Dell EMC VxFlex Ready Node, Руководство пользователя

Страница: 34 / 354

Rebuild
VxFlex OS initiates a Rebuild process in response to failure. Forward Rebuild refers to
creating a new copy of the data on another server. Backward Rebuild refers to re-
synchronizing one of the copies.
When a failure occurs, such as on a server, device or network failure, VxFlex OS
immediately initiates a process of protecting the data. This process is called Rebuild,
and comes in two flavors:
l
Forward rebuild is the process of creating another copy of the data on a new
server. In this process, all the devices in the Storage Pool work together, in a
many-to-many fashion, to create new copies of all the failed storage blocks. This
method ensures an extremely fast rebuild.
l
Backward rebuild is the process of re-synchronization of one of the copies. This is
done by passing to the copy only changes made to the data while this copy was
inaccessible. This process minimizes the amount of data transferred over the
network during recovery.
VxFlex OS automatically selects the type of rebuild to perform. This implies that in
some cases, more data will be transferred to minimize the time that the user data is
not fully protected.
Rebuild throttling
The rebuild throttling policy determines the priority of rebuild I/Os versus application
I/Os when accessing SDS devices. The possible rebuild throttling policies are no limit
on rebuild I/Os, limit concurrent I/Os per SDS device, Favor application I/Os and
Dynamic Bandwidth Throttling.
Rebuild throttling sets the rebuild priority policy for a Storage Pool. The policy
determines the priority between the rebuild I/O and the application I/O when
accessing SDS devices. Please note that application I/Os are continuously served.
Applying rebuild throttling will on one hand increase the time the system is exposed
with a single copy of some of data, but on the other hand, will reduce the impact on
the application. One has to make a decision and choose the right balance between the
two.
The following possible priority policies may be applied:
l
No Limit: No limit on rebuild I/Os.
Any rebuild I/O is submitted to the device immediately, without further queuing.
Please note that rebuild I/Os are relatively large and hence setting this policy will
speed up the rebuild, but will have the maximal effect on the application I/O.
l
Limit Concurrent I/O: Limit the number of concurrent rebuild I/Os per SDS device
(default).
The rebuild I/Os are limited to a predefined number of concurrent I/Os. Once the
limit is reached, the next incoming rebuild I/O waits until the completion of a
currently executed rebuild I/O. This will complete the Rebuild quickly for best
reliability, however, there is a risk of host application impact.
l
Favor Application I/O: Limit rebuild in both bandwidth and concurrent I/Os.
The rebuild I/Os are limited both in bandwidth and in the amount of concurrent
I/Os. As long as the number of concurrent rebuild I/Os, and the bandwidth they
consume, do not exceed the predefined limits, rebuild I/Os will be served. Once
Architecture
34 Dell EMC VxFlex Ready Node AMS User Guide