Commentary: Sometimes, The Old Ways are the Best
So I had an interesting day recently: It was the culmination of a few weeks' work preparing a company for an important internal webcast.
When I say company, I mean bank, and when I say important, I mean one of the few banks that is doing very well at the moment.
A little history: This bank came to me a couple of years ago for a proposal to deliver their annual staff meeting over their Intranet. They span a dozen countries and have several thousand staff. They looked at my (superbly well written!) proposal and decided that it looked easy enough to deploy a Windows Media multicast themselves.
So they tried.
For readers of this magazine some of the errors they made were schoolboy grade, but their 40-strong IT department felt confident that, after reading my proposal, they could do it in-house.
Feeling a bit put out that the proposal had been too clear-cut, and that they had used my "sales time" as a way to get some direction to try to do this without me, I was (behind closed doors) quite amused when they failed to multicast and ended up with a few thousand users rolling over to unicast connections to their video server, thus causing huge problems across their corporate WAN.
I was even more amused when they came to me with their cap in hand and asked me to supervise the event end-to-end this year to avoid the CEO and Chairman looking like fools (again).
Now, the broadcast industry has very high production standards. Broadcast clients will always expect everything to work in sync with their workflow and perfectly every time. There is little margin for error.
That said, the broadcast industry has the luxury of being the end client of every telecom network they contract to use. They demand quality and should expect quality.
Corporate enterprise multicast is a completely different beast.
In a banking environment, the corporate network has not been put in place to help receptionists lookup recipes or surf job sites. It is not there to enable emailed arrangements about the weekend get to the managing director from his wife (although that will surely form part of the requirement). No. A corporate banking intranet is put in place specifically to enable banking. Thousands of transactions with split-second timing and accuracy that enable the world's commerce to flow.
These networks are not "a bit of fun." They are serious, financially and economically impacting systems that must be secure, highly available and perform their primary function (trading) at ALL times.
So trying to squeeze marcomms out of such a system needs a particular kind of love and care.
Last year's attempt to multicast, failing to huge quantities of unicast, crippled their network for several hours. Once the notification of the webcast of the CEO and Chairman's presentation (which critically for the mass of staff contains information about their annual bonuses) was published, there was no turning back.
When the operators didn't really understand Windows Media Multicast and it all went crazy at the core of their network, they not only swamped the core network, causing huge problems for highly latency-sensitive financial data flow, but they also found that they couldn't control their servers and encoders and stop the broadcast. The flood of stream requests prevented them from reaching the unicast server and stopping it.
Lots of egg on their faces. And a few questions about why they tried to go it alone, rather than using the consultant.
So this year we spent a few days (on the clock, I hasten to add) gradually deploying a test stream over their global WAN. We incrementally added regional offices and took reports about their user experiences and network traffic flows.
We brought up logging systems to answer the "So, how many users?" questions that always come within 3 minutes of the end of an event. We checked for multicast broadcast storms, something that was the bane of multicast from the 1990s to about 1997 and gave it a poor reputation. We checked address allocation, and that there was no conflict with their financial data multicast flows. We finally ran test loops throughout the day of the event and gradually picked up their audience to the live event at 5 p.m. today.... and it went like clockwork.
In fact the network managers were slightly "freaked out" since they realised we had several thousand viewers connected, but at the same time could only see 400Kbps of traffic on their 1Gbps network. Despite understanding how to configure and manage multicast they didn't fundamentally know how it worked so found it challenging to understand how so many Gbps of data could be received while only a few Kbps were being sent.
All in all, the most amusing part of the event was when a laptop that was having its VGA output scan-converted and rebroadcast on their internal COAX network to a handful of reception area plasma screens decided that, in the middle of the CEO's bonus comments, it needed to switch to screensaver, bringing up amusing photos of the IT guy's family holiday across the company's public areas until he was found, and rushed to the laptop to log in again.
Multicast has some way to go yet, and seeing Flash, Wowza, and Microsoft keeping the flag flying for multicast is very heartening. It is an increasingly important way to reach across corporate intranets, and for me the interesting thing is that despite Windows Media beginning to be considered as a "deprecated" medium, this event and also a second deployment I have just been asked to conduct across several dozen corporate sites reminds me that when something works it "just works."
All the new adaptive bitrate Technologies are very important in the consumer market, but unicast in high volume wont been seen "dead" for several years in the corporate intranet environment. These networks are not built for video-they are built for business. We (with our video tools) are welcome guests, but will be asked to leave if we compromise the core requirement for those networks.
It's good to see multicast still featuring in the newer servers, but many corporates are only just now starting to trust Windows Media Multicast, and that technology is a decade old (or more). We should not assume that new is great in all environments... sometimes the old stuff that just works is the best.