Saturday, January 15, 2005

Broadband for Business article

This article appeared in CMA Newsline during 2004

Broadband for Business By Ian Grey, Membership Services Director, CMA

Many of us will have looked at providing DSL connectivity to home workers and branch offices. The lure of cheap DSL Modems/Routers, £20-£40 a connection and the possibility of 2Meg downloads to rival Megastreams certainly looks attractive compared to the prices we pay for leased lines.

Quite a few of us will have piloted DSL for some locations or workers and had mixed results in the process. A friend of mine used to refer to disastrous implementations as a “partial success”, generally interchangeable with the robust Army expression “SNAFU”. After our own partial successes, we will have either decided to live with the limitations or abandon it due to the uncertainty.

Some of us who are early adopters and have actually implemented DSL for business may have found that our concerns were unfounded, or at least are for now. However, there is always that nagging doubt in the back of our minds that the gaps in the SLA might indeed backfire on us and perhaps we should have put in that expensive tail after all (or paid for always-on ISDN). SLAs are often worth the paper they are written on, i.e. there may be cause to get compensation out of your Telco on occasions of partial successes, however it never reflects the true cost to the business of the lost productivity or goodwill as a result of sustained outages. Energis have been rumoured to have provided a 100% availability guarantee 24/7, which as any fool knows is an accident waiting to happen. BT on the other hand, are much more straight forward with their guarantees, generally being based on terms such as “four hour response”, which means that they might actually get round to turning up mid-morning on Monday if something goes wrong mid-afternoon on Friday. (Unless it is DSL, of course, when they won’t even do that).

Those of us who have a home Broadband connection would have concerns that the consumer experience is at best variable and could be very unsatisfactory if the typical problems that occur at home impacted at enterprise level. (They often do, but for different reasons, of course!). Typical problems include: unable to connect to the Radius Server for authentication, packets disappearing into oblivion (whether the Internet at large or the providers’ home servers), FTP transactions that slow down to a crawl and never complete, DNS responses that deny that they have ever heard of well known sites, POP poisonous payloads, PING response times that would challenge a snail and throughput that would sometimes put a 300 baud modem to shame. One of our Managers resents spending £4 to ring BT in Bangalore every time an email refuses to download blocking the POP account, an event apparently common enough to be a nice little earner for BT Openworld (& forthcoming churn for that particular customer).

Fortunately, relief is at hand. The use of DSL to connect a location into a Business Grade service is now starting to become viable, widely available and affordable. Using what Thus christened PaDSL (Private Access DSL), cheap tails into robust IP networks with the ability to implement Quality of Service now looks realistic. It is also possible to choose a contention level to suit your pocket, from 1:1 uncontended through 5:1, 10:1 and 20:1, even 50:1, depending on what the supplier offers in their portfolio. Where DSL isn’t available, Frame Relay can be used with one PVC into the IP cloud, giving effectively any to any if the cloud is MPLS. There are also dial-up ISDN possibilities into MPLS, however the solution is a bit of a dog’s breakfast if you want it to be “always on” rather than ad-hoc for nomadic staff as BT is still going to want 1p a minute from any OLO unless something like Friaco is used.

There are two flavours of Business DSL, using what BT call IPStream and DataStream. IPStream uses BT’s Colossus Network to transit to the destination, whilst DataStream uses ATM to get to where it is going. The services are point to point, however you can’t make the other point another Branch site, it needs to be either a switching network (such as an MPLS cloud) or a fat pipe into a Head Office (essentially the same arrangement as a non-BT ISP). How secure and robust these solutions are depends on who you talk to and their vested interests, however IPStream is essentially BT’s part of the Internet whilst DataStream could be thought of as parts of BT’s CellStream ATM backbone. All DSL traffic is initially carried over DataStream, however the ATM traffic types are not the same and IPStream generally gets the raw deal. The following diagram should make it clear as fairly thin, watery, diluted mud. (The icon top right represents “here be dragons…”)

(Nope, I haven't embedded the diagram, sorry. Do you want it on a plate?)

There are variants of this diagram and the DataStream proponents will give you a good story on why IPStream leaves something to be desired. Similarly the IPStream proponents will also give you a good story on why DataStream is unnecessary expense. Most people will also tell you that contention is not a problem but they may include caveats such as “today” or “at present”. It may not be but if I am trying to maximise my utilisation, you can bet that other network managers that I am contending with are thinking the same way.

It seems that contended ADSL into MPLS is about to hit the big time, there is talk of a 256k/256k pseudo SDSL (i.e. capped ADSL 512k/256k) appearing widely by the summer in addition to the real standards based SDSL piloting in Coventry and London.

I have been disappointed that DSL has not been seen as an opportunity to make Frame Relay tails even cheaper- whilst it isn’t the future it would have made a nice stop-gap for 5 years or so whilst IP matured. (My Boss agrees, as he intends to retire by then!) It seems that over the pond, the Merkins are breathing fresh life into Frame to make it interconnect to the brave new world.

Frame has it’s own problems, of course, particularly management of PVCs (permanent virtual connections) where a truly meshed network requires n! PVCs (n factorial, i.e.n*n-1*n-2 etc), each of which has to be configured and paid for.

The nice thing about frame, however, is the committed data rate, where a 64k connection can sustain a throughput of (say) 32k indefinitely, but will burst up to the maximum as required for short periods.

And the good news is…… it is coming to DSL! I have seen one provider presentation showing availability of bandwidth guarantees over DSL available as a product during 2004. Given a choice of contention (& uncertainty) versus non-contention (& more money), the third way, a more meaningful (and indeed measurable) minimum bandwidth guarantee wins hands down.

No comments: