Automic Workload Automation

Hello All,

For a long time, I've been using the SYS_BUSY_?? function to retrieve the percentage of workload usage from the AutomationEngine to display it in reports, graphics and so.

But actually, I always had questions about what kind of information this usage carries and I'll count on my community fellows to guide me through them:

1) What is considered on the workload usage calculation (db performance, number of executions, number of connections, ...)?
2) In the usage graphic of the System Overview window, what is the math behind it (60 min all wps average, for example?)?
3) Since the sys_busy cannot retrieve the information from a particular WP, is is possible to retrieve the columns B.01, B.10 and B60 (System Overview > AutomationEngine) data from the db? Which table(s) contain this info?

Thanks!!

Simone Martineschen said:

3) Since the sys_busy cannot retrieve the information from a particular WP, is is possible to retrieve the columns B.01, B.10 and B60 (System Overview > AutomationEngine) data from the db? Which table(s) contain this info?

Hi Simone!

I'll jump in quickly on #3:
Yes, this information is in the db.  Actually, almost all the information from the AE is in the db.  I don't have a specific location at this point, but here are two suggestions:
1. In the directory <AEHome>\Utility\db\_structure\HTML you can see all of the tables with brief descriptions of what they contain. Things that start with "A" are usually archived stats, "E" is normally things in the Activities window, "O" are object definitions, "JP" are workflow definitions, "MQ" are Message Queue information--temporary workspaces as to what is connected.  With that info you can rule out a bunch of tables and start exploring for the correct info.  I'd start with the MQ tables.
2. PhilippElmer is running an entire course on the database schema and advanced topics like this at Automic World this year (Nov 3, 2015).  This is a first for the US, and anyone could gain some valuable advanced info from this class.

Hi Timothy,
Thanks for your reply!
Yes, I usually use the db structure documentation and I tried some of the MQ tables but the information was not clear enough. I'll double check them.
Any thoughts on the other questions?

Oh, and I'm a big fan of PhilippElmer blog! Great insights on DB queries.
PhilippElmer

The person who explained the B01 B10 B60 columns best to me was Mitch_Denis_38.  I'd hate to mess up your understanding by translating his explanation through my own filter from several years ago.   :D

I'll ping him and see if he can take a minute to reply.

Hi all,
Sure, I can put in what I understand of these columns.

The biggest mistake that we see customers make is that they think that the B.01, B.10, and B.60 columns directly translate to CPU usage and capacity.  In fact all that they are showing is the amount of time during the last 1 minutes, 10 minutes, and 60 minutes respectively that the process was working.

For example, if you are looking at the B.10 column and it shows 50 for WP1, that simply means that WP was working for 5 minutes of the last 10 minutes.  It doesn't really have anything to do with capacity - in fact it could be that they were running one database statement for 5 minutes and it would show 50 for the B.10 column (hopefully if a single db statement is taking 5 minutes, you are aware of what is causing it ;)), or it could be that it ran 300 things that took 1 second each or 3000 things that took .1 seconds each.  

Should you be alarmed at a high number in the B.01, B.10, and B.60 columns?
Not necessarily.  If there are no other symptoms accompanying this, there's probably nothing to worry about.  If there's no high CPU usage at the same time for the processes, there's probably nothing to worry about.  If the B.01 is at 100, but the B.10 is at 1, and the B.60 is at 1, there's probably nothing to worry about.

When should you be concerned?
If there's a large upswing in the numbers very suddenly.  For example, if your B.01, B.10, and B.60 are usually at a relatively low level and then are consistently higher (perhaps after a restart or slowdown of some sort) and don't go down, you should probably contact Support.  

Usually, there's no reason to panic if the numbers are higher than you think they should be.  I think there's only been five or six times over the last eight years that I've seen a correlation between the Usage columns and actual problems with the Automation Engine.  Usually we'll notice other behaviors before checking the usage columns.

Thanks Mitch_Denis_38!  This is a great post that will probably help many people.

One other related thought on usage--since it is difficult to determine what "high usage" entails you should consider utilizing the "resources" on the job "Attributes" tab.

I'll give you an example:  One customer had a job that required a lot of DB processing power on a SQL agent.  While the job was being worked in the AE or was waiting, the SQL agent did nothing, and the WPs didn't do a whole lot. As soon as it was needed the DB would spike to 100% usage on RAM, and this was a critical process.

So, we set the 'resources' to 10,000 on this job and set the agent to limit usage to 10,000.  Almost all other jobs had usage of 1.  This made sure that the job had everything it needed to complete.  Something similar could also be accomplished with Queues.

Thanks Mitch_Denis_38  and Timothy_Dodd_84  for the enlightment on usage data. Much easier now to correlate. I still could not retrieve the system overview values from the DB. I've got no luck with the MQ tables. Any ideas?

Hello All,
Did anybody manage to retrieve the system overview > Automation Engine > usage information from the MQ tables or any other ways?

I'm interested in this as well. It's rarely the OS resources that get over utilized in our environment, but we see our core processes (WPs/CPs) get pegged at 100% pretty often. It would be extremely helpful to monitor this somehow. We're dealing with a pegging CP at this very moment actually!

I brought this up previously here: https://community.automic.com/discussion/686/how-do-you-measure-your-system-usage-over-the-long-term

I note in that thread that I had been told usage metrics are computed on the fly and actually not stored in the database, which is really too bad. They are output in the logs every so often, though. Scraping the logs for this information isn't ideal if this is the only way to get at this information.

Like Simone, I have looked through the DB schema but haven't come up with where this stuff is located if it's there at all.