cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
1399
Views
0
Helpful
10
Replies

bandwidth usage

catalystexpress
Level 1
Level 1

Hi All,

Can you please help me to understand on how i should be dealing with below issue

We have a 6Mbps line from verizon, we have around 150 users in our network and this link is mainly used only for browsing with video, games blocked...

I did a simple test by downloading two big size files at the same time, please note there was very few users accessing the internet as it was after office hours

362MB -- took 25min to download -- so the throughput is 2024 kbps 

440MB -- took 47min to download -- throughput is 1308 kbps

I understand there would be many TCP overheads, but with very less users accessing the circuit am i right to say i should have got atleast a throughput of  2500 kbps or above...

the vendor says they see 4 Mbps of traffic during production, in this case yes i understand we would get anywhere between 1.3-1.5 mbps

but during non production hour shouldnt we expect much better results...

how do i calculate if 6 Mbps is the correct bandwidth for 150 user network... assuming all 150 would go online same time and would only be allowed to browse but no video,games...facebook and other social network sites....

thanks a lot for all the help...

2 Accepted Solutions

Accepted Solutions

Disclaimer

The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.

Liability Disclaimer

In no event shall Author be liable for any damages whatsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.

Posting

Yes, you should see better performance if there's less network load.  Actually, it's often better to test before production hours.  With different time zones and users at home, after business hours likely decreases your link's network consumption, but remember much other Internet resources are still shared; often need to see their load drop too before your performance increases.

Since you didn't describe your particulars, don't know whether you might be running across a PPoE connection which, without allowing for, can impact TCP transfers.  Also not knowing your physical and logical bandwidths, and your ISPs oversubscription policies, your TCP performance can be impacted by ISP oversubscription drops.

PS:

Just noticed what Richard described - he's right - again many variables that can impact TCP performance; he further describes many of them.  What I've just noted, above, can impact other traffic too.

View solution in original post

catalystexpress wrote:

Thank you Joseph, thanks for the valuable comments...

yes i use the speedtest.net to evaluate the link and its been good, however my question is should'nt we be getting a better throughput when the test is after production hour ...

as you say, i have to figure out per user bandwidth to come up with a justification

thanks again

cheers...

Have you considered that just because it's out of hours for you doesn't necessarily mean it's out of hours for the server you're downloading the file from?

You don't say where you did your tests from (across the road? across the country? across the world?) but there are so many factors which could effect your download speeds from a given host which are out of your control they can't be listed without someone speaking up and saying "You forgot about....". The simplest one is what resources does the OTHER end have, and how busy is it? if the server you're downloading from has a 100 meg link and 200 people are downloading from it at the same time, you get 500 kb/s each on average - which won' come close to flooding your link.

If you really want to stress test your link and see if you're getting the bandwidth you are paying for, I can suggest two methods.

The first is to find a big software mirror which is local to you - use traceroute and find the one with the shortest number of hops and the best RTT times - and download a big file using a proper FTP client (not a browser) like Filezilla, as it is more accurate in reflecting what your speeds are at any given monent, as well as your average.

The second is to find a large torrent file, tweak your firewall rules to allow your test machine to participate in peer-to-peer sharing, and download a large file that way (say, a Linux ISO or something similar - I am NOT advocating movie or music piracy).

The second is my preferred method as it tends to remove the problems of the site at the other end having limited resources, but you need to find a reasonably active torrent for this method to be successful. The first works, but you've got to be pretty sure of your software mirror - I don't know where you are so I can't make suggestions, but most educational institutions have reonasbly large software mirrors without much in the way of bandwidth restrictions, so your local university is a good place to start.

Cheers/

View solution in original post

10 Replies 10

Joseph W. Doherty
Hall of Fame
Hall of Fame

Disclaimer

The      Author of this posting offers the information contained within this      posting without consideration and with the reader's understanding   that    there's no implied or expressed suitability or fitness for any    purpose.   Information provided is for informational purposes only  and   should not   be construed as rendering professional advice of any  kind.   Usage of  this  posting's information is solely at reader's own  risk.

Liability Disclaimer

In     no event shall Author be liable for any damages  whatsoever   (including,   without limitation, damages for loss of use,  data or   profit) arising  out  of the use or inability to use the  posting's   information even if  Author  has been advised of the  possibility of  such  damage.

Posting

What kind of connection?  What's the nominal bandwidth, both up/down if asymmetrical?

You're correct, TCP has many factors impacting its performance.  You can't really assume what bandwidth throughput you might obtain unless all parameters are known and controlled.  You might try one of the speedtest Internet sites.

Are there any firewalls or proxies in-line?  They usually only cause a minimal performance impact, but that's not always the case.

As to whether 6 Mbps is the correct or sufficient bandwidth, a simple rule is if network is waiting on you, you have the right amount of bandwidth.  If you are waiting on the network, you could use more.  Simple rule, and the latter is often the case but unfortunately, the cost of the latter often has us as the former.

From a business perspective, if the Internet usage has a business purpose, measure the cost of paying employees to wait vs. the cost of the additional bandwidth.  Sometimes your overall costs can be reduced by paying for more bandwidth and making employees more productive.

Another simple guide, what are people used to?  If your users have 10x better performance on a high speed home connection, poorer network performance on the job can negatively impact morale.

