dslreports logo
site
 
    All Forums Hot Topics Gallery
spc

spacer




how-to block ads


Search Topic:
uniqs
2811
share rss forum feed

FirebirdTN

join:2012-12-13
Brighton, TN
kudos:1
Reviews:
·Comcast

Bufferbloat-long read.

I am starting this thread because I have had a couple discussions both recently in one thread (wrecking the OPs thread) and with other users back when I was suffering intermittent connectivity.

Starting a few months back (December) my internet would just "cut off" for brief periods of time. It took 3 months, several tech visits, a lot of lost hair, and finally stumbling upon a phenomenon I had never ever HEARD of to figure out what was going on.

Once I discovered what was happening, I consulted the man who coined the term himself (Jim Gettys), and he confirmed my suspicions.. Problem has been solved ever since.

I initially had a DOCSIS 3.0 emta (4x4). Basically what was happening was whenever my upload path was saturated, it rendered the internet connection unusable for anyone else on the network. Other users would experience "Page can't be displayed", "unable to connect to email server", gaming disconnects, etc. Cable, equipment, signal levels, etc had all been ruled out as a source of the cause. That left the modem itself. I replaced it with a DOCSIS 2 loaner modem, and the "outages" went away completely. Feeling confident that I found the problem, thinking it was a bad modem, I swapped it out for another DOCSIS 3 emta (8x4), and the issue instantly came back, only WORSE.

It was at this point that I just stumbled onto bufferbloat, and realized the symptoms matched what I was experiencing completely. Studying up on it, I realized it could be mitigated by enabling bandwidth shaping on my router to keep from utilizing the modem's built in buffer by limiting my bandwidth to *just slightly* under my provisioned speeds. Once I did that, all my issues disappeared, and my performance even improved! DNS lookups were instant, no matter if the connection was idle, or I had 3 other people on the internet at the same time.

In discussing the phenomenon with some users via PM I was told that not everyone believes it is real. That is understandable, as it will ONLY manifest itself when a link is saturated (and usually only noticeable to the end user if the UPLOAD path is saturated). If you are the sole user, or have very little upload traffic on your internet connection you may never seen an issue. Again, its only when your link is saturated that it becomes an issue.

So, what does this have to do with anything, really you might wonder...Along with excessively large buffers existing in OSes, routers, Wifi connections, and 3g, I have found that cable modems are really bad about it as well. And to make matters worse, the *faster* the modem is capable of, the worse the issue gets as the modem is operated at relatively slower speeds. The sad thing about this is the ISP's have the power to do something about it by controlling the active buffer size in our modems, but to the best of my knowledge, none of them do.

Most everyone should know what a buffer is and what it does, so no point explaining that. What is important to know is when you are sending data at a rate SLOWER than what your internet connection can accept it, the buffer isn't used at all. That data comes in and goes right out. Now, lets say you have a 1gbps connection to your modem. But you only have a 10mbps upload speed. When you try to upload a large file, its pretty obvious that you are going to try and send faster than your connection can accept the data; what happens is the modems buffer starts to fill. Once the modem's buffer is full, THEN the computer realizes it sending data way to fast and slows down the transfer until it finds an equilibrium. The point which the equilibrium is found will result in a buffer that is always full (until the transfer is finished).

Now here is the real problem. Due to falling memory prices (and the non availability of small enough memory chips) buffer sizes have been increasing in broadband equipment. The problem is consideration isn't given to the proper sizing of buffering in the equipment. What they should be doing doing is sizing the buffer based on a delay factor, not a size factor.

Here is an extreme example, but it should make the issue readily apparent:

In our example above with our 1gbps connection to the modem, but a 10mbps upload speed, lets say we are uploading a 1GB file. Lets also say we have a 100mb buffer. When we start the transfer our PC will essentially send data at a 1gbps rate, effectively filling that buffer in a short time. Once our buffer is full and the transfer is underway, lets pretend PC #2 makes a DNS lookup to view a website. The DNS lookup gets thrown in at the end of the buffer, and has to wait until it transverses the buffer before it gets out to the internet. In our example, PC#2 has to wait 10 seconds before it can even LOOK UP the address of the website we want, and we haven't even started loading the page yet. Now if that buffer was only 10mb, the delay would be 1 second for the lookup. This is only an issue when the buffer is full (link saturated). When the buffer is empty, the buffer could be 1000mb in size and it wouldn't matter, there would be no added delay to our DNS lookup. Stressing this point again-this is ONLY an issue when the link is saturated.

