Wednesday, October 17, 2012

Good cloud, bad cloud for Africa



Recently I attended a cloud computing summit in Nairobi, Kenya. The theme was "Re-imagining the evolution of ICT technologies in East Africa”. While attendance to the conference was limited, it was interesting to listen to the cloud messages from the global IT vendors: CISCO, FUJITSU, MICROSOFT, etc. I'm always amazed to see how those global companies deliver exactly the same message in Africa than they deliver in the developed world, without any consideration for the differences in the markets.
But before I can share my opinion, I need to explain the difference between public and private cloud for those readers not familiar with the subject. It is mainly based on a difference in the implementation and business model.

Public clouds are based on large data centers owned by cloud service providers (CSP). They "rent out" IT resources on a "pay-as-you-go” or "pay-per-use" basis to the public. That public is basically anyone who has a credit card and can pay for the service. A good example of a public cloud is Amazon Web Services (AWS).
The key point with "public clouds" is that the cloud IT resources are shared by all the customers. You can have applications from different customers running on the same server with the appropriate security to prevent them to see each other's application. This is achieved by using a technology called "virtualization," which enables applications to be moved freely from server to server, and to run different applications on the same server.
It is this sharing of IT resources between a large number of customers that allows CSPs to offer their services for such a low rental price. Typically computer server utilization in a cloud runs at 90% of capacity, as opposed to the 10-20% of capacity typically seen in a traditional enterprise installation.  Private clouds, on the other hand, are reserved for a single enterprise. Their IT resources are dedicated to that enterprise, and no other enterprise or customer may use them. The reason it is still described as a cloud is because the enterprise itself uses the same virtualization technology as the CSPs to optimize utilization of their servers. The goal of a private cloud is primarily to reduce the cost of the IT resources (by reducing the number of servers required), along with a lower cost of management of the IT infrastructure through automation allowing users to allocate IT resources in a self-service mode. While the enterprise may also use the "Pay-per-use" business model through an internal billing system, that is generally not the primary reason for an enterprise adopting cloud technology.

The question then becomes: how does an enterprise choose between a public or a private cloud?


When the scale (mainly the number of servers units) of an IT infrastructure increases, the management cost per unit decreases. For a given scale (A), we have seen that the unit cost of a private cloud is lower than for a traditional infrastructure. We can then easily compare that cost with the unit cost of a public cloud, which is simply the rental price charged by the public cloud CSP for one unit of IT. The public cloud unit cost is lower, since they benefit from the economies of scale offered by their large IT infrastructure.  B is the scale at which the unit cost of a private cloud becomes lower than the unit cost of a public cloud. So if we just look at the cost factor, one enterprise should decide for a public cloud if their IT infrastructure scale is smaller than B.
Now let me go back to that summit and explain why I think the global vendors are selling the bad cloud to Africa with no shame about doing so!

Africa is a continent of SMEs, with only 20 enterprises in Africa making more than $3B/year (McKinsey Global Institute report 2010: Lions on the Move). That means the majority of enterprises in Africa have a small IT infrastructure (when they have one at all!), significantly below “Scale B” (above). The only viable solution for most enterprises in Africa, then, is the public cloud. Unfortunately there are very few public clouds in Africa. A few smaller CSPs are offering public clouds, mainly in South Africa, but none of the major CSPs have plans to deploy a cloud data center in Africa any time soon. The main reason is probably that the market is not large enough and not yet ready from the CSPs' perspective.  The Middle East-Africa market opportunity for infrastructure as a service (IaaS) (Gartner, Public Cloud Services, Worldwide and Regions, Industry Sectors, 2009 – 2014, June 2010) is unchanged from 2010-2014 at $100M/year which probably represents about $30M for Africa!!

The global IT vendors at the cloud summit were selling the private cloud to African enterprises using fallacious arguments to justify their private cloud offerings: "It is no longer a question of whether enterprises will use cloud computing - they already are". But that is in the West, not in Africa! As I have shown, this private cloud solution is inappropriate for African enterprises, because their IT infrastructure does not justify it. In fact, in Africa “Scale B” is probably larger than in the developed world because of the lower labor cost. IT labor being cheaper, IT teams can manage larger infrastructure for the same cost, pushing the limit B where a private cloud becomes justified even further away. But more importantly, the IT expertise to manage such complex technology is not available. That doesn't seem to bother those vendors, since they will be glad to offer those expert services in addition to the infrastructure even if that solution is inappropriate for most African enterprises.

The hope for Africa lies in two factors. First, the rapidly decreasing cost of broadband internet and the increasing speed of networks enables African SMEs to run their applications in public clouds located in the US or Europe.  Many African SMEs are doing this already today. As long as your application’s transactional data transfer is small (e.g. for an accounting application), this is quite possible.
The second factor is the emergence of smaller CSPs, who are willing to invest in cloud data centers in Africa. They have probably come to understand that the cloud financial model in Africa is different from that in the developed world, where all enterprises have invested in IT infrastructure along with the associated IT skills investment. A significant cost reduction is required for a developed world enterprise to justify the cost of moving to a public cloud. So CSPs in the West are building mega data centers in order to offer the lowest possible cost and convince enterprises to place their IT infrastructure in a public cloud. In Africa, the situation is different. Most SMEs have no IT infrastructure or a very limited investment in IT. For those enterprises, a public cloud is the only way to get affordable access to IT. So any CSP in Africa who can provide a good cloud service, even if it comes at a higher rental cost than their peers in the developed world, will be offering a good value to African SMEs, because those potential clients will be comparing the investment needed to acquire an IT infrastructure (including the skills acquisition) with the cost of "renting" an IT infrastructure from a public cloud.
Consequently, the African CSPs don't need to make major investments in mega data centers. In fact, a small data center will be able to offer acceptable renting prices. Furthermore, because most SMEs at this stage lack existing IT, they don't need much power or storage to begin with. So CSPs are in a position to serve more customers with the same infrastructure than they could in the developed world, where customers require more power and storage for their legacy IT needs.

African entrepreneurs should take this opportunity to offer public cloud services for SMEs, since the public cloud market has been abandoned by the major CSPs. Do so before the big players realize their mistake. African innovators can understand African needs much better than the Silicon Valley companies. Africa, this is your opportunity to take!

This posting was also published in NextBillion