Tech Lover 9

It’s obvious that one of the primary reason enterprises are reluctant to move resources to the cloud, and cloud computing, is the inherent bottlenecks and unreliability of the Internet – as the primary networking mode service delivery channel. The Internet possesses several critical problems that expose its weaknesses in providing adequate network performance and quality levels. These include:

• Peering point problems. Peering points separate Internet networks for the purpose of exchanging traffic between customers of each network. Peering is normally a settlement-free “swapping” (exchange) of data from one network provider to the other – forming the backbone of the Internet. Unfortunately, perhaps largely due to the settlement-free nature, there has been lack of economical incentives for network operators in sufficiently upgrading their peering routers to handle increasing network traffic. (Depeering is another process that I will not go into). While the “first mile” (from data center to the Internet/WAN) and the “last mile” (from end-user to the Internet/WAN) have received significant investments, the “middle mile” (i.e. the Internet) has lagged behind.

• Border Gateway Protocol (BGP), the Internet’s inter-network routing protocol determines how data packets travel from one network to another in the cloud. The problem with BGP lies in its slow routing capabilities, especially when making fine distinctions between the traffic on multiple routes. This results in lack of performance and reliability of service delivery in the cloud.

• TCP bottlenecks. TCP is the Internet’s primary communication protocol – providing reliable, ordered delivery of a stream of bytes from a program on one computer to another program on another computer. However, TCP was designed for reliability rather than efficiency and often causes drag (delay) in packet delivery due to its multiple round-trips to set up and tear down connections. This is a overhead that can be especially detrimental to the performance of SaaS and PaaS-based enterprise applications.

So, if these are the main concerns for good network performance and reliable service delivery on the Internet, what can be done to mitigate these? Well, there are probably several different measures that can be taken, and depends on both the cloud service provider, network provider(s) and the enterprise using the cloud computing services and include:

• Using a overlay network, e.g. from CDN providers like Akamai or Limelight Networks. These providers use edge servers and multiple (proprietary) optimization schemes to improve network performance and quality levels of content delivery.

• Use managed IP network services. Many network providers offer private IP connections, e.g. from a particular data center (cloud provider) to an enterprise, promising quality of service and much better performances than over the public Internet. Today, the MPLS (Multi-Protocol-Layer-Switching) data-carrying mechanism is becoming prevalent in modern networks for high-performance delivery.

• Consider the new FASP (Fast and Secure Protocol), originally designed by Aspera, for improving efficiency and throughput. Amazon AWS plan to offer FASP in its services.

Having said this, it becomes obvious that companies do not have to rely completely on the Internet for service delivery if they aspire to use public cloud computing services. There are several ways to improve network performances and reliability as I have pointed out – and there are probably even more ways available!