When can we say live stream buffering issues are based on viewer side - region/ISP congestion issues or whether buffering is due to local network issues?

Hi all,

We recently held a live event where we experienced a lot of buffering on youtube and we're trying to understand where the issue came from. The situation was as follows: We were pushing 5 different streams (different languages) to Wowza cloud which was distributing to various destinations including Youtube. One point to note is our wowza settings were going to Akamai first so we could use Primary/Backup ingest setup (we came to later know that this "route" ads an extra step in the delivery that can cause latency/buffering). On one system (Imac 2017), the destination was our main Youtube channel, for a few hours things were generally OK, however at one point during the event we reached appx 60K concurrent viewers. This is the time when we started seeing from our side and viewers side lots of buffering. Unfortunately, at the exact same time, Youtube was giving us an internal 500 server error, which prohibited us from viewing our dashboard and stream health, but we did notice our Wowza stream health had lots of incoming/outgoing datarate fluctuations. The buffering issue wasnt the case on the other channels that we were streaming to. Because the buffering was so bad, we were constantly switching on/off our wirecast feed to resume the stream, where it would switch between Akamai's Primary and Backup. Strangely, once the archive video was processed, we didnt see much issue except for this switching between Primary & Backup. Once the concurrent views dropped back to appx. 5K, things again settled back down and we didnt have any more issues.

So, we're wanting to understand a few things:

• What could have been the root cause(s) for this issue?

• Is this an issue from Encoder to Youtube or Youtube to User issue? Or both? (We had plenty of local bandwidth, but got buffering and "steam not available" message on our player)

• Should we assume that everyone had buffering? We didnt specifically ask in comments if anyone was seeing the feed well, so we arent sure.

• How do you avoid (and test) for something like this from happening in the future?