Flash High Level Customer Presentation
Flash High Level Customer Presentation
Flash High Level Customer Presentation
IBM® FlashSystem™:
The World’s Fastest Storage
Smarter
Computing
IT efficiency is essential to outpace
data growth
3
Top challenges for IT managers
With limited budgets, where should IT managers invest for the best overall return?
4
Target the bottleneck in the IT stack
Storage disk speed is the bottleneck that’s slowing everything else in the IT stack
5
Flash unlocks the storage bottleneck
7
Flash benefits
Boost performance/efficiency without re-architecting applications or the
storage infrastructure
85%
85% reduction
reduction 90%
90% reduction
reduction 25-100
25-100 µsµs latency
latency 80%
80% reduction
reduction
in
in batch
batch processing
processing In
In OLTP
OLTP times
times for
for deterministic
deterministic In
In energy
energy usage
usage
times
times performance
performance
8
Compare flash performance / efficiency
Compare the capabilities of ½ a rack of flash to a traditional 15K drive implementation
Performance
Flash can service up to 5.9M IOPS
Disk requires over 98 racks of 15K RPM drives
Nearly a 200:1 performance/capacity
advantage
Density
Flash provides 500TB of usable capacity
Disk requires almost 4.5 racks of 600GB drives
Nearly a 9:1 capacity/rack space advantage
Environmental
Flash uses <10 KVA
Disk drives would consume ~6,000 KVA
Nearly a 600:1 energy/IOPS advantage
9
Flash works now
10
iTalk celebrates the Year of the Rabbit with
a 6x flash performance boost
Business challenges • With one easy change, we vastly improved
Handle high call volume to China from performance.
the United States on Chinese New Year
• It is by far the most successful, highest-impact piece of
Solution hardware in our system.
Implement IBM’s flash to combine
performance, capacity, and reliability at • Probably the best investment of technical dollars that
a competitive cost we could have made!
Jeff Liegel, Chief Engineer, iTalk
11
Mercury Technology boosts cloud performance
IBM’s flash achieved reliable acceleration for
Mercury’s Oracle cloud applications
Business challenges
Eliminate storage latency and • IBM is the only company that has such a long history in this
achieve the higher performance technology.
needed to accelerate Oracle cloud apps
• The other vendors are newer players, none of whom have
Solution been working in the Oracle world.
In-server flash offered dedicated
storage at microsecond latencies Brian Day, Vice President of Sales at Mercury Technology
IBM’s flash reduced Mercury Technology’s batch processing times from 5 hours to 1—
a 5x improvement with just one change in the storage tier!
12
UFC knocks out production bottlenecks with
IBM flash
Business challenges • Flash has just completely eliminated [bottlenecks].
Extreme media streaming requirement Everything’s flowing smoothly.
Mike Saindon, Production Engineer
Solution
• Before we put the flash in, we couldn’t use our asset
Implement a high-impact, nondisruptive
RamSan solution which was installed in a management system to its maximum capacity and intent.
single day Christy King, Vice President of Digital, Technology and R&D
With flash, streaming batch jobs completed up to 70 percent faster and improved performance
and efficiency on other disk storage platforms by 40 percent!
13
13
Penn State slashes backup time by 80 percent
the TMS RamSan delivers vastly better IOPS
performance than two hundred 15K disk drives
Business challenges • We literally just turned it on and that’s
Accelerate nightly backups to all the performance tuning we did. It
meet windows just worked out of the box.
14
14
Banco Azteca integrates flash into its SAN for increased
performance in its credit processing business
Business challenges
Speed up application
processing without
architectural changes
Solution
Integrate flash seamlessly into
the existing infrastructure
~100µs ~100µs
Processing
CPU state
~5000 µs
Waiting
Time
1 I/O request
18
Flash’s microlatency drives CPU utilization/efficiency higher
1. Issue I/O request (~ 100 μs)
2. Wait for I/O to be serviced (~ 200 μs)
3. Process I/O (~ 100 μs)
• Time to process 1 I/O request = 100 μs + 200 μs + 100 μs = 400 μs Speed application
• CPU utilization = Wait time / Processing time = 200 / 400 = ~50%
response up to 12x
simply by changing
~100µs ~100µs
Processing
storage latency!
CPU state
~200 µs
Waiting
Time
1 I/O request
19
The real cost of latency
20
Let the IBM® FlashSystem™ family help you
1. Collect some easy-to-gather performance information about
your environment
• Host: IOstat, perfmon, sar, and so on
• DB: Oracle AWR, DB2 db2support, and so on
• Storage: Symmetrix WLA, NAR files, and so on