...To Be Continued...

-Alan

FirebirdTN

join:2012-12-13
Brighton, TN
kudos:1
Reviews:
·Comcast

3 edits
Click for full size
Click for full size
Click for full size
The reason for this post was most recently a user asked whether they should upgrade their modem from a 4x4 to a 4x8 model.

At first the choice seems fairly obvious...the 8 channel isn't much more money and will actually help performance in an overly congested node.

However, on the flip side of the coin there is that bufferbloat problem. Without full technical specifications it is hard to make a comparison, but generally, the 4x8 modem buffer is going to be sized at LEAST appropriately when operated at its rated speed (I don't know what a 4x8 tops out at but its 300+mbps).

That is fine, but if you operate that modem at a substantially slower speed, all of a sudden the appropriately sized buffer is now way to large. Packet delay is a function of transfer speed and buffer size. If the buffer size stays the same, but the speed is decreased, it will take longer for the packets to get through the buffer, increasing latency.

At the top of this reply are 3 screen shots I took, all under the same conditions.

In all cases my provisioned speed was 16/4.

The first screen shot is on a D2 device. Notice the latency increases under saturation.

Second screen shot is on a D3 4x4 device. Latency gets fairly high as well, but not only does the connection lag, but it experiences total network failure when under saturation.

Last screen shot is on a D3 4x8 device. Latency gets ridiculous (3 full seconds), and causes total network failure for even LONGER periods of time.

Once traffic shaping was implemented on my router to throttle the connection to just under my provisioned speeds, all that went away COMPLETELY. By implementing shaping it effectively turned off the modem's buffer.

So back on the "should I get a faster modem" question. It depends. If you have a large network and/or a large amount of network traffic, frequently saturating your link AND you have no way to throttle your network throughput to your modem, then NO it would not be a good idea to get that faster modem just for the sake of getting more bonded channels. If you CAN implement traffic shaping on your network (effectively disabling the modem's buffer), then YES, getting a faster modem might help performance in heavily congested nodes.

For more information I highly recommend visiting this page, and the website in general:

»www.bufferbloat.net/projects/blo···calIntro

Or you can watch any of Jim Getty's tech videos I believe available on Google TechTV and/or youtube.

-Alan


graysonf
Premium,MVM
join:1999-07-16
Fort Lauderdale, FL
kudos:2
reply to FirebirdTN
More reading material:»en.wikipedia.org/wiki/Bufferbloat

I've been using traffic shaping for years to avoid this problem and the routing solution I have been using it with clearly explains that you can't solve the problem without surrendering a small amount of bandwidth, ie properly setting the rate limits to be just below your capabilities.


NetDog
Premium,VIP
join:2002-03-04
Parker, CO
kudos:79
reply to FirebirdTN
Bufferbloat Demo at IETF86

»www.youtube.com/watch?v=NuHYOu4aAqg


You can also check out homenet working group at IETF.. »tools.ietf.org/wg/homenet/
--
Comcaster.. Network Engineer with NETO

FirebirdTN

join:2012-12-13
Brighton, TN
kudos:1
Reviews:
·Comcast
Just an update to add some clarification/clear up some misconceptions:

Bufferbloat is present in any number of devices: OSes, routers, 3g, WiFi, but the biggest issue with Bufferbloat will always be with the device at the bottleneck, which will usually be your broadband device that connects to your ISP (ie: Cable modem, DSL Modem, etc)

This guys example says it way better that I ever could:

»betanews.com/2012/10/04/its-time···erbloat/

-Alan


JohnInSJ
Premium
join:2003-09-22
Aptos, CA
reply to FirebirdTN
Interesting term. In TCP/IP networking, not giving your ACK packets any way to get out because your uplink is saturated will kill your download speed.

Because the sender (the computer on the other end, that you are downloading FROM) won't send you any more packets until you acknowledge (by sending them an ACK packet) the packets you've gotten already.

This is why when you flood your uplink your downlink suffers/stalls. Traffic shaping (giving ACK packets about 10% of your uplink via any number of techniques) will fix this, but it has to be done at your egress router to be 100% effective.

I guess someone needed a less technical way to describe how tcp/ip networking works.
--
My place : »www.schettino.us


workablob

join:2004-06-09
Houston, TX
kudos:3
Reviews:
·Comcast
reply to NetDog
said by NetDog:

Bufferbloat Demo at IETF86

(youtube clip)
can also check out homenet working group at IETF.. »tools.ietf.org/wg/homenet/

Very nice video.

