What will be the network bandwidth required in between a geographically redundant 1+1 DMA-E500 setup (as per the attached diagram), managing M&C works for broadcast downlink, MCR and compression along with a max of 20 concurrent SRM booking with resources from both sites? 
I've checked the answers from the link below and am still in need of more details to match the above elaborated scenario in order to have an acceptable justification for the required bandwidth.
Yes Ben, we will have one DMA-500 on the main site, and a hot backup on the backup site. By resources, I was refering to the resources within SRM service workflow that will be manageable by DM.
Hi Naveen,
Firstly, of course this is not exact science, because there are many unknown variables. Variables that you do not know, nor anybody else. But let's see if we can break it down.
Assumptions:
- we have a DMA-500 on the main location
- the DMA-500 does not use the inter-site link to communicate with the managed products, so we do not have to account for traffic from the data-collection plane.
- the users of the system are located at the Main Site, so we do not have to account for the client traffic.
Or in other words, the traffic on the Inter-Site link is (under normal operation) purely used to synch the back-up DMA with the main DMA.
Therefore, under normal operational circumstances, that means that there would be traffic flowing from the main to the backup site for the following:
- Alarms: whenever alarms are detected and registered by the main DMA, those alarms would also flow to the backup DMA - that should be minimal, we have maximum 500 devices, and if properly set up and configured.
- Performance Trending: that could be more heavy, i.e. the question is really on how many metrics the user will activate trending. Note that if you have products that you manage that contain tables of data, and you enable on one of those metrics (which represents a column), it then depends on how many rows there in that table. Furthermore it also depends on the nature of the metric (i.e. DataMiner stores the data intelligently and will not continuously store data if the values are not changing - if you have a bitrate, each value can be a new one that needs to be stored, if you have something like a temperature, that would only change more occasionally).
- Configuration Changes: if a user changes configuration (e.g. adding an element to the system, changing a Visual Overview graphic, creating or changing an alarm template, etc.), that needs to be synched up as well with the back-up system. The question is again how much that will happen, and how much of that will be the more heavy activities (e.g. very rich and graphical Visual Overview designs). This category would also include traffic from SRM bookings, that information would synch as well. Again this largely depends on the type of SRM, the nbr of bookings and all that.
I believe those are the most important considerations. Given all known variables, and assuming reasonable use of the system, I would say that a 250 Kbps capacity should acceptable. Of course again with the caveat here that it really depends on how the system is used.
I also would like to add that we need to check what happens if we have a fail-over event. In that case, will the polling traffic go from the back-up DMA straight to the products that are managed, or will that flow across the inter-site link across the main site? Will the users have to connect with the back-up DMA via the inter-site link. If that were to be the case, then maybe that specific scenario also needs to be looked at.
And then there is the consideration of what the constraints are on that inter-site link. Is that a leased line? Is that shared with other applications or is that reserved and dedicated bandwidth? How easily can be it be scaled up (in which case you could go with a more conservative estimate, and scale it up based on actual usage data)?
Hi Ben, Rahimi here from Astro. Thank you for the elaboration above.
Just to zoom in further on Naveen’s shared diagram above (minus the connection of the remote site), we will be having DMA-E500 main on main broadcast site (AABC) and DMA-E500 hot backup at the backup broadcast site (ACBC). However, the elements connected are 250 devices at the main site and 250 devices at the backup site (estimate only). The entire traffic along with devices should be using the inter-site fiber link for communication with the main DataMiner at any one point in time. Similar situation in a fail-over incident just that the backup site DMA will become the main one then.
Also, the DataMiner client (user interface) will be mainly used at the main site (AABC) but then in times, we would require to manage DataMiner from the backup site (ACBC) as well.
So in this event, what should we start with as the fiber bw requirement?
HI Rahimi,
Welcome to the community, glad to have you here. Ok, gotcha. 250 elements are on one location, 250 on the other, so A) the main DMA is active and will be polling 250 local elements, and will go across the link also for the polling of the other 250 elements, or B) after a failover, the backup DMA will be active and it will the same as the main, but in the reverse direction. So that means that in addition to the aforementioned traffic and estimate, there will be some added traffic because of the polling of 250 elements, and the occasional users that will connect to the system from one site across to the other site.
The client communications is not going to be really adding so much (assuming a regular user type of profile, and not a designer / dev profile that would be updating Visual Overview graphics all the time), but the key thing for that type of traffic is that it is more burst type of traffic (e.g. more bandwidth is good to smoothly and quickly connect the client to the system, after that, you really don’t need that much). But if the link is somewhat limited, it could take slightly longer for the client to connect, or download a new version of a graphic for example (which is only when it is new, after that it is cached by the client, it’s not like it does it each time again and again).
The polling traffic is more continuous, around the clock. The tricky thing there is of course that this can largely vary depending on the type of protocol used by the devices, how fast data is collected, etc. So we can only make assumptions here.
Based on the client and inter-node traffic, I would go with 500kbps now (to make sure the client is not delayed to much in connecting and loading the more sporadic data it needs), but then we have to add also the polling traffic. I’ll send you some information in a separate response in a just a second for that.
Just to clarify, you have one DMA-500 on the main site, and a hot backup on the backup site? Or is there a DMA-500 on each site, with each of them have a redundant counterpart on the other site? Because I got a little confused about the statement “with resources from both sites”.