Has an IT professional in your organization told you that Software-as-a-Service (SaaS) just does not work for your enterprise? I am going to help their argument; then I’ll tear it down.
The biggest problem for a professional engineering organization with SaaS is latency. Also called disk-access-time , data transfer rate, bit rate. Rapid communication between processing units and data storage pools is critical for today’s professional computing efforts. If you don’t believe me, try opening (and editing) a large GIS or CAD file via a VPN from a network storage device operating in a satellite location from across the country.
For example – Our realized transfer rates from our facility in Portland to Amazon Web Services (AWS) in Virginia is 3.5 megabits per second (mb/s). We have a 100 megabit per second dedicated service pipe to our facility. This latency is not Amazon’s fault, but rather it is a function of the speed of light and the number of Internet “hops” or network transfer points between WeoGeo and AWS. From Florida, we average about 7 mb/s to the same facility. This issue is one of the fundamental reasons AWS released their Import/Export feature to S3.
To put this into perspective, in a typical disk drive on a desktop computer (circa 2008) operating at 7200 rpm operates at a disk access transfer speed of 560 – 2400 mb/s (70 – 300 megabytes per second). That is 2 – 3 orders of magnitude faster than accessing the same file via the Internet. The last time desktop computer users “suffered” thru 5 mb/s disk access transfer speeds was in 1987 when IBM released the PS/2 with a 5 MegaByte ST-506 Seagate Technologies hard drive.
When will latency be reduced to the point where we might consider “the internet is the computer”? That is kind of hard to say. Even with ubiquitous broadband access at >100 mb/s to all business, we will stuff suffer the “speed of light” problems – photons can only go so fast. In addition, every time you have to go through an Internet junction or telecommunication switch (i.e. hops) you will increase the packet transfer times. A report from the National Broadband Coalition (also covered here) suggests the pipe speeds to small and medium business will not approach those of hard disk drives until sometime between 2015 and 2020 (see table below).
This suggests that the internet-as-the-computer to replace your desktop is still some time away, maybe as long as 3 – 5 years. This latency argument is what many in corporate IT departments would use to strike against SaaS, PaaS, or IaaS services within your organizations. (I am purposely ignoring security, but will address this issue in another post). I would counter that those arguments are very similar to the ones once used against the IBM PS/2 as a corporate workhorse back-in-the-day; and ultimately I believe they are rooted more in bureaucratic inertia than true cost/benefit analysis to the enterprise.
Internet-based services have a place in today’s enterprise environment. Depending on the use case, they can be more efficient in managing your limited IT dollars. These services can also provide greater, timelier, software support, which together with the lower costs, increases the bottom-line productivity of your organization. More importantly, these managed, off-site, services will have a greater place in your organization tomorrow. Data transfer speeds will increase to a point that the latency issues will be negligible for the services you require. As a decision maker in your organization – are you planning for greater productivity and enhanced profits tomorrow, or adding to the “mainframe” infrastructure of today?
Several major imaging vendors send a chunk of their current image processing to low cost countries like China and India to complete their large-scale projects. It seems that there might be a better way to accomplish such geo-processing tasks that still require eyes then to incur the time and expense of sending these tasks overseas. Perhaps Mturk and some smart programming might offer a different approach.
