Developers Guide

7 Dell EMC Storage Manager Scalability Solutions | 3395-BP-SC
1.4 Geographic distance
When the Data Collector is separated from the arrays by distance, the added network latency can possibly
magnify the transaction times, reducing the total number of volumes and arrays that can be managed by a
single Data Collector. While certain client transactions can tolerate up to 75 ms of network latency before
crossing the 10-second threshold testing goal, other types of transactions can exceed the goal with just 30 ms
of latency separating them. While individual results may vary, benchmarking shows that the DSM client
operations routinely finish under the 10-second goal where network latencies are less than 50 ms. Of course,
every environment is different, so administrators must adjust the number of arrays each Data Collector
manages to suit their specific environment.
1.5 Failure domains
When supporting multi-data-center replicated environments, if more than one Data Collector is required, Dell
recommends separating arrays into failure domains. In other words, arrays that use each other for disaster
recovery, fail-over, or business continuance, must be added to the same Data Collector. In hub and spoke
replication architectures, this means dedicating at least one pair of primary and secondary Data Collectors per
spoke. Keep in mind that based on latency between sites, multiple Data Collector pairs may be required
between each site for larger environments.