Proof
Points
Two separate rounds of independent testing have each
confirmed that the Green Platform Anti-Vibration Rack™ can
increase I/O read performance of hard disk drives by over
200% for random data.
Test Round 1: Lab Testing at Sun Microsystems
Lab testing of our prototype AVR-1000™ with Sun Microsystems
conducted in 2008/2009 demonstrated that increasing
vibration caused a 65% reduction in I/O throughput and
almost tripled the energy required to complete a given I/O
task in servers and storage devices that use or interact
with HDDs. HDD’s performance are most system’s bottleneck,
because the vibration penalty on HDDs degrades the
performance and energy efficiency of the system.
Sun then tested the ability of Green Platform’s AVR™ to
mitigate vibration by placing the rack on a shaker table.
The following chart displays the results of this testing
where the upper (red) line represents vibration IN and the
lower (green) line represents vibration OUT. Sun determined
that Green Platform’s AVR™ solution effectively reduces
vibration measured at the storage device by a factor of 10
to 1000X across the critical frequency range.
Green Platform Reduces Vibration
10-1000X
Test Round 2: Data Center Testing by
Q-Associates
In late 2009, Q Associates in partnership with Green
Platform Corporation performed a series of tests and
benchmarks to determine the effect of Data Center vibration
on an end-to-end compute environment. The testing shows that
by reducing the level of vibration in storage systems, one
can expect faster system performance and proportionally less
overall energy use. These tests revealed that ambient
vibration inherent in a world-class, raised floor Data
Center caused performance degradation of up to 246% for
random reads and up to 88% degradation for random writes for
an enterprise class storage system. A prototype Green
Platform AVR-1000™ anti-vibration rack was tested within the
same environment and shown to significantly reduce or
eliminate the detrimental vibration effects resulting in
significantly increased performance. Since storage and
storage access are a system bottleneck, increasing disc
throughput for I/O intensive jobs reduces the time required
to complete the job. Less system computation time results in
shorter system runs and less computation and total Data
Center energy use.
The objectives of the testing were:
Determine how typical levels of Data
Center vibration affect storage system performance in a
side-by-side comparison of a typical metal rack vs. the GPC
Anti-Vibration Rack™:
-
Measure
the throughput performance of random and streaming reads
and writes under varying vibration loads.
-
Perform
suite of Macro Benchmark tests to determine actual
compute system performance difference of simulated
loads.
-
Perform
suite of Micro Benchmark tests to determine the actual
compute system performance difference of simulated
loads.
In excess of 300 individual tests were
performed and recorded. The simple compute environment
consisted of a Sun X4440 host, Sun 7110 Unified Storage NAS
using 2.5” 10,000 rpm SAS drives and 16 ports Gigabit
Ethernet switch. The two separate physical environments used
for testing. The first consisted of a sound isolation room
with no external vibration, and the second was a Tier 1 Data
Center equipped with both a standard Chatsworth Products Inc
(CPI) metal freestanding rack (APC style) and a Green
Platform Model AVR-1000 Anti-Vibration Rack™.
The most significant impact of
vibration on HDD throughput was for 1M blocks of random data
I/O, where dissipation of vibration by the GPC AVR-1000
yielded improvements ranging from 56% to 246% for read
operations and from 34% to 88% for write operations.
Micro Benchmarks were performed for random reads and writes
for 2k, 8k, and 1M block sizes as well as single stream
reads and writes of 1M and 10M block sizes. Single stream
reads showed a performance improvement of approximately 10%
and single stream writes showed a performance improvement of
approximately 5% when the storage server was moved from
metal rack to composite Anti-Vibration Rack™. Macro
Benchmarks were performed for File-Server, Varmail, Web
Proxy, and Web-Server workloads containing both random and
sequential data. By moving the storage server from a
standard metal rack to an Anti-Vibration Rack™ in the same
environment, performance increased from 2% for a web server
workload up to 22% for Varmail. File server and web proxy
performance increases were 16% and 20% respectively.
For streaming writes, measured
performance differences were less than 4% regardless of
environment or external vibration caused by sound up to
120dB. Even though only Sun Microsystems products were
utilized for these tests, vibration is not an issue isolated
to just one vendor or one specific Data Center location.
This is an industry issue that we expect impacts all Tier 1
and Tier 2 hardware vendors to an equal or greater degree.
Likewise, the move from raised-floor to flat-floor Data
Centers is likely not going to mitigate future vibration in
the Data Center due to new sources of vibration such as
overhead chillers and air handling units. Based on previous
industry studies, 2.5” SAS drives are generally less
sensitive to vibration than 3.5" SATA drives. As such, it is
expected that storage arrays with 3.5” SATA drives would
show even greater performance improvement with AVR than what
was demonstrated here.
Our Value Proposition
ROI Model
