This is an interesting topic raised, and i have searched the forums for related material, of which there is enough!

I will comment here though
The following is to the best of my knowledge and experience, and various aspects my be incorrect.
Telkom runs a vast network, and in terms of technical profficiency at the datacentres, larger POP's and INX's they are pretty damn hot.
You dont just set up a load balanced transparent-proxy network layer for the kinda traffic they must be pushing... easily! And lets be honest (and think of the scale here) it works, and most of the time well! (Every one of you peeps is proxied, all your porn, warez, music, mindless rubbish, research and work related stuff)
Unshaped as far as i know and i think i tested this once, does not route via a proxy.
I do not advocate or agree with telkom, there strategies, monopoly, pricing, policies and underhanded tactics, telkom is the atypical
incorrigible! Bully, and they know it.... Policies such as the one we find ourselves discussing now, '
transparent-cache', are a forgone conclusion and in reality simple economics..! for TELKOM
A quick google search with the right keywords can turn up loads of interesting information on a variety of topics. i'll comment on some as we go.
Firstly for those who are going - "
Huh, transparent-cache ?" / "
But dont we eat squid ?" - check these links.
My explanation (
somewhat skewed, incorrect, and simple - but i hope it gets the idea across)
Telkom has initial caches (this is assumed) at each Edge Router (They are the localised gateways for a region), these are scattered around the country and everyone connects through one. These initial caches take care of - some of the load from
HTTP(
port 80) and very little
FTP(
21) requests / traffic.
They are slaves to what i think are between 4 - 6
master caches (could be alot more - ie ... clusters) These master caches are the transparent-proxy's that you all see on your outgoing requests! should the request even get that far... So if the localised cache cant deal with the request it will query one of the master caches, and if that cannot service you, it makes a full request to the server in question GET's your content and delivers it to your screen - but not before caching it (maybe replicating it across the trans-cache-squid network) for the next ZA'peep wanting that same content.
They are great in concept, supposidly increasing your browse speed and file tansfer speeds for smaller files! From the view of a large corporate behemoth, set on profit margins... They are an easy, practical solution saving more and more as the throughput increases "
Economies of scale" is the magic term here, and i will illustrate my point in a minute.
"I will leave you all to research the Pros, Cons and finer details..."
TENET
TENET - "
Tertiary Education Network"
Is the association formed to facilitate, monitor and control an agreement with telkom to supply bandwidth to campuses country wide.
I reccomend checking the link out and reading some of the documentation there about the network and agreements in place, very interesting stuff some of which i will highlight now.
Now check this out:
Transparent Cache Bypass List
Quote: "
In terms of the GEN2 agreement, TELKOM operates a large web caching appliance that transparently caches most web requests from most GEN2 sites. This started under the HEIST agreement in October 2004, and reduces the requirement for shared SAT-3 bandwidth by about 10%. TENET bore the upfront cost of the cache appliance and bears the monthly service charges from TELKOM." -
Telkom has vast experience in this field and i did read a couple of other references that validated telkom's experience and proficiency in setting up large distributed trans-cache networks.
TENET Transparent Caching Report - PDF
Quote: "
If TENET were merely to reduce the IBF to 1.30, the monthly Telkom invoice would fall by some R275,000 per month. Five months’ of such savings would pay for the cache appliance outright. There is thus no doubt that the introduction of caching would generate significant economic benefits for the institutions." -
Now i URGE! everyone to read this document in full!, based on the TENETreport and the amount of traffic TELKOM must be sending and recieving a quick mental calculation highlights the savings telkom must realise with this approach
Some comment on thread posts so far:
- As far as i know all SAIX / TELKOM ADSL Iternational traffic goes out over fibre.
- This end of the month madness, is very wierd and telkom must do some sort if purge or cache maintenance to cause such wierdness. (this is really a guess...)
- This could all be related to telkom implementing these new IP ranges and i just spoke alot a garbage...
Anyway enough rambling, i hope i made some sort of point, and that atleast one of you finds this interesting.
