Connectivity: the weakest link in cloud computing? Cloud computing makes you so dependent on the network, and often even the public internet. So how can it be a sensible option? If the comms slow down, the user experience takes a dive. Lose the link completely, and you’re stuffed. We have been hearing such questions and objections from nay-sayers ever since commercial hosted …
I have worked with a lot of small to medium sized businesses who to varying levels have come to see their internet connection as somewhere between important to vital. As more and more services move to the cloud the network connectivity and availability becomes ever more vital. When things have gone wrong with the network - and they will - small companies have exactly zero leverage with the ISP.
It is not too bad when there is a clearly identifiable fault where for example there is a complete loss in connectivity. The ISP recognises there is a fault and gets around to fixing it. Where it gets very, very difficult is where there is an intermittent fault or a fault not easily identifiable as being with the ISP. In these cases I have seen ISP's disown the fault completely or provide the bare minimum effort to try and identify the problem.
For domestic users moving to the cloud is even more of a gamble. There's an intermittent fault with Virgin Media broadband connectivity at the moment where it looses connectivity for a few moments. It's a bit of a pain when your browsing the web but a page refresh is acceptable. However, even this occasional intermittent fault is completely unacceptable to my two teenage kids who loose connectivity to whatever on-line mayhem and violence they happen to be involved with at the time. As a customer I can report the fault but I certainly can't expect a quick remedy (but I can expect to be told to reboot the router every time I call).
All I'm saying is that as you move more services/data to the cloud the network connectivity is more and more vital. As a consumer or small business you have very little influence on the levels of service from your ISP. I might use the cloud but I will have a disaster recovery plan in place so that if I loose connectivity, I don't loose everything.
A company I support has this very problem, the connection speed is absolutely abysmal because of never-ending packet loss.
Phoned BT who 'checked' the line and then announced it must be an issue with the LAN. Cue more testing at my end, and it still appears to be a problem outside the network. A big company would stamp it's feet and make BT jump to, but a 4 man outfit just isn't able to do that.
They've asked me about Cloud before and I quite honestly told them why I thought it would be a really bad move for them to start relying on Cloud based services (and this was before the Internet issues).
Funnily enough they seem to agree completely, and I can't see them being convinced to move for a while!
ADSL service over copper are not very reliable for a whole raft of reasons - underlying dial-up technology, copper wiring issues, interference and poor fault resolution times. A reasonable estimate of ADSL annual availability is around 99% - 3-4 day's outage a year. Single sites might go 15 months without an outage, but large estate of sites will experience faults with 1/15 of their sites each month.
Because of the dial-up nature of ADSL, most (80%) faults are cleared by rebooting the router - hence why ISPs tend to lead with that, rather than trying to wake up an engineer and get them to test the line or look at an exchange. Not every customer has tried it, even those that say they have.
Complex intermittent faults are always harder to diagnose and fix - and the customer's investment in their connection is often proportionate to the amount of support resource aimed to resolve the problem. Buy into an enhanced or superior care package and you can access engineers during the night and expect more attention, with quicker fix times and better compensation for failure to hit service restoration.
DSL is still relatively rubbish though.
Chosing a different connectivity mechanism can make things more reliable. EFM is based on multiple copper pairs - meaning a single faulty pair can degrade, rather than disconnect, the service. Business-grade EFM services can be had for a couple of grand a year.
If we assume that an ADSL line is 99% available, and an EFM line 99.7% available - and we combine that with an understanding of the cost of an outage (usually proportionate to revenue associated with the site and the IT assets involved), then we can generate an idea of the value of site that might justify chosing a more reliable connection.
We can do the same for a fibre-based Ethernet (~99.9% available, ~£8,000 a year rental for 10Mbps).
Many businesses have experienced the application creep where the Internet used to be limited to email and browsing, which is not often critical to the business. Increasingly Internet-facing business apps have appeared and been adopted without revisiting the connection and how valuable it is to the business.
I'll be interested to see the overall reliability characteristics of FTTC, because for me, this is more important to SMEs than capacity/throughput. Cabinets with single power feeds, lack of engineers trained in VDSL equipment, but less copper in the mix...
It's not just the network.
It's everything that's outside of your immediate control, or that you cannot make, at least, a direct effort to fix.
It's levels of support telling you to do silly irrelevant things your end; its' marketing people telling you they'll sort it out (they always lie; they did when they sold you this stuff and it won't change). Ultimately, it's a press release saying that only a small number of customers is affected, that the supplier is working with blah blah blah to get things back on line, that lessons have been learned, and...
OK, So you are one of the small number of customers, seething. The press releases will not help.
How come the commercial world is so naive?
(well... my theory is that it is because it is now run by bullshitters that were promoted by bullshitters, and that these guys actually believe each other. It does not bode well for the future)
There are so many ways connectivity can fail -
http://www.infosec.co.uk/page.cfm/action=Archive/ArchiveID=16/EntryID=270
The benefits of a fully redundant cloud data centre are only any good if you can reach it. But there's more at stake than just access to the cloud. We're rapidly replacing a wide range of well-proven infrastructure with faster, more convenient but much less reliable alternatives - and also converging all our communications onto fewer and fewer of those less reliable channels. There'll be a tipping point very soon.
If your connection is not available for a short while then its a bit of an inconvenience. If it's not available to your clients that's a bigger concern as more products and services are sold on-line.
The advantage of cloud services for client facing businesses is that the cloud infrastructure vendors provide much more reliable services than are available for the same cost to the office. We're big users of Amazon Web Services. Our servers run in the US-East availability zone - the one affected by the outage in April but we we unaffected by it such is the redundancy built into services like Amazon, Azure, Rackspace and so on.
Because connectivity to SaaS services (our own and others) is important we have a backup. We have both Virgin *and* BT to the office at a marginal cost of 15/month. Our routers fail-over if one or the other is not available. Someone might be disrupted for a few seconds though it rarely happens and (so far) it has never been the case that both services are unavailable. And even if that were to happen we have smart phones with data contracts that could be used in an emergency.
In my view this is an old argument and the benefits of using on-line service now outweigh the risks. On-line data can be stolen but its a fallacy to think an office is secure. There are solutions to mitigate risks and the solutions don't need to cost much.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