Thanks

Dave
--
I may have been born yesterday. But it wasn't at night.

nmbdude

join:2010-06-09
North Miami Beach, FL
reply to JohnInSJ
While this is a very real problem, it's also nothing new. This existed in the days of Docsis 1.1 modems, too.

I personally use a mini-pc router running FreeBSD and PF (QoS) to resolve the issue (and it works really well). I also have worked as a network engineer for an enterprise/business internet services provider. Obviously most non-technical users will not be able to set something like that up.

I've found in discussions with most other cable users that until I pointed the problem out to them, they never even noticed it was happening -- even if they were technical users, unless they happened to be major uploaders/torrenters.

The buffer sizes are configurable (by the Cable ISP), and they choose to have large buffers, either out of ignorance or out of ulterior motives.

Cable ISP's have two incentives to configure things this way:
1) You are much less likely to experience saturation when you have a higher speed service -- so this is a good incentive to upgrade from their 20Mbps tier to 50Mbps (or 100Mbps) (which also have higher upload capabilities)
2) Large buffers optimizes the link for throughput (maximum speed), smaller buffers optimize the link for latency (interactive sessions like VoIP, video conferencing, or administrative tasks like Remote Desktop or SSH. As well as for shared links with multiple users simultaneously surfing). When they optimize for throughput, it helps put up a nice number when you run a speedtest, which is great for marketing.

As far as incentive number 1 goes -- My service was recently upgraded from 50/5 to 105/20, and while I still use PF, there's plenty enough capacity that I would barely ever cause noticeable buffering without QoS.

For less-tech-savvy users, setting up DD-WRT really shouldn't be too difficult if you are experiencing this problem and want to mitigate it, and while it may not have as many options to optimize things as PF on FreeBSD, it will be a huge improvement.

Waiting for Cable Co's to "fix" this would be akin to waiting for the drug companies to make people healthier.


NathanO

join:2008-08-21
Moorestown, NJ
CableLabs is working pretty hard to test CoDel on DOCSIS: »www.cablelabs.com/downloads/pubs···work.pdf - Based on the video I assume Comcast is working on it as well.


JohnInSJ
Premium
join:2003-09-22
Aptos, CA
reply to nmbdude
said by nmbdude:

The buffer sizes are configurable (by the Cable ISP), and they choose to have large buffers, either out of ignorance or out of ulterior motives.

They're not unique. Same issues over ADSL modems (I recall when we had this problem on AT&T/PacBell, and then later Sonic.net, over RT connections.)

Large buffers provide performance advantages to 99.999% of all home broadband users. The ones saturating their uplinks should know how to fix it.
--
My place : »www.schettino.us


graysonf
Premium,MVM
join:1999-07-16
Fort Lauderdale, FL
kudos:2
said by JohnInSJ:

The ones saturating their uplinks should know how to fix it.

Exactly.

nmbdude

join:2010-06-09
North Miami Beach, FL
reply to JohnInSJ
said by JohnInSJ:

They're not unique. Same issues over ADSL modems (I recall when we had this problem on AT&T/PacBell, and then later Sonic.net, over RT connections.)

Yes, it happens on pretty much any consumer-grade internet service -- DSL, Satellite, 3G, and Cable.

said by JohnInSJ:

Large buffers provide performance advantages to 99.999% of all home broadband users. The ones saturating their uplinks should know how to fix it.

Not really. Buffering past 200ms is probably harmful in essentially all cases. TCP Adjusts it's sending rate based on packet loss, buffering prevents that loss and convinces the sending side to continue increasing the speed until eventually it exhausts the buffer, after latency has become so high as to make all other interactive sessions unusable.

In fact most business-grade routers (Gig-E/10GE) have buffers smaller than 10ms, and are fully capable of maxing out at 1000 or 10,000 Mbps. There is simply no reason to have 1000ms of buffering in any standard broadband link.

But yeah, I agree that if you are smart enough to notice the problem, you should be able to fix it -- and probably in a better way than the generic untuned one-size-fits-all queue management that your ISP will pick.

FirebirdTN

join:2012-12-13
Brighton, TN
kudos:1
Reviews:
·Comcast

1 edit
reply to nmbdude
said by nmbdude:

While this is a very real problem, it's also nothing new...

I appreciate everyone's input. I do apologize for being long winded above. The main reason I started this thread is basically I was told by a user in another thread that bufferbloat has nothing to do with cable modems, and I wanted to try and set the record straight without wrecking the OP's thread (I think it was too late though).

