Paranoid about the NSA? The case for dumping cloud's Big 3
Can you achieve security through the obscurity of regional ISPs?
Internet Service Providers (ISPs) may be the most important public cloud providers of the next decade. Hosting your data with an ISP has a number of advantages over choosing the dominant American cloud providers: advantages that run the gamut from technical to political.
ISPs have been in the co-location business practically since the internet began. Many have offered hosted services (typically e-mail and web server space) for at least as long as the World Wide Web (and the browsers required to interpret it) have been around.
The idea that ISPs might be interested in hosting cloud services is thus not particularly novel, nor is it particularly hard for them to stand a cloud service up today. HP will cheerfully sell you cloud servers, and all of the software to make it go, too.
Almost every start-up wants in on this too. Maxta is in bed with Mirantis to form a hyper-converged compute + storage "Openstack cloud in a can" offering. Yottabyte is building its own fully hybrid cloud unicorn thing... even newly out of stealth Springpath is planning a pre-canned cloud in a box via partnerships.
Companies that don't quite qualify as start-ups anymore like Nutanix are all over this, and the big boys are trying to buy in too. Look to Cisco's Metacloud purchase or the Dell-Microsoft franken-pesudo-Azure thing.
A quick web search shows that many ISPs already have their own public clouds. BT in the UK, Verizon in the US, Telus in Canada and Optus in Australia are all examples of ISPs in the four major Register reader countries that offer cloud services.
Chances are almost all of the majors (and most of the not-so-majors) either have a cloud offering today, or will within the next 18 months. So why are they doing this?
The technical argument
The technical argument for ISP-based clouds is actually pretty simple. If you are a business that is located only in one restricted geographic area, you can probably have all your business locations using the same ISP.
It's not particularly hard to get great deals on really high-speed data links to your ISP – what's hard is getting high-speed data links to the internet at large for cheap. If I want to slam a large amount of traffic back and forth between several locations, all within the same province and a data centre that is owned and operated by that ISP in that same province, there's a really good chance I'll be able to pay "on net" rates for my cloud traffic.
On net rates are cheap. I can then set up something with the ISP such that any of my traffic that does go out to the internet at large gets charged to me per gigabyte, but because that will probably be just some general browsing it's not going to cost much. The real bandwidth is between me and my data centre, and by staying within my ISP's network, that traffic won't break the bank.
If I maintain a back-up link through a second internet service provider, then whenever I failover to that connection, the traffic will have to go through "peering" and thus "over the internet" to get from me to my ISP's data centre. In this case, those data costs would be the same as if I were using one of the big American cloud providers, but I don't have to use the back-up link unless my main ISP's link has fallen over.
One of the other advantages to ISP-local clouds (or, really, any regional cloud provider located physically near you), is that physical proximity enables services that are otherwise difficult.
Take data recovery as an example. For those using cloud services simply as an offsite back-up repository there is always the problem of getting your data back when a restore is required. Downloading can take a long time. A local cloud provider can pop the data onto a drive and courier it over the same day. A cloud provider in another country takes longer, if they even offer that as a service.