DCE polling engine feature.

[DCIM_Support]

Hi team,

Could you please highlight the DCE feature?

What does DCE do if it hove no time to ask all devices which are monitoring during scan interval? Does DCE use multithread polling engine?

In the other words. How can we take the time which DCE spent to poll devices parameters for single device or for all connected devices?

Thanks for answer.

(CID:102924577)

[DCIM_Support]

Hi Dmitry,

I believe it is multithreaded but I will need to verify with engineering. Is there a specific reason for this question? Other than a customer simply wondering if it is multithreaded, what is the objective behind this question?

Steve

(CID:102924582)

[DCIM_Support]

Hi Steven. The question is rising from the trouble in the huge configuration which contain 3000 devices which are distributed over hundreds locations. Today the customer uses physical DCE which does not work enough swift. DCE does not poll devices at the adjusted time. The customer is looking over to replace DCE with a virtual appliance with extra CPU cores and memory. The solution is reasonable at the case of multithread polling engine only.

(CID:102924724)

[DCIM_Support]

Hi Dmitry, With 3000 devices, it may end up being slow especially if you also have NetBotz, surveillance, remote shares, backups, etc. Each of these would be their own thread however I have yet to be able to get an answer on the polling in regards to multi-threading. I have sent another e-mail but if the system is bogging down there are only 2 things you can do. You can, as you're suggesting, use a VM and provide it with more resources or 2, lighten the load on the current system. Please note that lightening the load does not mean removing devices but you may want to increase polling cycles, decrease anything to do with surveillance (image size, frame rate, etc), Knowing if the system is multi-threading may not really help in this endeavor but as I mentioned, I have a few e-mails out to engineers and I am waiting to hear back. Steve

(CID:102924727)

[DCIM_Support]

Hi Steve, The system does not include NetBotz, surveillance, etc. Only SNMP devices. DCE client and web interface works good. But DCE catches changes of devices states extremely slowly. In addition to huge size the system has couple features. The customer claims to use only encrypted communications, so DCE communicates with devices over SNMP v3 with encrypt only. Also the system uses distributed communications which have different traffic capacity. A part of channels are very slowly. My suggestion is. DCE use single thread SNMP daemon and poll devices one by one. Because of communication troubles DCE has no time to poll device just at time. If my suggestion is correct the customer has no reason to increase CPUs. If i am wrong the solution may help.

(CID:102924766)

[DCIM_Support]

Hi Dmitry, If the devices are not APC, then they would not have priority scanning. Such devices must wait until the next polling cycle. If the system is set to 5 minutes, that means it could take up to 5 minutes to see a state change. If they are APC devices, SNMP version 3 will indeed slow things down a little as there is not only about double the amount of data being transferred, you have to encrypt and decrypt everything which will take extra time. Extra memory and processors may help in the processing of all this data. Steve

(CID:102924805)

[DCIM_Support]

P.S. If they want to test this, try giving them temporary keys for a VM so they can test in their environment if they'd like. A VM does not have to be activated to use temp keys s

(CID:102924806)

[DCIM_Support]

Hi Steven, Thank you for answer. The configuration includes 2000 APC devices and 1000 not APC ones. The priority scanning is configured for APC devices but it does not help enough. As i suggest we have to know how much CPU and memory resources are used on DCE now to calculate CPU cores and memory which have to be granted to virtual appliance to achieve necessary performance. Could you please advise how can we catch CPU and memory utilization on the physical DCE?

(CID:102924909)

[DCIM_Support]

Hi Dmitry,

I am told:
 "DCE uses an asynchonous (non-blocking) scan engine". 

Steve

(CID:102924735)

[DCIM_Support]

Hi Steven, Thank you for answer. The configuration includes 2000 APC devices and 1000 not APC ones. The priority scaning is configured for APC devices but it doers not help enough. As i suggest we have to know how much CPU and memory resources are used on DCE now to calculate CPU cores and memory which have to be granted to virtual appliance to achieve necessary performance. Could you please advise how can we catch CPU and memory utilization on the physical DCE?

(CID:102924932)

[DCIM_Support]

Hi Dmitry,

Priority scanning is a process where an APC device is configured with DCE as a hidden trap receiver. When there is an event on the APC device, a trap specific to this feature is sent from the device to DCE. DCE then sees this message and polls the device for alert information. The alert is then verified and DCE alerts (or not) based on it's configuration and essential alert DDFs. It is designed this way because on older systems (ISX manager) the device might send a false alarm and manager would simply relay that false alarm.

This feature causes the manager and now DCE to verify the alarm before sending out what may be a false alarm. This feature is not implemented on 3rd party devices to this point as we do not have the correct OID to write ourselves in as a trap receiver to all devices and the specific communications that is expected at this time from the device is not available in all systems. This may change in later versions of DCE but I can not verify at this time.

As for performance, I did also just realize I may have forgotten to mention that if you're using remote repositories, this will both take up CPU and memory but will also cause much more network traffic. The web page for DCE has a netstat feature (status/troubleshoot network/netstat) that can give some potential insight as to what is happening there.

As for the usage specifically (memory, cpu, etc), engineering is looking for a way to make this information visible in a future version of DCE, maybe 8.x. It's not as simple as looking at a web page but I am sending you an off-line e-mail about it and perhaps you could find this helpful.

Steve

(CID:102924955)

[DCIM_Support]

This question is closed for comments. You're welcome to start a new topic if you have further comments on this issue.