As far as it being nothing new...maybe...but I had never experienced it, or even *heard* of it prior to March. It wasn't until I pulled my hair out for three months trying to figure out what was wrong with my internet when I stumbled onto it.

One thing that still hasn't been explained though...not that it is important, but I am damn curious:

I don't normally generate enough upload traffic to even come across this issue; The ONLY reason this issue surfaced for me is my wife had an outgoing email with LARGE attachment stuck in her iPAD since the day after Christmas. (I didn't know that much about apple stuff, and its her's so I never looked at it in all that time). Anyway, evidently when it was trying to send that email out it would cause my internet to lag. I get that now. I understand completely why that was happening. What I don't get is why did it not only lag, but caused total internet failure but ONLY when I used DOCSIS 3.0 modems?

Probably never will get an answer to that one, but if anyone has a guess I'd love to know. Even Mr. Getty's couldn't explain why that was happening.

-Alan


Streetlight

join:2005-11-07
Colorado Springs, CO
reply to FirebirdTN
You can also get information on buffer bloat from Steve Gibson's video podcast on the Twit network episode 345

»twit.tv/show/security-now?page=8

Actually, it looks like it's on p 9, but the link works.


netcool
Premium,VIP
join:2008-11-05
Englewood, CO
kudos:107
reply to FirebirdTN
This was published last month regarding the different active queue management algorithms that are being considered by CableLabs:

»www.cablelabs.com/downloads/pubs···_3_0.pdf

An interesting read with a lot of good info.


NetFixer
Freedom is NOT Free
Premium
join:2004-06-24
The Boro
Reviews:
·Cingular Wireless
·Comcast Business..
·Vonage

4 edits
reply to FirebirdTN
said by FirebirdTN:

One thing that still hasn't been explained though...not that it is important, but I am damn curious:

I don't normally generate enough upload traffic to even come across this issue; The ONLY reason this issue surfaced for me is my wife had an outgoing email with LARGE attachment stuck in her iPAD since the day after Christmas. (I didn't know that much about apple stuff, and its her's so I never looked at it in all that time). Anyway, evidently when it was trying to send that email out it would cause my internet to lag. I get that now. I understand completely why that was happening. What I don't get is why did it not only lag, but caused total internet failure but ONLY when I used DOCSIS 3.0 modems?

Probably never will get an answer to that one, but if anyone has a guess I'd love to know. Even Mr. Getty's couldn't explain why that was happening.

-Alan

I see a big difference in upstream buffer size if I switch between my normal SB6121 DOCSIS 3 modem with three bonded upstream channels, and my backup DCM202 DOCSIS 2 modem (with only a single upstream channel). When rate limiting on a local router or switch is disabled*, the Netalyzer test consistently shows ~700ms upstream buffering using the SB6121 modem (when my upstream speed was 2mbps instead of the current 3mbps, I would see ~1100 ms upstream buffering). When I use the DCM202 modem the upstream buffering is ~90 ms. Anecdotally, it would seem that more upstream channels may mean more upstream buffering.

OTOH, Comcast's PowerBoost also seems to be part of the equation. With the SB6121 and its three upstream channels, I frequently see upstream PowerBoost results of ~20mbps on my 16/3mbps account; but when I use the DCM202, my upstream speed is a fairly constant 3mbps (no PowerBoost effect at all). If it weren't for lack of native IPv6 support, I would probably just use the DOCSIS 2 DCM202 modem due to its more predictable performance. I keep seeing hints that Comcast is thinking of retiring PowerBoost; I would love to see that on my account (if for no other reason than that I might also find out if the excessive upstream buffering is perhaps totally caused by PowerBoost).

* FWIW, the primary deleterious effect I would see from excessive buffering was with VoIP traffic. I used to counteract that using rate limiting in a Netgear GS108E switch on the common WAN side of my routers, but the primary NAT/IPv6 router that I currently use did not work well with rate limiting on its WAN interface; instead of smooth rate limiting, I would see dropped packets (and its own internal QoS/rate limiting had its own set of problems aside from being outside the path used by my servers and VoIP service). My solution was to replace the Netgear GS108E with a ZyXEL GS105S switch that uses priority based QoS instead of rate limiting. With that solution, the large ~700 ms upstream buffering is still present, but it no longer causes problems for VoIP traffic.
--
A well-regulated militia, being necessary to the security of a free State, the right of the people to keep and bear arms shall not be infringed.

