EMC's benchmarking...
The previous post was maybe a little more technical than
people want to see, so here's the deal in hopefully, non-
technical talk. EMC's benchmarking for SPEC hides a couple
performance problems for their Celerra NAS product by
one thing they don't do for the benchmark and one thing they
do, do, for the benchmark:
1. EMC benchmarks with RAID turned off. RAID (Redundant
Array of Inexpensive/Independent Disks) allows a system
to have data duplicated, so if one disk fails the data
can be regenerated from other disks. Obviously, with
RAID *on*, you spend more time creating the duplicated
data so your numbers may not be as good. EMC always
benchmarks with RAID *off*, NTAP always benchmarks with
RAID *on*.
2. EMC benchmarks with *lots* of memory and NVRAM. Memory
is used to store data so you don't have to go to the disks
as often, thereby getting better numbers. NVRAM (Non-Volatile
Random Access Memory) is just another kind of memory that
keeps its information stored even if you lose power to the
system - it is used in addition to memory by both NTAP and
EMC so the data remains consistent on disk if power fails to
the system. EMC puts lots of memory in the Celerra and also
puts lots of memory and NVRAM in the Symmetrix (where the
disks are) to get better benchmark numbers. NTAP benchmarks
with much smaller memory and NVRAM configurations. Basically
EMC creates a very costly configuration because of all the
memory/NVRAM and disks to bump up their numbers.
Benchmarks are good for comparing technology as long as
you are comparing apples and apples. EMC and other companies
rarely want to compare apple-to-apple with NTAP because
they know the results will not be flattering. NTAP
really does have the best technology - the best use of
software on a given hardware platform.
This is just benchmarking though, I think the more important
point is the ease-of-use features of a product that was
built from the ground up as a "data appliance", there
isn't anything comparable out there. |
