Cloud computing is the hot topic wherever we turn these days . I picked up a few magazines to read in my flight back home from Dresden few days ago, and had an overdose of cloud computing in the next eight hours. Also, my employer is a big player in cloud computing , and hence I get a healthy dose of it via internal communication channels too. I also had some twitter exchanges on the topic. So at this point - I wanted to share my thoughts on this topic with fellow SCNers on how I see this playing out.
I don't see a lot of customers rushing to move their mission critical production systems to cloud - even to private clouds within their firewall. It will probably take a few early adapters to pave the way before there is a mass migration. This is especially true for systems that handle sensitive data. Even if customers want to do this - it is complex to get around the legal hurdles. Some countries, especially in EU, have hard and fast laws on where the servers can reside. Considering the fact that there are bigger problems in the world like a crashing economy, unemployment and so on - I doubt if law makers will jump into reworking these laws any time soon. However, not everythng in production systems is sensitive. And it might make sense to move them to a cloud - even to a public one. So most probably a hybrid of public and private clouds is what most customers will end up with.
I wonder if SAP will have solutions to keep sensitive data separated from the rest of the data . For example - a customer's SSN, bank account and email id maybe considered private info, that should be in a server located where it is legally approved. But the rest of the data (orders, deliveries etc) - which is probably the much bigger chunk - could be held anywhere else. So if SAP has a solution to keep sensitive data and the rest logically connected, and physically separated - maybe the legal hurdles might not pose such a big trouble after all. I might be over-simplifying this, but I think some such paradigm needs to evolve for cloud computing to really catch on.
However, production systems are only one part of the puzzle - there is still considerable value in moving Development and QA systems to cloud. The ability to quickly deploy a development or QA system for new projects is something I have encountered several times in my consulting career - and this could quickly change for the better in the world of clouds. However, this needs the ability for data centers to scale up and down "on demand". This scaling part is not easy, especially in private clouds. Data center vendors will probably be happy to scale up, but not quite happy to scale down capacity. There might be "lock in" periods before a scale down can happen. There are other challenges too when development and QA are on cloud and Production is not - like refreshing QA from production, or moving transports between these two systems, as well as backup and recovery. I also don't know if solution manager can handle such a landscape with one foot in cloud and one foot outside. I am not at all a big expert in this area, and have pinged a few colleagues from that area to understand their thought on this. I will post a followup when I get some undertanding of that.
I am especially intrigued on whether it makes sense to keep backups on the cloud or not. From a disaster recovery perspective, it might be an excellent idea to do this. However, from a security perspective - this poses similar risks as puting the production data on the cloud. So I wonder what the right answe would be.
Not all computing have the same hardware requirements - for example, high performance number crunching required in analytics applications need something different from a high volume social networking site, and a single customer might need them both. I believe there are sophisticated optimization tools that can reduce this pain to a good extent, but end of the day - you still need to sclae two different hardware platforms. I am trying to find out how hardware companies are approaching this challenge.
I think a lot of customers might want to move their desktops to cloud in near future. This is a big pain in today's world to manage desktops - to keep on top of licenses, upgrades and all that. Also, this would mean - the client side machines could be really dumb (as in "less powerful"), and you could run old PCs for much longer. This will need reliable networks / internet connection. And I assume that some non-cloud desktop will exist to serve as back up option if networks are down. It will be interesting to see how computer manufacturers will respond if such a trend gains momentum. I think SAP is well poised to take advantage if cloud desktops become common place, especially due to their determined efforts to unify UIs across the suite.
How does this affect SAP consultants? I think new specializations will develop in the consultant and SI markets who specialize in designing and developing systems on the cloud. And if cloud catches on, then client facing consultants would certainly need enough knowledge to factor this into their design and decision making. SIs will probably have specialized offerings to maintain and enhance private clouds for customers. Most applications in the market today already have some form of roadmap to adapt to the cloud - but I think one of the earliest adapters amongst these will be BPM. If processes become cloud based, then BPM by extension should be one of the first to adapt to cloud based paradigm. And due to the big concern on data privacy etc, security consultants might again have the wind on their backs - like the time when SOX came into being.
A big criticism I have seen on data centers is on the "green" front, that their cooling systems etc cause a lot of harm to the environment. Many companies, including SAP, have announced dedicated initiatives to cut it down drastically. If these companies, keep their promises - then it might actually make a case for cloud computing to expand its foot print ( assuming optimizing a large data center owned by a company with a commited sustainability strategy is more plausible than trying to optimize several small data centers or individual servers).