When governments fear people, there is liberty. When the people fear the government, there is tyranny.


telcodad
Premium
join:2011-09-16
Lincroft, NJ
kudos:15
reply to FirebirdTN
This week's "Translation Please" column by Leslie Ellis just happens to discuss "Buffer Bloat":

May Your Buffer Never Bloat
By Leslie Ellis, Multichannel News - May 13, 2013
»www.multichannel.com/blogs/trans···er-bloat


DavidA

@torservers.net
A related problem with Comcast is that when their peers (Tata, Cogent, GX, etc) fill up, the latency jumps up by 50+ ms, but packets do not actually get dropped... contributing to buffer bloat.

I know that Comcast uses the Cisco CRS-1 and ASR9k routers on their provider edge... could some of the Comcast engineers on this forum see about maybe replacing some of the queuing with drops to reduce the effects of buffer bloat?


NetFixer
Freedom is NOT Free
Premium
join:2004-06-24
The Boro
Reviews:
·Cingular Wireless
·Comcast Business..
·Vonage

1 recommendation

said by DavidA :

A related problem with Comcast is that when their peers (Tata, Cogent, GX, etc) fill up, the latency jumps up by 50+ ms, but packets do not actually get dropped... contributing to buffer bloat.

I know that Comcast uses the Cisco CRS-1 and ASR9k routers on their provider edge... could some of the Comcast engineers on this forum see about maybe replacing some of the queuing with drops to reduce the effects of buffer bloat?

Of course then there will be the inevitable complaints about dropped packets being unacceptable and Comcast better do something about it...or else.
--
A well-regulated militia, being necessary to the security of a free State, the right of the people to keep and bear arms shall not be infringed.

When governments fear people, there is liberty. When the people fear the government, there is tyranny.


NetAbuser

@torservers.net
Sometimes you win, and sometimes you lose. You can't please everyone all the time, and the ultimate fix is for Comcast to upgrade their peering capacity across the board, but that just isn't happening with their current administration and mindset.

Given what we've got, having Comcast turn down some of their buffering to remove bufferbloat is the best thing we can hope for.


NetFixer
Freedom is NOT Free
Premium
join:2004-06-24
The Boro
Reviews:
·Cingular Wireless
·Comcast Business..
·Vonage
said by NetAbuser :

Sometimes you win, and sometimes you lose. You can't please everyone all the time, and the ultimate fix is for Comcast to upgrade their peering capacity across the board...

I suspect that the "ultimate fix" for Comcast will be to assimilate the rest of the entertainment industry, with the rest of the Internet following in the wake.
--
A well-regulated militia, being necessary to the security of a free State, the right of the people to keep and bear arms shall not be infringed.

When governments fear people, there is liberty. When the people fear the government, there is tyranny.


PacketLost

@torservers.net
Or Comcast's end game of getting all of their network "abusers" criminally charged.


NetFixer
Freedom is NOT Free
Premium
join:2004-06-24
The Boro
Reviews:
·Cingular Wireless
·Comcast Business..
·Vonage
said by PacketLost :

Or Comcast's end game of getting all of their network "abusers" criminally charged.

Not really. Comcast is a for profit company and getting paid is far more important than retribution.

I received a snail mail notification the first of this month that my DTA had accidentally been programmed to provide more channels than my current TV plan was supposed to receive. There was no threat of prosecution for "theft of service", or service termination (because that would have been ridiculous since Comcast programmed the DTA). There was only a sales pitch that they would be happy to change my TV plan to match the current DTA config (with of course an increase in the monthly bill). I plan to do nothing and allow the DTA programming to change at the end of the current billing period (as the notice said will happen) because I had no idea what channels I was supposed to get in the first place, and I can't remember the last time I watched anything on TV except for occasionally the local news and weather. I suppose that if my wife starts complaining next month that some channel that she watches has disappeared, then I will deal with that if/when it happens.
--
A well-regulated militia, being necessary to the security of a free State, the right of the people to keep and bear arms shall not be infringed.

When governments fear people, there is liberty. When the people fear the government, there is tyranny.


DarkLogix
Texan and Proud
Premium
join:2008-10-23
Baytown, TX
kudos:3
reply to FirebirdTN
And this is why QOS should be used on asymmetric links and at bottle necks.

If you have a node that has one interface that is vastly faster then its other interface then some form or prioritization is needed to properly manage the flow.

