Besides the basic concerns like full HA (two
contollers on separate chassis), full redundant power, premium on site support
(hardware, software and performance issues), excellent monitoring, alerting,
plugins and easy to use dashboards, best raid and shelf protection, what
else?:
·
Large Global Cache with
active/active LUN processing across controllers with automated
load-balancing
·
Active/active LUN access
from any front-end port
·
Independent scaling of
resources (CPU, cache, front-end and back-end ports, and drives) with drives
detached from controllers
·
Single pool of resources
with better than N+1 availability
·
Enough controller
head-room to allow the system to meet performance requirements even when
controllers fail
·
Capacity scaling to PBs
and performance scaling to 100,000s of IOPS
·
End-to-end data
integrity checks
·
Host connectivity
support for all current technologies ( NFS, FC, iSCSI and
FCoE).
·
Efficient use of flash
both as data drives and controller cache extension
·
Near zero second RPO
asynchronous replication that does not impact production
workloads
·
Synchronous replication
that provides continuous application availability when the array goes
down
·
Cloning to enable
parallel processing without impacting production workloads with split and
background synchronization
·
Consistent high
performance with front-end host requests prioritised over back-end array
processes
·
Quality of Service
policies and cache partitioning to prioritise critical
applications
·
Rapid non-disruptive
firmware updates with minimal impact to the array (i.e. ports and cache remain
on-line during upgrades)
·
Non-disruptive
migrations to next generations arrays
·
Efficiency features,
like de-duplication, compression and snapshots, without affecting performance.
Both for file and block level.
·
Does the array use
Copy-on-Write or Redirect-on-Write snapshots?
o CoW snapshots will add significant
additional IO load to the array for each unique write that
occurs
o RoW will not have this overhead,
but performance will be impacted due to the metadata overhead
·
Other
concerns?
No comments:
Post a Comment