My Computer and I: One Way Out

My Dell Studio XPS 8100 has served me well for years now. Sooner or later, however, problems will arise, often ones that are difficult to troubleshoot. That goes with any computer system, anyway.

I have no formal training in information technology, yet I try to troubleshoot computer problems myself. With that arises a somewhat crude scientific investigation, testing one component at a time until the issue is located and fixed. So, what is the problem this time?

The problem:
While playing World of Warcraft, the game would go through random cycles of lag spikes and catch-up, lasting up to minutes at a time, sometimes disconnecting completely. While surfing the Web, webpages would randomly take minutes to load, if at all, and refreshing would sometimes retrieve the webpage in full.

Background information:

• TCP (Transmission Control Protocol) is a protocol that handles the connections between computers. As it allows reserving of resources, as well as ensures the integrity of the data transported regardless of conditions, it is ideal for the transfer of files, as well as conveying instructions between client and server.
• If a TCP connection suffers packet loss, latency can easily result, as the sending computer needs to resend the packet that was lost.
• The quality of the physical line affects the connection, as noise and attenuation can reduce the throughput of data, and even corrupt it, resulting in packet loss.
• A router can only forward a packet when the sending line is free. If it is busy, yet there are packets arriving, they must be held in a buffer. The time the packet spends in this buffer adds to latency.
• When buffers start filling up, network congestion results. Once a router's buffer is full, it must decide to do with the excess packets, and some may be dropped as a result, causing packet loss.
• An ISP (Internet Service Provider) is not free to use any Internet link at any time. Unless they own it, they must enter into a contract with the owner to lease capacity, then establish a point of presence (PoP) at a node on it by installing their own equipment.

The circumstances:

• My telephone line is owned by Telstra, and my ISP is iiNet. Among others, iiNet operates capacity on the Southern Cross and SEA-ME-WE 3 submarine cables, in addition to those running throughout Australia.
• It has been years since I last replaced both my modem, a NetComm NB6PLUS4Wn, and the line filter for my landline telephone.
• The World of Warcraft server I play on is located in Sydney (realmpop.com reports Melbourne), with latency normally falling as low as 11ms.
• Many of those I play with on World of Warcraft did not report any lag at the same time as my problem.
• I got a much better experience tethering my desktop computer to my mobile phone, such that it uses the phone's 3G connection.
• The lag spikes would happen around 6pm-11pm, during which is peak Internet usage in Australia. It is unfortunate that my World of Warcraft raids are scheduled during this time.

The experience:

• I rebooted the router (yes, it is a computer too), many times.
• I disconnected the telephone and line filter from the telephone line.
• I called iiNet about the problem, where they performed things such as line tests, modem configurations remotely (which scares me somewhat) and port refreshes.
• To gather more information about what was happening, I did a ping test both on pingtest.net and to my server using WinMTR. The results were very worrying:

|--------------------------------------------------------------------------|
|                              WinMTR statistics                           |
|              Host       -   %  | Sent | Recv | Best | Avrg | Wrst | Last |
|--------------------------------|------|------|------|------|------|------|
|         NB6Plus4Wn.home -    0 |  804 |  804 |    0 |    0 |   10 |    0 |
| ***.****.****.on.ii.net -    5 |  673 |  640 |   14 |   86 |  381 |   61 |
|  ***.***.****.on.ii.net -    5 |  689 |  660 |   14 |   88 |  369 |   62 |
|  ***.***.syd4.on.ii.net -    6 |  657 |  620 |   14 |   94 |  444 |   61 |
|   55497.syd.equinix.com -    6 |  657 |  620 |   15 |   84 |  370 |   62 |
|   No response from host -  100 |  161 |    0 |    0 |    0 |    0 |    0 |
|           103.4.115.*** -    6 |  661 |  625 |   14 |   84 |  358 |   60 |
|________________________________|______|______|______|______|______|______|
WinMTR v0.92 GPL V2 by Appnor MSP - Fully Managed Hosting & Cloud Provider

I found out that there was significant packet loss happening, so I began to suspect network congestion. Looking through the iiNet Network Status page, I noticed an event noting the cutting of a submarine cable between Perth and Singapore, and access to international sites slowing down as a result. That by itself would definitely explain my Web browsing experience, but not that for World of Warcraft, as the server is located domestically.

After some more research, I found Greg's Cable Map, which plotted every international submarine cable in the world. Looking at Australia's connectivity, I found something disappointing:

The cables currently running off Perth, of which there are only one.

It turns out that there is only one cable leaving Perth, which is SEA-ME-WE 3. Compare that to Sydney's situation:

The cables currently running off Sydney. Some go through Guam, while others end up at the United States.

The sorrow is compounded by the fact that there are at least three Perth cables currently under construction (Australia-Singapore Cable, APX West and Trident), with all of them originally due for operation this year. We could certainly have used one around this time!

Putting everything together, it seems that, when SEA-ME-WE 3 went down, all of iiNet's international traffic needed to be funneled through Sydney. They would have provisioned capacity throughout their network assuming there was a path out of Australia through Perth. With that gone for the meantime, iiNet's Sydney routers may well have been pushed beyond capacity during the evening peak, requiring them to drop packets (among them, mine bound for Blizzard's Sydney servers).

Many of my online friends seemed perfectly fine with their connections; that may well be because they use ISPs that do not depend on the Perth-Singapore route. For example, TPG's sole link to the outside world (mind the pun) is the PPC-1 cable to Guam. They would have generously provisioned their Sydney routers to handle the traffic of the entire nation.

Implications:

• Network congestion can contribute significantly to latency. You know the situation is dire when packets are being dropped.
• For latency-sensitive applications such as online gaming, any amount of packet loss is highly detrimental to the experience.
• Network congestion, or any Internet issue, for that matter, is out of any end user's control. They are beholden to the commercial interests of their ISP and, in turn, the ISPs of their ISP (tier-1 carriers).
• Australia needs more routes out now than ever. The completion of the cables running from Perth will allow the country's ISPs to diversify their international links more.
• There are currently no cables running from Darwin. A cable or two here can add more much needed redundancy.

Bibliography

• iiNet (2015), Event 4449509 | iiNet Network Status - Internet Service, web page, viewed 25 November 2015, <http://www.iinet.net.au/status/4449509>.
• Mahlknecht, Greg (2015). Greg's Cable Map, web page, viewed 31 October 2015, <http://www.cablemap.info/>.
• Singapore Traffic From Perth Routing East Coast (2015), forum post, viewed 11 November 2015, <http://forums.whirlpool.net.au/forum-replies.cfm?t=2453648>.
• TPG Internet (2015), About TPG Network, web page, viewed 25 November 2015, <http://www.tpg.com.au/about/networks.php>.