DX Unified Infrastructure Management

Hi,

What is the recommended configuration for qos data in data_engine and nas retention period. We have 180 days for raw data and 1 day for historic data which I feel is not proper. I am trying to change these values and we have enough space in our DB. Currently we have around 60 % space free. With this configuration, what do you this can be the values for

1) Delete raw data older than

2) Delete historic data older than

I am looking to have atleast 6 months of qos data.

And what will be the impact of increasing the alarm retention period. Currently, the rentention period for alerts is 30 days. Can we increase that to say 180 days? Will this cause any impacts on the DB?

-kag

I shouldn't think altering those settings will cause any issues for you. For QoS, it is determined more by demands of the business rather than the DB. For example SLA calculations. Of course, if you store a lot of data in your raw tables, their size will grow, their maintenance operation durations will grow, etc. Still, I shouldn't think there's an issue there. I'm keeping a year's worth of raw data.

For nas it's just pretty much about the DB size, I'd say. The transaction db and the alarm db are on the same server with nas, so those will grow and if your alarm volumes are high, I guess it might impact things. Don't know if there's any best practices regarding that. There's scarcely any 'best practices' with UIM really..

Also noteworthy, in data_engine configuration you can also configure threshold for the dn tables (daily average). This is not in the IM gui and if I remember correctly, also not in the AC gui.

-jon

Thanks Jon. Just another question. What is the difference of preserving the qos on raw data against history. I understand the raw data is kinda heavy where as the historic ones are the compressed form on the raw data (HN Table).

-kag

That's right. RN_ stores all samples, the amount depending on sample interval, for the period defines. Hourly averages are stored in the HN tables and daily averages in the DN tables.

-jon

Thanks Jon. Having said this, it is prudent to have an extended period of the HN Table. In my case, I would like to change raw data to say 100 days and historic to 200 days and alarm retenction to 90 days from 30 days. I hope this should be a good represenatation of numbers. I have about 800 gigs of storage on my DB and with the current configuration I am assuming this would give a good represenatation. And I am sure there will be no more addition of QOS as I am trying to drill down few metrics.

-kag

You should be careful when reducing the retention period for raw data though. If you go from 200->100 or something like that, it'll have to do the hourly aggregate for every object for 100 days and remove the data as well. It will do this at the configured data_engine maintenance time, which is something like 00:20 if I remember correctly. It might be a good idea to do this chunk by chunk, like 5-10 days per day, especially if you have a large number of qos objects and your database isn't partitioned.

-jon

Hi Jon,

If my understanding is proper, I think we should not change the raw data and leave it as it is and probably try increasing the historic data say from 1 day to 10 days and observe for a while. And repeat this cycle until a desirable amount is achieved. Please correct me if I am wrong.

-kag

You can change it upwards at once, no problem there.

So, doing this will not impact anything?

-kag

If you only increase the threshold for historic, there shouldn't be any side effects.