You are viewing a single comment's thread from:
RE: Accelerating STEEM with Intel Optane
Excellent! I know that the main priority of D-App creators (and me) currently is to ensure that mass onboarding can be supported by the blockchain - so any testing that can be done to assess how much load nodes running Optane can sustain would be greatly appreciated.
Agreed. That being said, simulating large amounts of load is going to be a a difficult task. @anyx has a full RPC node running which uses Optane for the RocksDB storage. I think his node will be the best to test the network performance. Right now I feel we are limited to test only the disk I/O. Testing the capability of the blockchain like STEEM also includes how good or bad the components like Appbase and RocksDB work. I think the blockchain as such which was battle tested from the days of "Graphene" and Bitshares-2.0 (& now Chainbase) - infact the right way to address Bitshares / STEEM family of blockchains is to call them as Chainbase ( https://github.com/steemit/steem/tree/master/libraries/chainbase ) based IMHO.
I see, ok. I know that the current network is alleged to be only running at a small percentage of capacity, so it may be that the actual limitation for onboarding comes not from hardware, but from resource credits and delegated steem power available for actually creating functional accounts. It would still be good to have some real world data on that though.. Plus what would occur if the faster nodes go down and we have significantly more users on board that we do currently.
The D-App teams need to be confident in advance that their marketing efforts will not result in network problems soon after.
completely agree. End to End bench marking is an item I was always requesting. There are tools available to benchmark blockchains and its possible to perform the tests. Personally I have been trying to setup a means to effectively test atleast with Tinman but with my office work I have not gone far. But with everyone's support we can do the benchmark with "Caliper". I hope Intel team can support us again once we have the caliper based test suite ready so that a formal bench-marking can be done. :-)