Hello, this is SkyPeople CTO.
SkyPeople is maintaining the SR since June 27th.
As we promised on commitment, here we are sharing the Full-node system monitoring and the system we are maintaining for TRON.
Full-node machine of SkyPeople is located in AWS Seoul.
We receive this system monitoring data from AWS-CloudWatch.
Full-node uses x1.32 instance.
TronScan-API server uses t2.xlarge instance.
It is communicating with Full-node, DB Backup, and internal API.
Tron-Telegram that will be launched soon uses t2.medium instance.
Additionally, there are SSH instance, backup instance, test instance, etc..
These instances like backup instance, test instance, etc.. are generating and deleting repeatedly through the constant test.
As a Summary,
Full Node — x1.32
TronScanAPI — t2.xlarge
TronTelegram — t2.medium
x1.32 instance is the highest performance in AWS.
123 core, 2TB memory, 4TB SSD
This is the Full-node system monitoring and personal opinions below :
- CPU Usage
CPU usage rate is moving within 1%.
2. Memory Usage
Command is free –g, and this function is for memory check with GiB unit.
Total memory is 1,921 GiB, and currently using 7 GiB.
3. Network In
Sometimes a lot of network activity is detected. It’s unsteady.
It seems like new node is requesting the synchronization.
Network In is very stabilized at this point.
Compared to “SkyPeople Game Service Network” that hits 2 billion bytes shown in the picture below, Network In usage is very low as it is shown in the picture “Network In” above, which tells us that we have a large scale of capacity.
(Plunge point is the time when we have a game maintenance)
4. Network Out
Network Out is also detecting an amount of network. It is regular occurrence.
This spike is the time when SkyPeople send the rewards out.
SkyPeople proceed the reward giving out through the self node in script.
Network Out is also stable. With the SkyPeople game services, there are 2 billion bytes of instance in Network Out.
5. Disk Size
Output-directory is the folder to save the TRON blockchain.
Currently, it is at the 142M and gradually increasing.
It would be good to compare and analyze with the next report.
6. Error & Failover
So far, there is no error report.
SkyPeople is maintaining the 0 block missed.
In case of unexpected crash, we prepared a script that detects a process down and reboot the process. It never activated, not even once.
TRON recommend SR for x1.16 instance.
SkyPeople chose x1.32 instance.
TRON full node is using very low performance now.
Node has been maintained in testnet, and it is known as requiring a few system resources.
When the mainnet launched, I worried about the exchanges’ transaction. I expected that there would be a huge amount of “TransferContract” and “VoteContract” occurrence. However, there weren’t so many transactions afterall.
Although there aren’t so many transactions occurred, it seems to be TRON node source code is well-made, because they are keeping the resources without any troubles until now.
With this amount of usage, half of the recommended system requirement x1.16(64core), which is 32 core would be quite enough to maintain.
SkyPeople will keep this infra.
We are looking forward to see the elements to use the infra resource like the TRON Virtual Machine, etc.
And that will be the next time I share the system monitoring again.
Skypeople Official Twitter :
Skypeople TRON SR Official telegram Group :