Hi, I know some of you are very experienced as to the working of networks. I wondered whether there is some accepted way of determining bandwidth needs based on the network traffic over time. For example, looking at the figures for the network traffic through the server interface, we have hourly, daily and monthly figures. If everything were linear, taking the hourly figure and dividing it by 3600 (or the daily by divided 3600*24) would give us the required bps, but this average is pretty meaningless. Those of you who have experience and education in this area, where can I look for guidance (links?) and do you have any rules of thumb you'd care to share? I'm actually looking for this for a web server not VoiP, but any info is welcome. It seems obvious to me that taking the per second average and multiplying it by some kind of seat of the pants number must give a decent idea? WHat is that magic number? Thanks in advance for any ideas. /r
http://www.asteriskguru.com/tools/bandwidth_calculator.php ~ Andrew "lathama" Latham lathama at gmail.com * Learn more about OSS http://en.wikipedia.org/wiki/Open-source_software * Learn more about Linux http://en.wikipedia.org/wiki/Linux * Learn more about Tux http://en.wikipedia.org/wiki/Tux On Thu, Jun 24, 2010 at 3:46 PM, Randy R <randulo2008 at gmail.com> wrote:> Hi, > > I know some of you are very experienced ?as to the working of > networks. I wondered whether there is some accepted way of determining > bandwidth needs based on the network traffic over time. For example, > looking at the figures for the network traffic through the server > interface, we have hourly, daily and monthly figures. If everything > were linear, taking the hourly figure and dividing it by 3600 (or the > daily by divided 3600*24) would give us the required bps, but this > average is pretty meaningless. > > Those of you who have experience and education in this area, where can > I look for guidance (links?) and do you have any rules of thumb you'd > care to share? I'm actually looking for this for a web server not > VoiP, but any info is welcome. > > It seems obvious to me that taking the per second average and > multiplying it by some kind of seat of the pants number must give a > decent idea? WHat is that magic number? > > Thanks in advance for any ideas. > > /r > > -- > _____________________________________________________________________ > -- Bandwidth and Colocation Provided by http://www.api-digital.com -- > New to Asterisk? Join us for a live introductory webinar every Thurs: > ? ? ? ? ? ? ? http://www.asterisk.org/hello > > asterisk-users mailing list > To UNSUBSCRIBE or update options visit: > ? http://lists.digium.com/mailman/listinfo/asterisk-users >
ISP 10% rule is what you are asking about.... expected that average usage is 10% of total subscribers with bursts higher.... ~ Andrew "lathama" Latham lathama at gmail.com * Learn more about OSS http://en.wikipedia.org/wiki/Open-source_software * Learn more about Linux http://en.wikipedia.org/wiki/Linux * Learn more about Tux http://en.wikipedia.org/wiki/Tux On Thu, Jun 24, 2010 at 3:46 PM, Randy R <randulo2008 at gmail.com> wrote:> Hi, > > I know some of you are very experienced ?as to the working of > networks. I wondered whether there is some accepted way of determining > bandwidth needs based on the network traffic over time. For example, > looking at the figures for the network traffic through the server > interface, we have hourly, daily and monthly figures. If everything > were linear, taking the hourly figure and dividing it by 3600 (or the > daily by divided 3600*24) would give us the required bps, but this > average is pretty meaningless. > > Those of you who have experience and education in this area, where can > I look for guidance (links?) and do you have any rules of thumb you'd > care to share? I'm actually looking for this for a web server not > VoiP, but any info is welcome. > > It seems obvious to me that taking the per second average and > multiplying it by some kind of seat of the pants number must give a > decent idea? WHat is that magic number? > > Thanks in advance for any ideas. > > /r > > -- > _____________________________________________________________________ > -- Bandwidth and Colocation Provided by http://www.api-digital.com -- > New to Asterisk? Join us for a live introductory webinar every Thurs: > ? ? ? ? ? ? ? http://www.asterisk.org/hello > > asterisk-users mailing list > To UNSUBSCRIBE or update options visit: > ? http://lists.digium.com/mailman/listinfo/asterisk-users >
> ISP 10% rule is what you are asking about.... > expected that average usage is 10% of total subscribers with bursts > higher....But remember to plan well for those bursts and ensure you have sufficient excess capacity. Certain events can have a significant effect on your burst pattern: some fellows are kicking a ball around in South Africa for three weeks, which is having an understandable effect on bandwidth usage globally. Same happened a couple of years ago during the Olympics. Regards, Chris
Randy R wrote:> Hi, > > I know some of you are very experienced as to the working of > networks. I wondered whether there is some accepted way of determining > bandwidth needs based on the network traffic over time. For example, > looking at the figures for the network traffic through the server > interface, we have hourly, daily and monthly figures. If everything > were linear, taking the hourly figure and dividing it by 3600 (or the > daily by divided 3600*24) would give us the required bps, but this > average is pretty meaningless. > > Those of you who have experience and education in this area, where can > I look for guidance (links?) and do you have any rules of thumb you'd > care to share? I'm actually looking for this for a web server not > VoiP, but any info is welcome. > > It seems obvious to me that taking the per second average and > multiplying it by some kind of seat of the pants number must give a > decent idea? WHat is that magic number? > > Thanks in advance for any ideas. > > /r >It depends on the charging structure your ISP is using. From an ISP point of view when they purchase transit links all pricing is done on the 95th percentile basis. So for example you might purchase a 50Mbps connection over a 100Mbps interface. You would get a guaranteed 50Mbps but could burst up to the full interface speed. Bandwidth measurements are taken normally every 5 or 10 minutes. The 95th percentile is calculated by discarding the top 5% results and the next highest reading is the 95th percentile. If this is below the 50Mbps then all is fine otherwise you will receive a bill for overusage. For a web server this is probably a good start but would depend on how spiky your bandwidth graphs are. You might want to lower the speed if you tend to have lots of people downloading on a particular day of the week for example as the 95th percentile would be much greater than your average bandwidth. If you just look at your bandwidth graphs you will probably get the best idea of a suitable figure.