Of all the EMC press releases that appeared last week, it was this one on some ‘New Cloud Storage Service’ that caught my eye. Not because I know anything about cloud storage however, it was the mention of AT&T, as a major telecoms provider, that renewed my continued wish that, as a consumer, cloud storage would start getting closer to my network at home. Before I explain, let me describe (by way of example) the limits of my knowledge on storage, lest you infer that I know even a smidgen as much about storage as some (ok, most) of my EMC colleagues. My little Windows Home Server (at home, obviously) started running out of space and I went to Amazon and bought the largest/cheapest disk I could find - when it arrived and I shoved it into my HP MediaSmart Server. Done. Finished. That about exhausts my knowledge on storage – you have been warned and obviously my views on storage do not reflect those of EMC gurus.
Back to cloud storage. If I want to stream a video to my XBox I will have no bandwidth cost and no latency if I stream it from my home media server – as opposed to streaming it off some server in, say, New Zealand – where the bandwidth costs will be high (think of all the middlemen in between) and the latency will be a problem. But, streaming it off my local network is not going to work – firstly I don’t have any videos on my local network and secondly I don’t have the local storage to keep them. The location of data that we stream for video is maybe not as far as New Zealand, but it is a long way away – which makes no sense.
ISPs have been bleating for years that the BBC iPlayer consumes way too much bandwidth and, at 12GB/s that is served, it would seem that it is the ISPs who are paying. While the BBC may be of the opinion that it is easier and cheaper to download on-demand directly from their own servers, it is conceivable that HD content would consume all available bandwidth. To me, it seems that the costs are not in the local loop (between the local exchange and your house), but rather in the Interwebz further upstream, where the demands of thousands of consumers need to be channelled across fairly fat pipes. I don’t see, in the near future, that it is cost effective to have a single, say 30GB, video streamed on demand to possibly millions of users. If digital distribution becomes the way of the future how could you do a worldwide simultaneous distribution of ‘Night in the Museum 12’ in super-HD from a single point of distribution without crashing the Interwebz and upsetting paying customers? The key, I believe, lies in making use of the local loop.
Imagine, for streaming video on demand, that you could put the top 200 movies (in HD format) in the local telephone exchange? You could cater for 95% of viewers from the local loop, which is essentially a fixed cost. You could take a movie being released (to DVD) on a particular Monday and distribute it to all of the exchanges overnight when latency is not an issue and the only contention is with *** downloaders. The idea of storage closer to the destination is not new – proxies have been trying to do this for a while, but are a bit dated can’t handle dynamic content and the more sophisticated authorization requirements. Bit torrent is also suited to sharing of more local bandwidth but doesn’t optimise costs.
So, the intriguing thing about the EMC announcement is the involvement of AT&T. While the Synaptic Storage as a Service is very enterprisey and not really related to my desire to view low cost HD video on demand, the idea of a network provider (who also owns some local loops) getting into cloud storage makes me jump to conclusions as to where they are going to locate storage in future. Obviously centralised, secure storage is desirable, but if you have a huge big worldwide network then why not put the storage where it makes the most economic sense (which includes ISP costs of providing bandwidth)? Once you put your storage on the local loop it can be used for up and download of data. Why not make a backup to the local loop that eventually manages to copy itself elsewhere?
One of the principles behind cloud computing is the ease at which resources can be provisioned and maintained. You could deliver storage in a crate to the exchange, plug it in and let the software sort itself out - you only need to send an engineer to the site again to rip it out and put in a bigger one. So just maybe the storage cloud experts and their network buddies are working on some stuff that will make a difference in our lives.
I may be off the mark, but the abstractions that the cloud presents us allows a whole lot of ideas to be bandied around that may have been too much of a challenge before.
Simon
@simonmunro