> There is no way that this can be true unless there is some very very bad craziness going on. A real mindfuck
System Center Operations Manager and all derived products such as the various Azure Monitor and Azure Logs offerings have craziness going on.
For example, a typical metric is stored with a 300:1 write amplification because the database schema they use repeats the fully qualified machine name, counter name, and counter instance name per measurement. This prevents the collection of a useful set of metrics at a useful frequency.
Logs are similarly generated using suuuuuuper verbose JSON, capturing everything except the critical fields you actually wanted, and then stored uncompressed record-by-record, keeping the full field names for each and every row.
What incentive do you think they have to improve on the efficiency of the wire format or the storage format if they charge by the gigabyte?
System Center Operations Manager and all derived products such as the various Azure Monitor and Azure Logs offerings have craziness going on.
For example, a typical metric is stored with a 300:1 write amplification because the database schema they use repeats the fully qualified machine name, counter name, and counter instance name per measurement. This prevents the collection of a useful set of metrics at a useful frequency.
Logs are similarly generated using suuuuuuper verbose JSON, capturing everything except the critical fields you actually wanted, and then stored uncompressed record-by-record, keeping the full field names for each and every row.
What incentive do you think they have to improve on the efficiency of the wire format or the storage format if they charge by the gigabyte?