s. Whats the overhead for the 24/3? This was directly connected to the router with a Macbook Pro

Try »speedtest.comcast.net/

rolande
Certifiable
MVM,
join:2002-05-24
Dallas, TX
ARRIS BGW210-700
Cisco Meraki MR42

rolande to TestBoy

MVM,

to TestBoy
said by TestBoy:

Do the VRADs operate like a HFC cable network does?
I forget the term but basically if no subscribers on a node are watching a particular channel then the system does not send it.
When a customer requests something a channel is allocated and the receiver is told where to get it. When another customer requests the same one the system tells that box to tun to the previous allocation.

I guess my question is... does a VRAD get sent all of the streams and from there they are allocated or do the VRADs requests streams be sent to them from the head end when a customer wants it?

Yes. The Uverse TV service uses IP multicast for its video feeds. It is an IP based service but acts much like switched video. The video streams are only sent when a customer requests it. If more than one customer on a VRAD wants to watch the same channel, they are both served by the same video stream from the same multicast group.

The caveat is the guide thumbnails and when you initially switch channels the video stream uses TCP which is unicast. But after ~10 seconds or so the STB switches from the unicast feed to the multicast transparently.

weaseled386
join:2008-04-13
Edgewater, FL

1 recommendation

weaseled386

Member

To be honest, I don't know the specifics of it. I install the hardware, and ATT comes days/weeks behind me. They handle the fiber cross connects, splice the field side of the lightning protectors, and turn it up. Not all 7330's are installed equally. Some are turned up for video, and some are internet only. You are correct about port capacity... It'll handle 192 subscribers based on no one being bonded.

Typically a VRAD will have a dedicated 1GB link, but internet only systems can have an expansion shelf. IE: two 7330's (384 ports) sharing the 1GB link. All of the techs I've talked to say 10GB links are possible. I assume that would require additional 7450's in the CO's.

Before anyone gets their knickers bunched about the over subscription, it is nothing new! I don't know of a single residential service that isn't. Even POTS (SLC96 and SLC5) can be knocked down to 2 DS1's per 96 (DS0) customers. Business T1/T3's are a different story. They're 1:1 on SONET. The current trend is pulling away from SONET, and going to IP circuits on Ciena 5140's.

rolande
Certifiable
MVM,
join:2002-05-24
Dallas, TX
ARRIS BGW210-700
Cisco Meraki MR42

rolande

MVM,

said by weaseled386:

Before anyone gets their knickers bunched about the over subscription, it is nothing new! I don't know of a single residential service that isn't. Even POTS (SLC96 and SLC5) can be knocked down to 2 DS1's per 96 (DS0) customers.

No knicker bunching going on here. I know full well the task of oversubscription management. When you can accurately predict the uplink utilization based on customer density and service distribution you can save a huge amount of money on infrastructure at the aggregation layer. It all comes back to managing cost. The question is what the oversubscription model looks like and in reality what the available headroom is at peak utilization. If they are able to average 60% utilization or less on the uplink while being oversubscribed 5:1 or less, they are probably in pretty darn good shape.

It still has to be more difficult to predict utilization with Uverse DVRs constantly recording shows at any time of the day or night. I assume that AT&T has visibility into that scheduling info anyway, but it would still be a massive exercise in constant data management and measuring.

weaseled386
join:2008-04-13
Edgewater, FL

weaseled386

Member

said by rolande:

said by weaseled386:

Before anyone gets their knickers bunched about the over subscription, it is nothing new! I don't know of a single residential service that isn't. Even POTS (SLC96 and SLC5) can be knocked down to 2 DS1's per 96 (DS0) customers.

No knicker bunching going on here. I know full well the task of oversubscription management. When you can accurately predict the uplink utilization based on customer density and service distribution you can save a huge amount of money on infrastructure at the aggregation layer. It all comes back to managing cost. The question is what the oversubscription model looks like and in reality what the available headroom is at peak utilization. If they are able to average 60% utilization or less on the uplink while being oversubscribed 5:1 or less, they are probably in pretty darn good shape.

It still has to be more difficult to predict utilization with Uverse DVRs constantly recording shows at any time of the day or night. I assume that AT&T has visibility into that scheduling info anyway, but it would still be a massive exercise in constant data management and measuring.

Yeah, with a CCIE tag I knew you'd have a grasp on it! That was aimed more at the non-techie readers.

The first gen of DSLAM's (Alcatel 1000's) were provisioned with 4 DS1's per 48 customers. When 3 shelves were installed we'd link them, and 144 people would share a DS3.

The second round used Alcatel HD7300's. Each shelf served 216 customers, and would have a DS3. It wasn't uncommon for them to link two shelves together -- 45M shared between 432 customers.

The 3rd round was Alcatel 7300 LP-UD's. Once again DS3 fed for 216 people. Very common for 2 systems to share the link.

As for the 7330's, once a system hits 50% the RT is upgraded (if possible). I don't think I've ever been to a site that all 192 ports were utilized.

For FTTH we've been installing Ericson BLM1500's. There is some issue with them, and they've been manufacture discontinued. For all jobs moving forward we're tossing in Alcatel 7342's. I haven't had the chance to see one yet, but my inner nerd can't wait
Paralel
join:2011-03-24
Michigan, US

Paralel to vengance01

Member

to vengance01
This certainly helps shed light on the "upgrade" nonsense I'm having to deal with at the moment with regard to getting the 45/6 Power tier where I'm located.

rolande
Certifiable
MVM,
join:2002-05-24
Dallas, TX
ARRIS BGW210-700
Cisco Meraki MR42

rolande to weaseled386

MVM,

to weaseled386
Yeah, definitely not my first rodeo, so to speak.

So, the assumption is that the first "3 rounds" were all the legacy ADSL service running in an RT or the CO and no video services. I could swear I was informed in the past that each RT was fed with a pair of OC-3's. Either I misheard or someone was guessing as fact.

So it is a single uplink connection per VRAD/7330? I guess being the phone company has its advantages when it comes to the circuit repair response time. This certainly sheds light on the availability expectations. After all, it is an unregulated consumer service.

Just curious, what is the MTTR if the VRAD 7330 hardware takes a direct lightning hit or some such catastrophe that renders it a smoking pile of metal and silicon? What about uplink circuit failure/agg router (7340) failure? I assume there is at best a 4-8 hour response window plus time to repair/obtain spare hardware and re-image the configs?