api.deathwing.me | Hive Node Stats and Updates

Hey everyone, it's been sometime since I posted the stats of api.deathwing.me one of the most used and fastest public nodes available on the Hive blockchain for the time being.

Bandwidth Stats

Above, you can find the 6 month bandwidth statistics for the node. there is a substantial increase in bandwidth utilized every month. It went from approximately 1TB being transferred to clients (i.e. apps and people using the node) to 7.6TB in November with vnstat estimates being 7.19 TB for the month of December.

Jussi Traffic Analyzer

Thanks to the help of @blocktrades, I was able to get Jussi Traffic Analyzer working on my node. It was Docker logging engine that was messing up the parsing with the analyser, after some edits to Docker, it now works flawlessly.

Below, you can find the last 24 hours (from December 14th to December 15th) statistics of all calls served by api.deathwing.me -- please do note that calls with <100k responses have been filtered out due to the sheer size from this screenshot.

To explain what these mean:

In the last 24 hours, 33.41 million get_block calls were served, these calls are for apps, bots and scripts to see what is in each block and process their own data accordingly. 2 million get_accounts calls were served, these are for when anyone goes into the account information for anyone. Same as get_accounts, almost 2 million posts were served.

In total, for this past 24 hour period, api.deathwing.me has served more than 52 million calls. On average, there were 605 calls per second being received and served by the node.

Upgrades

For some time now, api.deathwing.me had an issue where on random occasions (sometimes during times where more-than-average incoming calls were monitored) would respond with "Internal Error" to any call. These calls were being stuck on Jussi level, in other words, the call wasn't even properly being forwarded to blockchain app, hived.

In a conversation, @blocktrades mentioned that this can be due to the HTTP proxying rather than WS (WebSocket) so thanks to his recommendation, I switched over Jussi's connection to hived server to be made through WebSocket, rather than HTTP. After 24 hours, this seems to have fixed most of the issues as my tests show no internal errors at the moment. Thanks to this, the stability of call responses on the node should be better.

Closing Thoughts

Unfortunately, my DHF proposal for global API node infrastructure was not funded. This means all costs in terms of maintenance and overall server costs are ever-increasing, especially due to the IPv4 shortages affecting the pricing of IPv4 addresses.

If you would like to support me and my work on Hive, you can spend a witness vote on me to vote me as a witness :)

You can visit these links to vote: