A Numbers Game

Once the regulations have been promulgated, the operators will still be required to build a central database - at a shared cost of up to R50m . This database is needed because the three-digit prefix on a telephone number will no longer be sufficient in determining the network a subscriber uses.
Click to expand...
R50m to 'build a central database'?! How much does it REALLY cost to deploy a database capable of holding say, 120m records (40m people, x 3 for redundancy)? I'm no database expert, but that sounds ridiculous! Even the ridiculously-high cost of some of the proprietary database products out there, plus the machine costs, would hardly justify that high a price tag, particularly for data that would be largely static?! Any volunteers for an open source database project solution... :p
 
The big thing that worries me about number portability is how are consumers going to know which network they are calling and thus what they are being charged? If your fixed line number has been ported to a cell phone how is a consumer going to know he is now paying mobile rates to call you? I think this is a big issue and nothing has been said about it at all.

Perhaps we can look forward to annoying messages announcing "you will be billed at Rx per second for this call" instead of ringing sounds.
 
mbs said:
R50m to 'build a central database'?! How much does it REALLY cost to deploy a database capable of holding say, 120m records (40m people, x 3 for redundancy)? I'm no database expert, but that sounds ridiculous! Even the ridiculously-high cost of some of the proprietary database products out there, plus the machine costs, would hardly justify that high a price tag, particularly for data that would be largely static?! Any volunteers for an open source database project solution... :p
Click to expand...

These databases, I believe, are quite costly due the the speed requirements placed on them.

This requires super hardware and even better software.

If this databases fails, or is slightly slow, consumers will experience it as "useless service provider being overloaded again"

Typical sub second response times are required.

Redundancy is also essensial and adds another major cost component to the system. If the system is down, nobody will be able to make a single call as number portability introduces a single point of catastrophic failure.

I do hope that neither Telkom, Sentech or iBurst gets their fingers close to this database!
 
mbs said:
R50m to 'build a central database'?! How much does it REALLY cost to deploy a database capable of holding say, 120m records (40m people, x 3 for redundancy)? I'm no database expert, but that sounds ridiculous! Even the ridiculously-high cost of some of the proprietary database products out there, plus the machine costs, would hardly justify that high a price tag, particularly for data that would be largely static?! Any volunteers for an open source database project solution... :p
Click to expand...

I'm not a telecoms expert, but I don't think it's simply a case of deploying a database...the db has to hook into the switching systems of the telcos in order to do the lookup and then to route the call. Plus the lookup and switch neeeds to be as fast as possible - who wants to sit and wait for 10-15 seconds when making a local call? I'd say R50m is reasonable...
 
The database is not the costly part, Its the hardware it has to run on. Theyre probably looking at a fully populated superdome type machine ( 64 PA-RISC or Itanuim CPU's, probably about 40-80 Gig of ram, and an EMC symmetrix or HP XP type fibrechannel disk array ).

The superdome goes for 20 Bars second-hand where a bar is Rm1.

Then you havent even considered a duplicate system for disaster recovery / redundancy ( hell, maybe even a triplicate ). For that setup i'd run the disks on a raid 1 mirror with bcv disks for the backup. That all costs just slightly more than anyone can conceive.

And then, their biggest cost would be telkom again. If its a central database all providers must have datalines to the db, at the appropriate speeds + all the backup circuits again......

So Rm50 is not all that much when you consider all factors.

PS: an HP superdome with oracle 10 hold the current speed record doing something like 4million database transactions per second!!!!

Obelix
 
If there's one thing that's pretty obvious, it's that some serious thought must be given to this endeavour, before bandying figures about in the press. Interestingly, this is supposed to be a shared project (presumably amongst the main providers, plus all the resellers - maybe even the SNO...). I'd love to be a fly on the wall at those project discussions!
 
Lets see - R 50 Million.

An absolutely fabulous P4 HT 3GB machine with 4 120GB mirrored hard drives will set you back R18000.00. (Round up to 20 000)

For R10 million, I could run 500 of these in a clustered segmented routing envionment, much like google are going.

That leaves 10 million to buy google's expertiese in setting it up, and 10 mil for connecting to the backbone, and 5 million for training.

Yup - not a bad budget. I get 5 mil in my pocket. Melikes.
 
And the remaining R10m? Your figures tote up to R40m...
/mebbe pass on to MyADSL forumites? :p
Any volunteers for a R10m open source project?
 
i'm an open source for R10m...

it is going to be interesting watching this unfold - projects which require co-operation and communication between competitors have their own intrinsic value
 
R50 million is pocket change for any single one of these companies so it really is no significant investment for a combined solution. When I worked at MTN in 1998 one of their datacentres was worth over R200 million and is probably worth far more these days.
 
You must log in or register to reply here.
Back
Top
Sign up to the MyBroadband newsletter
X