The ICT Optimist

For as long as I have been in the ICT industry, there always seems to be a technology buzzword that dominates the dialog but no one really agrees on the actual definition of the term, the details behind it or the way to measure the market.

In some cases these issues become resolved and the technology becomes mainstream and understandable. An example of this is the term “switch” when used to describe a network device that passes packets over multiple ports. If you remember when the term showed up over a decade ago, some companies praised the value of switching as a cut-through device versus a store and forward device (meaning the packets spent less time inside the devices buffers and ultimately had lower latency). The issue was that then every multi-port bridge as defined by IEEE 802.1 standards was renamed a “switch”. Soon anything with more than one port on it was a “switch” including things formerly known as routers, bridges, gateways,…  Over time the technical purists gave up and the industry just started calling multi-port bridges, routers and other devices layer 2, 3,or 4-7 switches. All that posturing and debate resulted in nothing more than a new name that honestly didn’t matter at all.

In other situations it just fades away as the next big buzzword appears and distracts the industry from the prior dialog. A good but forgotten example of this was the short lived infatuation of Asynchronous Transfer Mode (ATM) technology and other deterministic LAN technologies. The hype was around the idea that if we could make networks predictable, voice, video, and data could all work in a converged way over the same network. Obviously if you look at the fact that today, these types of traffic do work over the same network but not because we achieved deterministic networking but rather because we threw huge amounts of bandwidth at low cost at the problem and then layered on top a “good enough” class of service model. The hype of perfectly engineered networking was lost by the availability of cheap, pervasive capacity with some band aid type control mechanisms.

So here we are with another new buzzword generating a huge amount of hype… cloud computing. Also with cloud computing is a host of new acronyms describing “everything as a service”, XaaS, IaaS, DaaS, SaaS, PaaS…. These are buzzwords because we don’t have a clear definition of them. On sites like Wikipedia, there is a pretty detailed summary of all the possible elements and models that could be considered cloud computing, for example,  but all one has to do is read the literature from a few players in this area to see that the language and architecture of this space is far from uniform. For a pretty good dialog on this set of new “thing” and how vague they are, the register.co.uk hosted a pretty long webcast. Its done pretty well but the ambiguity this space has is obvious from the dialog. The question though is if cloud computing is in fact a sustainable and real paradigm shift or just simply a better marketing term for a set of different approaches to IT and compute architectures.

In terms of new and unique properties, I find that the basic elements are not new. Most definitions of cloud computing boil down to using :

1. A distributed computing environment – meaning many CPUs are used as a system with a network as the interconnect. The issue is that sometimes this is a classic grid computing model and in other cases it is a much more ad hock collection of interconnected computing resources.
2. A level of virtualization – meaning that the processes have little knowledge of the actual compute hardware they run on. Technology such as VMWare is often used for this but by no means is there a uniform architecture for virtualization yet and the degree of virtualization varies widely from one definition and system to the next
3. A set of services that can be accessed by software running in the cloud – meaning that many of the functions that one would normally have to implement in a discrete system are offered as “services” to any code running in the cloud. For example real time communications and conferencing could be offered to apps in a cloud as a function. Other traditional discrete services such as identity management or even complex functions such as location functions and tracking of client devices could make up some of that broad services suite. The issue here is that what services a cloud delivers is left to the particular specific cloud so an application implemented in cloud A might not have access to the same “services” on cloud B.
4. A web services or other IT friendly programming interface -  meaning that instead of low level compute centric interfaces, functions in this type of system are presented as programming interfaces that are native to the IT applications using the cloud. For example instead of making multiple low level calls to invoke a conference call, a programmer in this type of environment would simply execute a basic function such as “invokecall(phone number)” and the call would happen. All the complexity buried in the cloud on the system that specialized in real time communication.

The problem is that in all 4 of these elements, they may or may not be implemented in a system called a cloud.They may also be implemented in radically different ways and at different levels of sophistication. And in many cases they may not even be present. Yet in all of these permutations the system is called a “cloud computing” system or solution. Sounds like hype and buzzwords to me.

Does this mean cloud computing is hype and irrelevant? Not at all, the idea of pulling a distributed network based compute infrastructure together with good vitalization, embedded services and an IT friendly interface to applications is in fact pretty new. It is also very useful. I have spoken to dozens of small companies in the last six months that are able to achieve more rapid development of solutions and greater functionality because they can host their solutions fully or partially in one of many cloud computing systems out there. They also are able to focus on their core value as the cloud computing systems provide them a set of embedded services that if they had to create and implement internally would be far to costly and complex. All of that means that this idea of cloud based computing is of value, the issue is if it is a singular thing or simply a progression of a number of IT systems that happen to provide good synergy today. I am not sure that cloud based computing as a “new thing” is sustainable or even relevant but the sub elements and the ability to use them together are a good step in the right direction. What inevitably will happen is that more of the elements of the IT ecosystem will evolve and become available to a larger base of customers and additional improvements on other portions of the total IT framework will emerge to provide incremental value beyond the four elements mentioned in cloud computing today. Will that continue to be known as cloud computing or will it simply become part of the way we do IT in the 21st century. I am betting on the latter and with that betting that the term and hype over cloud computing will simply disappear under the broad evolution of all parts of the ICT ecosystem. In any event it will be interesting to see the evolution and how disruptive it is to our industry.

If you want to read a recent article on the pros and cons of cloud computing, Forbes.com just published a decent article that I think sums up many of the issues often overlooked and the reality of this hype not being something new.