Imagine if you will a Proxy server with a dialup modem and a gig eth link.
--
»www.change.org/petitions/create-···imcity-4


DarkLogix
Texan and Proud
Premium
join:2008-10-23
Baytown, TX
kudos:3
reply to FirebirdTN
said by FirebirdTN:

The ONLY reason this issue surfaced for me is my wife had an outgoing email with LARGE attachment stuck in her iPAD since the day after Christmas. (I didn't know that much about apple stuff, and its her's so I never looked at it in all that time). Anyway, evidently when it was trying to send that email out it would cause my internet to lag. I get that now. I understand completely why that was happening. What I don't get is why did it not only lag, but caused total internet failure but ONLY when I used DOCSIS 3.0 modems?

The real question is how big was that e-mail and would her e-mail provider even accept it.

As to why D2 was unaffected that's easy most D2 modems only have a 10/100 where as most D3 have a gig.

now what level of service are you paying for?
--
»www.change.org/petitions/create-···imcity-4


DarkLogix
Texan and Proud
Premium
join:2008-10-23
Baytown, TX
kudos:3
Personally I've never run into this issue then again I don't max my upload, the one app I use that could I have set just below my upload limit.

Then again I have 2 points of possibly node congestion

I have effectively 2gb/s to my router and then 1gb to my smc and then 25/5 (or whatever it is now) to comcast.
--
»www.change.org/petitions/create-···imcity-4

FirebirdTN

join:2012-12-13
Brighton, TN
kudos:1
Reviews:
·Comcast
reply to DarkLogix

said by DarkLogix See Profile
The real question is how big was that e-mail and would her e-mail provider even accept it.

As to why D2 was unaffected that's easy most D2 modems only have a 10/100 where as most D3 have a gig.

now what level of service are you paying for?

I don't know the size of the email, but it was way larger than the provider would accept. That is why it stayed in her outbox for three months driving me crazy.

BUT...It shouldn't matter if the attachment was 100 Gigs in size-It should NOT have caused total internet failure for ALL devices on my network [note, all devices COULD still access the modem during that time, they just could not get out of the net].

At the time my service was 16/2 I believe. Long story, but my service got knocked down in speed during troubleshooting, swapping modems, etc.

My current service is 50/10.

-EDIT- I don't see how the link speed would affect anything since even the relatively slow 10/100 ethernet link on the D2 modem was far faster than the 2mb upload speed I was provisioned at...

-Alan


DarkLogix
Texan and Proud
Premium
join:2008-10-23
Baytown, TX
kudos:3
Think about how far out of balance the 2 side of the link are

at gig it'll fill the buffer in no time

You think businesses have e-mail size limits for no reason?
we've have our wan freeze due to someone sending a 100mb e-mail (ya we now have fixed that missing size limit)

Sending something that big without QOS in place will jam the link
--
»www.change.org/petitions/create-···imcity-4


ropeguru
Premium
join:2001-01-25
Mechanicsville, VA
said by DarkLogix:

Think about how far out of balance the 2 side of the link are

at gig it'll fill the buffer in no time

You think businesses have e-mail size limits for no reason?
we've have our wan freeze due to someone sending a 100mb e-mail (ya we now have fixed that missing size limit)

Sending something that big without QOS in place will jam the link

Or worse, you are in a remote office using outlook to connect to the corporate exchange server and someone at corporate sends a 20MB email that all 70+ users in the remote office start retrieving at the same time. It basically shuts down ALL work and communication to corporate AND the outside world.

Not to mention what it does to the AD infrastructure and authentication during that time period.


DarkLogix
Texan and Proud
Premium
join:2008-10-23
Baytown, TX
kudos:3
said by ropeguru:

said by DarkLogix:

Think about how far out of balance the 2 side of the link are

at gig it'll fill the buffer in no time

You think businesses have e-mail size limits for no reason?
we've have our wan freeze due to someone sending a 100mb e-mail (ya we now have fixed that missing size limit)

Sending something that big without QOS in place will jam the link

Or worse, you are in a remote office using outlook to connect to the corporate exchange server and someone at corporate sends a 20MB email that all 70+ users in the remote office start retrieving at the same time. It basically shuts down ALL work and communication to corporate AND the outside world.

Not to mention what it does to the AD infrastructure and authentication during that time period.

I don't have to imagine, we've had people at corp send 100MB+ attachments to large distro groups (once to the everyone group)

we had set a max size limit of 30mb but corp removed it, and now we just have a send and receive connector limit at our local exch server.
--
»www.change.org/petitions/create-···imcity-4