I don't even see how a cloud storage could properly supplement a established backup solution. It has other usages- such as with PaaS, if the company's storage requirements change over the year for whatever reason, on-line storage could be used for part of year to supplement their own on-site hardware. For backups, Of course, hard drives can fail locally, off-site backup tapes can be unreadable, etc. but restoring and saving backups is hours faster than uploading hundreds of gigabytes of information, regardless of the speed of the internet connection (Tape drives generally transfer at close to 340Megabits a second sustained).
Online storage moves the task of keeping your data safe from the IT administrators on your company's payroll to IT administrators of another who don't have a inherent stake in the safety of the data beyond it being their job to do so. Most companies should already have a backup strategy involving on-site and off-site full and incremental tape backups that have been tested, so- at least in my opinion- adding Cloud storage to this isn't very useful in general, but it can depend on the company's specific needs. But that's the problem, very rarely do people in the position to make such decisions in a company objectively consider the technology on it's own merits.
and access speed is not an issue.
I never said anything about access speed... I was talking about transfer rates, A 20GB storage tape with a transfer rate of 340megabits per second in some of the latest incarnations is going to be faster than transferring 20GB over even the fastest internet connection, unless both machines are in a Gigabit LAN. If it takes overnight, as it often does, to make a backup of a system, it's going to take a lot longer to either backup/archive to or from a on-line storage location. And problems during the transfer either way could cause various problems.
Another thing I feel, particularly with SaaS and cloud storage, is that it could present an issue for companies is that it increases the surface of attack. Software companies are semi-constantly having their source code leaked from their own internal networks from enterprising hackers. Backups stored on a cloud service will have that data as well. Naturally, there will be passwords, authentication, maybe encryption involved, but the fact is that so too did many of the companies that had their Intellectual Property leaked or stolen, so this increases the surface area; Hackers could get access to the private data either by doing what they have always been able to do and find some on-site wireless access point that some employee installed temporarily and forgot to remove; take advantage of a modem forgotten in some backroom somewhere that was installed to help Joe In accounting get access to his client's documents using a VPS- now they can apply the same things to the cloud service and get a lot of company data if they succeed.
Basically, putting that sort of data- regardless of how many authentications and encryptions and hashes are placed as barriers- is a fundamental issue because a target you can see is a lot easier to hack into than one that is only maintained on the internal network of the firewall. This also opens up such companies to man in the middle attacks that can place themselves between the company and the on-line service and reconstruct the data. So cloud storage services, particularly those utilized by businesses, had better have good security certificates. Usually a company keeps it's confidential data- such as the source code to any software they sell, business documents, marketing planning, etc on an internal network that is not internet exposed. Introducing On-line storage in this computing means that the data will in fact be internet facing by design. There is a good reason that Company intranets are isolated as much as possible from the internet at large.
Anyway, that's what we're seeing and clearly it's a direction in which every company in the Fortune 1000 and well beyond are headed.
The problem is assuming that the direction that every company is going is inherently a good idea or is going to end well. How many other fads and hypes in computing have come and gone? FORTH was supposed to usher in a new era of computing in the 80's. Did it? No. What happened to the companies that embraced it? They struggled and/or failed. Java was similarly poised to usher in some new world of computing in the 90's. It didn't. XML was supposed to completely change how programs communicate with one another. It didn't. And what happened to many of the companies that based their entire existence on it's success? they either died or went into other markets. FORTH still exists, but it has found usages that aren't as gradiose as the original claims. Java still exists, but it has found quite a bit of usage that is nowhere near as grandiose as the original claims, too. XML found usage in the original concept of AJAX, and has managed to establish itself- though not quite as much as it was predicted- as a sort of data format. But even Forth, Java, and XML are exceptions; for each of these, there are four that failed in similar time frames. (ADA was similarly poised to be the thing that everybody needed to understand around 25 years ago). These are programming languages, but the fact is that the same trend is easily seen elsewhere. Each one has advantages that are hyped, but they all have disadvantages that contributed to the reality. Nobody can predict the future no matter how many economics degrees they have and when you have people stating things like "the cloud is the future" you almost get the sense that they hope if they repeat it enough times it's bound to come true. I'm not saying it won't be prevalent, but I'm saying that to assume it will- and bank a lot on it- is foolhardy at best. Technologies need to develop and mature before being universally adopted, especially in domains where it wasn't a use-case as a experimental tech.
At this point there are so many cloud-based services it would be difficult to tease apart those offering services on the cloud that are actually designed with the cloud in mind and those that offer a cloud-based service that has all the disadvantages of the cloud without exploiting any of the advantages. This is not unlike the situation with Java, where companies scrambled to come out with something- anything, written in Java, which was usually rushed and far more unusable than anything else they offered.
I'm not disagreeing, of course, that there are advantages. But it will take a few years for the companies offering this service to mature, and for those just doing it to make their service trendy (example: BMC's SaaS "Remedy" software) to fail and move on to other things. This is due to some of them simply misunderstanding the tech and expecting it to be a silver bullet of some sort.
Most companies want to follow these things, because they assume "hey, everybody else is doing it so it must be a good idea". When hype overtakes a person's ability to appraise a technology objectively, then it's going to start to be used for the wrong things.
The reason for this comes right down to the fact that many such companies have people in positions of power who, despite knowing absolutely nothing about technology, are able to form strong opinions about it. Let's say there is a company in 1996. I use 1996 because in hindsight many of the predictions promised are patently absurd-. Now, SuperCo. is starting on the latest version of their award winning Word Processing software, which is currently written in C. One of the higher ups- who assumes that all technologies have equal capability, decrees the new version shall be rewritten in Java. Why does he think this? What he was thinking is probably something like this- "Java is a standard, I know it must be, since I read about it in the press all the time, and since it's a standard, I can't get in trouble for choosing it, and it also means that there are lots of people with experience in Java, so if somebody working for me quits, as programmers working for me mysteriously do, I can replace them easily".
It's not unreasonable, but it is based on an assertion that the technologies being weighed- in this particular case- are equivalent. They are not. IT technologies differ in capability and have various trade offs. (SaaS is universally cheaper, but you can't really know if the company responsible did it competently or if they are basically just providing you a Remote connection to the program being run normally on some server somewhere in kiosk mode). If they were all equivalent, he'd be right on target. Just use whatever everybody else is using. But they aren't equivalent.
The actual needs have to be weighed against the technologies capabilities. In the case of SaaS, if you are going to do it, do it properly. Don't just wrap your client application in some thin silverlight wrapper. Companies wanting to buy Software as a Service need to weigh the pros and cons heavily. Sure, SaaS is cheaper, but things like Office 365 still offer disadvantages; for example, One example being that it is tightly integrated, meaning one can only get the full advantages of the software if they rely on windows or windows-based products. Outlook(client) works with Lync Mail Servers without issues, but there are several disadvantages if using the 365 version of the software. Whether these affect a company depends on the company. A company that uses Windows-based servers will probably be fine, but many companies use sendmail-based mail exchange SMTP servers or things like LYNC server which won't work as smoothly with Outlook in Office 365. The integration of course is a advantage in many cases too, but it depends on each individual companies scenario. Painting the Cloud as a solution to something- anything- without knowing the details of the scenario would be irresponsible, but that is what many tech pundits are doing with Cloud-based solutions.
For PaaS, you stated yourself essentially that there are only specific use cases for it; those with differing processing requirements throughout the year. Obviously, if a company doesn't fit that use case, they will get all the disadvantages of PaaS- such as Customer captivity and a lower confidence in data security- but without any of the advantages to their business model.