Its amazing that even when you take some time off for vacation, the macro level industry changes continue … you would hope that the industry would take time off and enjoy the nice spring weather …. just kidding <grin>

One of the most interesting events this week was the announcement at Interop of a broad alliance between Microsoft and HP to better compete with (or react to) Cisco. On GigaOM.com there is a pretty good comment stream summarizing the announcement and dialog at the keynote. The punch line is that as Cisco continues to move up the IT stack and focus on compute and collaboration technologies and solutions, the rest of the industry is reacting by forming alliances to better match the Cisco offerings.

Now we have the following alliances and large scale transactions just in the last few weeks….

IBM and Brocade

HP and Microsoft

Oracle buying Sun

Any one of these is big, but all of these huge companies shifting relationships and creating new industry structure means that there will be huge disruption and opportunity. The disruption is seen in the fact that almost every one of these big industry changes is at the expense of other prior relationships or even competitive positions. Additionally, whenever a new set of partnerships emerge they inevitably expect their customers to change with them even if the existing or prior offerings are working fine (note to customers -  be ready for a new round of sales calls).

The opportunity is that whenever a new ecosystem forms the large players inevitably overstate their capabilities and once they realize that they have gaps, they begin a process of finding the critical technology and solutions to complete their new image of the IT/Telecom converged space. It remains to be seen what gaps will be most critical in these new ecosystems but I can assure you that there will be lots. As an optimist I am excited by this as the best time for companies to emerge and new technologies to become valuable is when a macro level vision is incomplete without them.

Lets see if we have even more industry composition changes in the coming months. I would guess there will be and from that and the existing changes we should see repercussions throughout the vendor, channel and customer base. That will be exciting in any event.

Recently there has been a bit more dialog about the fact that Wireless networks (802.11 based Wifi specifically) are moving to a position of prominence in the enterprise market. I follow this with great interest because in my career dealing with Enterprise CIOs two of the most often asked questions are “when will I be able to rely fully on wireless access?” and “when will I no longer invest in broad cabling of my buildings?” The reason CIOs have been asking me this for the past 10 years is that when you take a hard look at user patterns on the edge of an enterprise (home, small office, hotel, airport, or even campus LAN) it is clear that with every year the use of Wifi as the primary access technology has increased. What is interesting is that this was not be design. In fact most enterprises offered Wifi as an overlay network that was optionally available but as soon as the workforce adopted laptops as their primary compute hardware the WiFi network became the primary network of the user base.  Additionally the workforce has become accustomed to WiFi in their home and when they travel and since most people conduct their work in more places than the fixed office, the users figured out,well before the IT groups did,  that you can indeed use wireless for business.

In a recent Network World article titled “Is it time to cut the Ethernet access cable?”, the topic was discussed and what was interesting was that the vast majority of end customers acknowledged that wireless is playing a much more primary role in their networks. Not surprisingly, the vendors where split. The wireless switching vendors suggested a wireless only experience while Cisco, the largest provider of wired LANs, asserted that you actually need wired, wireless and a cellular overlay. It was fairly interesting to note in this article that many of the customers interviewed wanted to remain anonymous when asked about wireless use, as if there was a shame or career risk to acknowledging that their business used wireless technology.

So when will we be able to unwire our enterprises? Well that is still a hard thing to predict. Clearly with advanced such as 802.11N, WiFi is a much more robust technology. There are still challenges in relying on Wifi that are not present in wired LANs. Specifically since WiFi operates in unlicensed frequency bands, the possibility of interference or jamming is a real risk. The reality though is that given the 2.4 and 5Ghz bands and the significant enhancements most Wireless systems have today, interference is a far less likely issue. Consider, if you rely on WiFi at home, how often the wireless network failed versus how often the broadband connection (DSL, Cable…) failed. In giving it some though I cant remember when a wireless issue stopped me from connecting but disruptions on the broadband link are a common event. In your experience in the office, consider how often your WiFi service stopped? Probably not very often and clearly more resilient than many of your business applications.

An additional concern that has haunted Wireless is that it is a “shared network”. Because the bandwidth is shared between all users on the access point, it is possible that an active user might occupy enough bandwidth to starve other users. One could deploy complex QoS in the wireless domain to try to solve this but the reality is that with 802.11N the amount of capacity on the “shared” wireless network is probably far greater than the capacity of the “shared” WAN link that 99% of your traffic is traversing. I just looked at my home network and my .11N networks are providing about 144Meg of capacity and my Cable link is 15meg downstream and 1 Meg upstream using a test suite on DSLReports.com . The bigger challenge today is to assure fairness on the broadband up-link (I sit on the board of company that does this called Smart Share Systems) as that is the actual congestion point in most cases.

In any event enterprises are still suspect of a fully unwired experience. They may not have specific examples of wireless failure but they do know that there is uncertainty when you move data over a shared wireless medium leveraging unlicensed spectrum. Another Network World article this week titled “Wireless Netowrks must overcome Interfearence , Latency and Security Challenges” lists the current concerns but on a closer read, it shows how broadly Wifi is currently used and in almost all of the examples used successfully today.  The reality though is that we have enough data to know that if you model your existing user patterns almost all enterprises will see that there has been significant shift towards using wireless and a dramatic reduction in the use of the wire-line LAN. A realistic approach would be to redistribute resources and focus to the new usage distribution and spend more time engineering, improving and enhancing the network people are moving to (Wireless) and looking for ways to decommission or reduce the footprint of the network they are moving away from (Wired access). How much you wish to shift is up to you but a good exercise would be to audit the current usage and ask if you are spending your capital and time on where your users are going towards or if you are simply perpetuating a network that is dramatically under utilized. If every year you ask and answer the usage distribution question and map your IT spend and resources to that trend, you will be aligned to the reality of your user communities true dependencies. As an added bonus you IT bill for access will most likely drop as you leverage more Wireless as it is common knowledge that the cost of a wireless network is far lower than wired and enables far better economics in terms of user productivity and Moves Adds Changes cost to IT.