smashing big data with aha hardware gzip
TRANSCRIPT
Smashing Big Data with AHA GZIP Hardware
www.aha.com
1
01 011110010010000101110001001101100010100010010100011000011010001110 1010000110100011001001010001100011100100100001010010000101110001
10 101000011010001100100101000110001110010010000101001000010111000101 0111100100100001011100010011011000101000100101000110000110100011
111011
1110101100
01101101101101101101110110
1110
0110
111011
1110101100
01101101101101101101110110
1110
01
10111011
111010110001101101101101101101110110
1110
0110
111011
111010110001101101101101101101110110
1110
01
10
101101101101
1110110010111110110010111111 11 10 1100 10 1111 10 1100 10 11
1111
111011 001011111011 0010111111111011 001011111011 001011 11 11
01111001001000010111000100110110
00101000100101000110000110100011
1011011011011011
101101101101101111
01110110111001
1011011011011101 11
01110110111001
1011011011011101 11
01110110111001
1011011011011101
01 100 100 101 100 1001 110
Sales and Contact Information• Website
- www.aha.com• Sales Contact
- [email protected]• White Paper to Presentation
- http://www.aha.com/Uploads/GZIP_Benefits_Whitepaper9.pdf
2
Introduction• Purpose
- The purpose of this brief is illustrate the benefits hardware GZIP compression on a 10G Ethernet link using the AHA 372.
• Outline- AHA Product Introduction- 10Gbps Web Server Experiment and Results- Energy Break Even Point of AHA372
3
Who we are…• AHA Products Group
- Business unit of Comtech EF DATA- Advanced IP Development Center
• Brief History- Established 1988 on Forward Error Correction- 1992 Created first CAM in Data Compression- 2006 Created First GZIP Compression IC- 2009 Created Fastest GZIP Compression IC
• Expertise- High Performance Data Compression - Forward Error Correction- Hardware and Software Development
4
AHA37XPCIe Family Features• Interface
- PCIe 3.0x8 interface• Algorithms
- GZIP, ZLIB, LZS• Board by Compression Speed
- AHA371 – 10 Gbps- AHA372 – 20 Gbps- AHA374 – 40 Gbps - AHA378 – 80 Gbps
5
AHA37[1,2] – 10-20 Gbps
AHA37[4-8] – 40-80 Gbps
Experiment Setup
• Client Emulator sends HTTP requests- Requests maximize throughput of 10 G link
• Webserver responds with GZIP’d page- Requests are run continiously
6
Supermicro X8DTL-6F Motherboard Intel Xeon E5620 CPU 3 X 2GB Kingston KVR13R9S8K3/6I Seagate Barracuda ST3160815AS Intel 82598EB 10-Gigabit AT2 Server Adapter
PowerEdge R720 Motherboard Intel Xeon E5-2643 3.30GHz (Total 8 cores) 24 X 4GB RDIMM, 1600MT/s RAM Western Digital WD800JD-75MSA2 Intel 82598EB 10-Gigabit AT2 Server Adapter
10 Gig-E
Page Responses• No GZIP• CPU GZIP• Hardware GZIP
Page RequestsClient Emulator Apache Web Server Power Meter
40 Gbps Compression: 2x AHA372 20Gbps GZIP Accelerators
Watts up? PRO ES Accuracy: +/- 1.5%
120 VAC
Scenarios and Measurements• Observed Metrics
- Effective throughput of 10Gb Link Better compression ratio, more effective throughput Experiment average between 1:2, 1:3, compression ratios
- CPU Utilization - Power/ Throughput (Watt/Gbps) or Energy/Bit (Joule/Gb)
• Compression Scenarios Ran- No Compression- CPU Compression : mod_deflate - AHA Compression : 2 x AHA 372 20Gbps
7
Experiment Results8
9 Gbps
15%
16 J/Gb 6.8 J/Gb
59%
35 Gbps2 Gbps
100 %
116 J/Gb
Throughput
CPU Load
Energy Efficiency
No GZIP CPU GZIP AHA GZIP
CPUs are Inefficient at Performing GZIP• CPU is Crippled
- Serving web pages drops from 9 Gbps to 2Gbps.
• CPU is Inefficient- 8 Core CPU at 100% load and
consumes 7x more energy.
9
9 Gbps
15%
16 J /Gb 6.8 J /Gb
59%
35 Gbps2 Gbps
100 %
116 J /Gb
Throughput
CPU Load
Energy Efficiency
No GZIP CPU GZIP GZIP
CPUs are not optimal for performing GZIP
GZIP Hardware Removes CPU Bottlenecks• 18x Throughput over CPU
At full load with 8 cores, CPU GZIP is only capable of 6% of the throughput of GZIP hardware.
• Available CPU cycles Not performing compression CPU can
perform tasks it is most efficient at.
• 17x energy efficiency- Less energy cost
10
9 Gbps
15%
16 J /Gb 6.8 J /Gb
59%
35 Gbps2 Gbps
100 %
116 J /Gb
Throughput
CPU Load
Energy Efficiency
No GZIP CPU GZIP GZIP
9 Gbps
15%
16 J /Gb 6.8 J /Gb
59%
35 Gbps2 Gbps
100 %
116 J /Gb
Throughput
CPU Load
Energy Efficiency
No GZIP CPU GZIP GZIP
GZIP Hardware is order of magnitudes more efficient at GZIP than CPUs.
Energy Break Even Point (eBEP)• Assumptions
- Card Price at Low Volumes- Continuous Data Stream- Power Usage Effectiveness 1.7- Normalized Linear Performance
Adjustment for Power Consumption • Hardware Pays Off FAST!
- 22 Days at Lowest Electricity Cost
11
Cost of Electricitycents (kW/h)
eBEP of GZIP Hardware Vs. CPU GZIP
15 22 days20 17 days25 14 days30 11 days35 10 days
GZIP Hardware Reduces OpEx and CapEx by creating optimal performing systems using less hardware (CPU).
AHA vs Other Compression12
bzip2, 47, 3.5:1
zlib, 70, 2.8:1
lzo, 640, 1.9:1
aha-gzip,3.33Gbps/Chan
, 2.7:1
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
1 10 100 1000 10000
Com
pres
sion
Ratio
to
1
Compresion Speed in Mbps
Comparative Compression Performance(256 MB Wikipedia Corpus)
GZIP Hardware still orders of magnitude faster than the fastest CPU compression and with good compression
ratios.
Integration• Plug and Play
- Driver installs in minutes- ZLIB library replacements- Apache Webserver API- JAVA API
• Full Engineering Support- Development engineering support for all appliance integration.
13
Sales and Contact Information• Website
- www.aha.com• Sales Contact
- [email protected]• White Paper to Presentation
- http://www.aha.com/Uploads/GZIP_Benefits_Whitepaper9.pdf
14