Bottlenecks, bottlenecks, bottlenecks!

Let me start this post by saying that as a technology professional, I hate bottlenecks.

In saying as much, I’m probably preaching to the choir; almost ALL of us struggle with bottlenecks in our daily lives.  Traffic jams are usually caused by a bottleneck – the reduction of lanes from 4 to 2, for instance – that results in more traffic per available lane.  Since two cars cannot peacefully occupy the same space on the road, a reduction in lanes during a heavy traffic period means that fewer cars can move through that obstructed zone at a time, resulting in a backup as faster-moving traffic piles up behind the slower-moving obstructed zone.  

Not only does the overall traffic flow tend to slow down due to bottlenecks, but the traffic flow tends to be more turbulent as well, often resulting in collisions.  In order to illustrate this point, I’ll turn to nature.  Consider a river: without obstruction or an increase to the volume of water in the river, it flows gently and calmly within its banks.  The character of the river can change dramatically, however, with the introduction of obstacles, such as boulders, that restrict and impede the flow of water, the narrowing of the river valley, or an increase in the volume of water flowing through the river.  As any of these factors are introduced, turbulence increases, and while water may rush through these restricted zones, even more water is often backed up behind the obstruction.

Technology works much the same way, whether you’re talking about a single computer or a data network.  My computer could have the fastest processor in the world, virtually unlimited memory, and cutting edge graphics components, but if I’m using a 4,200 RPM conventional hard drive, it will still take a long time to open Word.  At my house, I could subscribe to a 1Gb per second internet service, but will only achieve a fraction of that while browsing the internet on my iPad if my wireless router is only capable of 11Mb per second.

Within an enterprise network such as the one that the Iowa City Schools rely upon, the number of potential bottlenecks is staggering.  Consider, for instance, what happens if I use Google to search for a picture of Siberian Husky puppies from a laptop at Coralville Central Elementary School (note: this is a bit simplified):

1. I click “Go” to search for pictures of Siberian Husky puppies
2. My web browser sends my search request through my computer’s wireless adapter
3. The data from my request is transmitted through a wireless access point at Coralville Central
4. From there, the data travels to a switch in a network closet at Coralville Central
5. From there, it’s routed to the core switch at Coralville Central
6. Then, it travels over fiber optic cable to the core switch at North West Junior High
7. After being routed through fiber optics equipment to the core switch at NWJH, my data is sent via fiber optic cable to the ICCSD Core Data Center
8. At the data center, the request passes through a switch and is sent to the district’s wireless controller
9. From there, it’s sent to another switch
10. At this point, it’s sent to the district’s firewall
11. From the firewall, the request goes to our internet circuit to the Iowa Communications Network (ICN)
12. Now it travels to an ICN data center via fiber optic cable
13. From the ICN, the request is sent via the internet (broadly speaking) to a server at Google 
14. Google servers process the request, and then send results via a combination of data cached by Google and requests forwarded on to the websites delivering results

At this point, the entire process is basically repeated, in reverse.  When you consider that any of the bolded items are potential points of failure, and also represent potential bottlenecks.  A flood of traffic from (or to) Coralville Central, for instance, could overwhelm the switches in the building or the 1Gb fiber connection to NWJH.  A large number of devices accessing wireless could overwhelm the bandwidth available to the wireless controller, resulting in slowdowns or dropped connections.

In networking, as with traffic and rivers, bottlenecks lead to slowness, turbulence, and potential collisions.  If I send 100Gb of data through a 1Gb connection, much of the traffic will have to queue (wait in line) behind the bottleneck.  While this is occurring, network collisions can destroy some of those data packets (this is one of the reasons  – along with timeouts – why, when a network is slow, communications often fail to go through at all).

How are we addressing these issues? 

Dealing with points of failure and potential bottlenecks is always critical, but rarely more so than in advance of a major influx of new devices, such as we’ll see with our upcoming secondary 1:1 device program.  Currently, the district’s fiber optic networks have the ICCSD Educational Service Center (ESC) as the hub, which is directly connected to core switches at North Central Junior High, North West Junior High, South East Junior High, and the physical plant.  From each of those locations, additional buildings are served; in the example to the right, for instance, City High, Hoover Elementary, Twain Elementary, Tate High, and Grant Wood Elementary are served by fiber from South East Junior High.

Each of these connections between buildings is 1Gb, including the four core connections back to the ESC.  During normal usage, these connections frequently see usage rates of 50%; during heavy periods, they can easily be saturated, at which point the fiber becomes a bottleneck, slowing traffic and disrupting communications.  Further, by not having more bandwidth between buildings, the district has to maintain software and update distribution servers throughout the district, and we frequently must put off or stagger needed patch deployments so as not to overwhelm our 1Gb fiber connections.

One of the projects that we’ll complete this winter is an upgrade of our fiber network – and the switching equipment that supports it – to a 40Gb standard, which will increase our capacity by 4000%, and also make it possible for the district to centralize management, software deployment, and imaging solutions without concern about compromising the performance of our data network.

Along similar lines, we’re pursuing upgrades to our internet bandwidth and firewall in order to support more throughput than our current 1Gb.  Based upon our student enrollment of just over 14,000 students, minimum standards recommended by the State Educational Technology Directors Association would dictate that we should provide at least 1.4Gb internet bandwidth, with a plan to increase that offering to 14Gb by 2020.

Since the vast majority of our devices connect to the network wirelessly, we are also pursuing upgrades to our wireless network to support the speed and density requirements of our current and upcoming technology deployments, including offering 802.11ac service throughout our buildings, and support for a minimum of 60 concurrent devices within each of our classrooms.

All of these upgrades are being undertaken with the ultimate goal of providing a reliable, high-performance network that can serve the educational needs of our students and teachers, while also providing for the operational needs of the district.  An increasing amount of the work that we do is done online, through our PowerSchool student information system, Google Drive for content creation, content creation and delivery with YouTube, and online curricular resources such as those offered by Pearson, to name a few.  As we progress towards delivery of virtualized applications, a 1:1 environment, an evolution towards paperless schools, and other initiatives currently underway, the pressures on our network – and the need for that network to be robust and dependable – will continue to grow.

You can read more about our network infrastructure upgrades here.