Everyone, of course, generally prefers the fastest system/network performance, but again, in a business perspective, you need to balance cost against productivity.  Only then can you "right" size.  In other words, there's no one right size of bandwidth for x number of users.

Thank you Joseph, thanks for the valuable comments...

yes i use the speedtest.net to evaluate the link and its been good, however my question is should'nt we be getting a better throughput when the test is after production hour ...

as you say, i have to figure out per user bandwidth to come up with a justification

thanks again

cheers...

Disclaimer

The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.

Liability Disclaimer

In no event shall Author be liable for any damages whatsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.

Posting

Yes, you should see better performance if there's less network load.  Actually, it's often better to test before production hours.  With different time zones and users at home, after business hours likely decreases your link's network consumption, but remember much other Internet resources are still shared; often need to see their load drop too before your performance increases.

Since you didn't describe your particulars, don't know whether you might be running across a PPoE connection which, without allowing for, can impact TCP transfers.  Also not knowing your physical and logical bandwidths, and your ISPs oversubscription policies, your TCP performance can be impacted by ISP oversubscription drops.

PS:

Just noticed what Richard described - he's right - again many variables that can impact TCP performance; he further describes many of them.  What I've just noted, above, can impact other traffic too.

thank you joseph... i guess ours is ethernet ...thats what verizon describe the link...will follow all the suggestions ...thanks a lot...cheers.....

catalystexpress wrote:

Thank you Joseph, thanks for the valuable comments...

yes i use the speedtest.net to evaluate the link and its been good, however my question is should'nt we be getting a better throughput when the test is after production hour ...

as you say, i have to figure out per user bandwidth to come up with a justification

thanks again

cheers...

Have you considered that just because it's out of hours for you doesn't necessarily mean it's out of hours for the server you're downloading the file from?

You don't say where you did your tests from (across the road? across the country? across the world?) but there are so many factors which could effect your download speeds from a given host which are out of your control they can't be listed without someone speaking up and saying "You forgot about....". The simplest one is what resources does the OTHER end have, and how busy is it? if the server you're downloading from has a 100 meg link and 200 people are downloading from it at the same time, you get 500 kb/s each on average - which won' come close to flooding your link.

If you really want to stress test your link and see if you're getting the bandwidth you are paying for, I can suggest two methods.

The first is to find a big software mirror which is local to you - use traceroute and find the one with the shortest number of hops and the best RTT times - and download a big file using a proper FTP client (not a browser) like Filezilla, as it is more accurate in reflecting what your speeds are at any given monent, as well as your average.

The second is to find a large torrent file, tweak your firewall rules to allow your test machine to participate in peer-to-peer sharing, and download a large file that way (say, a Linux ISO or something similar - I am NOT advocating movie or music piracy).

The second is my preferred method as it tends to remove the problems of the site at the other end having limited resources, but you need to find a reasonably active torrent for this method to be successful. The first works, but you've got to be pretty sure of your software mirror - I don't know where you are so I can't make suggestions, but most educational institutions have reonasbly large software mirrors without much in the way of bandwidth restrictions, so your local university is a good place to start.

Cheers/

hi darren..........we are located in singapore and downloaded from microsoft (update patches) .

yes you are right we need to consider the other end resource ... infact i did take this into consideration and i never expected a very high throughput, all i expected was atleast of 2500 kbps but i got only around 1300 - 2000 kbps ...

i think with 6Mbps..2500 kbps would be a reasonable speed to expect when the link is not used heavily....

am planning for something simmilar as your first suggestion... and i also need to get a good monitoring tool to get the correct bandwidth usage per user...

this is a ethernet link...though am not very sure

thanks a lot for all the suggestions..and inputs....i really appreciate this....cheers...

catalystexpress wrote:

hi darren..........we are located in singapore and downloaded from microsoft (update patches) .

yes you are right we need to consider the other end resource ... infact i did take this into consideration and i never expected a very high throughput, all i expected was atleast of 2500 kbps but i got only around 1300 - 2000 kbps ...

The Microsoft Update site is not a godo example to use when testing link speeds - I've sen it absolutely race through downloads - literally max out my 100 meg internet link - on occasions, and on other occasions perform like molasses in winter - especially if it's patch day and they've had a big release of security updates or similar.

I can only repeat my previous advice - to get a real idea if your link is working at its maximum, find a software mirror or download site which is "close" to you and try to do a couple of simultaneous downloads.

Cheers.

Yes Darren...i agree with you...and am creating my test plan for the same.....cheers...and thank you very much for the suggestions....

Richard Burts
Hall of Fame
Hall of Fame

Just because it was after most users were off the network does not necessarily mean that you would get most of the nominal bandwidth. There are a number of factors that affect your throughput. Here are some of them:

- if there were a duplex mismatch in the data path from your PC to the provider then throughput would be lowered.

- if both downloads were from the same site (or from the same server) then throughput would be impacted.

- the TCP window size negotiated between your PC and the remote server could impact throughput.

- if there were retransmission of packets it lowers throughput.

- the distance from your site to the site you download from impacts throughput.

If you can account for these variables you might have more insight into your throughput.

HTH

Rick

Sent from Cisco Technical Support iPhone App

HTH

Rick

Thank you very much Richard, this is helping me to get a more clear vision on this issue.. will look through the points and work on it.....cheers and thanks again....

Getting Started

Find answers to your questions by entering keywords or phrases in the Search bar above. New here? Use these resources to familiarize yourself with the community: