Jeonghan Kim October 24, 2013 SW R&D Center, Device Solutions Samsung Electronics Co.
Legal Disclaimer This presentation is intended to provide information concerning Flash storage industry. We do our best to make sure that information presented is accurate and fully up-to-date. However, the presentation may be subject to technical inaccuracies, information that is not up-to-date or typographical errors. As a consequence, Samsung does not in any way guarantee the accuracy or completeness of information provided on this presentation. The information in this presentation or accompanying oral statements may include forward-looking statements. These forward-looking statements include all matters that are not historical facts, statements regarding the Samsung Electronics' intentions, beliefs or current expectations concerning, among other things, market prospects, growth, strategies, and the industry in which Samsung operates. By their nature, forward-looking statements involve risks and uncertainties, because they relate to events and depend on circumstances that may or may not occur in the future. Samsung cautions you that forward looking statements are not guarantees of future performance and that the actual developments of Samsung, the market, or industry in which Samsung operates may differ materially from those made or suggested by the forward-looking statements contained in this presentation or in the accompanying oral statements. In addition, even if the information contained herein or the oral statements are shown to be accurate, those developments may not be indicative developments in future periods. 2
Contents IT Trend for Flash Storage Flash Storage Applications Hybrid Array Server Flash Flash Appliance Flash Storage 활용효과 Cloud Computing Virtualization Big Data Virtualized Hadoop Change in SSD 산업인프라강화제안 3
IT Trend for Flash Storage - 3 rd Platform 시대 모바일 & Social 확산에따른 data 증가로 3rd Platform 시대의도래 사용자와애플리케이션의폭발 (explosion) 빅데이터, 클라우드컴퓨팅중요도증가 4
IT Trend for Flash Storage - Data Center Evolution 데이터센터가상화는지속적으로진화中 Multi-datacenter 구성시의 overhead 로인한 IOPS/TB 의저하이슈발생증가 Multi-datacenter 가상화 App App App App 단일 Datacenter 가상화 App App Server 가상화 App App App App App App Performance CPU CPU CPU Mem Mem Mem Disk Disk Disk Performance Gap IOPs/Spindle IOPs/TB Time Source: stec, Data Center Epochs 5
IT Trend for Flash Storage - Pattern of Workload VM 기반한 Cloud 컴퓨팅환경에서의 Random data access pattern 증가 Traditional Workloads Cloud Server Workloads Sequential throughput & IOPs important Utilization can be low More VMs drive random Random IOPs critical 24h x 7days Operation Sequential and Random Accesses More highly randomized accesses Source: HGST 6
IT Trend for Flash Storage - Storage Usage By 2015, 기업데이터의 80% 이상이 cold data Cloud storage 로저장및이동 10% 미만의 hot data SSD 활용 전통적인 FCoE, SAS storage array 의축소에따른 Storage 양분화가속화진행중 Source: SNIA, NVMe: The NextGen Interface for Solid State Storage FCoE : Fiber Channel over Ethernet 7
IT Trend for Flash Storage - Interface & Market SAS-SSD 및 PCIe-SSD I/F 기반제품이전체매출견인예상 (6B$/2016) SATA SSD 는 commodity server 기반의클라우드, 블레이드서버, 워크스테이션 시장에서꾸준한사용이예견 Source: IDC 8
IT Trend for Flash Storage - Flash Solution 사업역량강화 주요 Major 업체들의 flash 기반 startup 기업인수활발 SW & HW Startup 인수를통한 Solution 강화정책수행진행중 SSD Cache Software 2013.07 (Storage SW) 2013.07 Storage Hardware 2013.07 (Flash controller) 9
IT Trend for Flash Storage - Price 서버용 HDD vs. SSD 가격비교 2017 년 15K 고성능 SAS HDD 와비교시 SSD 의용량대비가격추월예상 HDD 와 SSD 의혼합사용형태인계층화 (tiering) 기능이용 All HDD 구성과비교시성능대비비용은이미추월 $/GB 15K RPM HDD Flash SSD 7.2K RPM HDD 2013 2014 2015 2016 2017 2018 2019 2020 Source: DELL 10
Contents IT Trend for Flash Storage Flash Storage Applications Hybrid Array Server Flash Flash Appliance Flash Storage 활용효과 Cloud Computing Virtualization Big Data Virtualized Hadoop Change in SSD 산업인프라강화제안 11
Flash Storage Applications - Overview 기업향 flash storage 기술분류 Flash Storage 종류 Hybrid Array Server Flash Flash Appliance 주요목적 전통적인 Spinning Disk 의대체및보완 Server-Centric 솔루션 주요워크로드에특화된 Array 솔루션 특징 SAS, SATA 용량과성능의절충안 기존인프라변경최소화 Auto Tiering PCIe 어플리케이션특화작업 usec latency, higher IOPs 고비용 / 저용량 DIMM/Proprietary 범용워크로드최적화 Tier-1 의대체 HA, Hot Swap 제약 Startup 기업다수포진 형태 관련업체 12
Flash Storage Applications - Hybrid Array Hybrid Array - 성능과가격의균형 SATA, SAS SSD 의 $/GB 극대화 auto tiering algorithm 의최적화 Data 의 I/O access pattern 파악 SSD/HDD 간 data 이동자동화 Tier0 PCIe SSD SAS SSD Flash in the server flash appliance Tier 1 Tier1 Tier2 Write intensive SSD 15K or 10K Read intensive SSD 15K or 10K Shares storage infrastructure $$$ RAID 10 High Endurance Enterprise SAS SSD Tier 2 $$ RAID 5 RAID 5 Tier 3 $ RAID 6 Mostly 7.2K and 10K Read Intensive Enterprise SAS SSD RAID 6 Tier3 Some 15K 7.2K SAS Source: DELL 13
Flash Storage Applications - Server Flash Server Flash - ultra low latency, ultra high IOPs per device PCIe 기반의 flash SSD 를이용하여 10 억 IOPs 달성 8 서버 & 64 PCIe SSDs 새로운 SW layer 의도입 CPU 와 PCIe SSD 간 data 의직접이동 100 만 IOPs, 20usec latency / PCIe SSD 10 억 Write IOPS 를위한서버구성例 Source: Wikibon, 2012 14
Flash Storage Applications - Flash Appliance Flash Appliance 성능보다는집적도향상및 GB 당 Cost 감소목표 SW 기반의중복제거및압축기능적용 $/GB 를 ¼ 수준으로감소 제품 특징 Max 44TB (19/20nm), 500TB (16nm) per 1U Skyera skyhawk $1.99/GB compressed, $0.49/GB comp and dedup 500 IOPs over 15 interchangeable 16Gb Fibre Channel or 10Gb Ethernet ports Violin Memory 6264 64TB (6264) 500 to 750 IOPs Nimbus 4 th Flash Appliance first to use 1xnm NAND HALO SW $0.78/GB (1/3 rd cost of comparable hybrid storage) Source: www.thessdguy.com, 2012 15
Contents IT Trend for Flash Storage Flash Storage Applications Hybrid Array Server Flash Flash Appliance Flash Storage 활용효과 Cloud Computing Virtualization Big Data Virtualized Hadoop Change in SSD 산업인프라강화제안 16
Flash Storage 활용효과 - Overview Flash Storage 적용 Domain 분류 Cloud Computing All flash based appliance in Cloud Computing QoS PCIe SSD in VDI Environment Cache Realtime NoSQL Low Latency Virtualized Hadoop Virtualization Overhead Virtualization Big Data 17
Flash Storage 활용효과 - QoS Cloud Computing 사용자관점에서의안정적인 QoS 환경제공가능 Higher IOPs 및 QoS 성능에촛점 Typical Multi-Tenant 환경 SSD 기반 Multi-Tenant 환경 Throughput Throughput Time Time Source: SolidFire 18
Flash Storage 활용효과 - 가상화통한 I/O 성능개선 VDI workload 고려한 PCIe SSD 기반가상화솔루션 유사한 OS image/application 사용을반복하는 I/O pattern PCIe SSD 를이용한 deduplication/cache 효과극대화 Read & Write Operations / sec PCIe SSD 적용통한 IO 성능개선 Source: NetApp 19
Flash Storage 활용효과 - Real-Time Big Data 처리 DRAM & SSD 기반의 hybrid NoSQL 등장 real time big data 분석목적 DRAM & SSD 의동시활용을통한 latency 효율향상 I/O 향상에따른실시간효과 DRAM DB index 저장, SSD large block 단위의 flash write 수행 Source: Aerospike 20
Execution Time (sec) Flash Storage 활용효과 - 가상화 + Big Data 수직최적화 SSD 를통한가상화기반의 Hadoop 클러스터성능개선 가상화 overhead + I/O throughput = 전체수행시간단축 Hadoop, VM, Hypervisor 각단계에서의수직최적화를통한성능개선진행中 Virtualized Hadoop Architecture 3,500 nhdd HDD vs. SSD Hadoop 벤치마크성능비교 VM Hypervisor Hadoop 3,000 2,500 vhdd nssd vssd 2,000 1,500 1,000 500 0 SORT WORDCOUNT TERASORT DFSIOE-READ DFSIOE-WRITE The system architecture of virtualized Hadoop cluster (Xen). The comparison of execution times of benchmarks between with baremetal and virtualized cluster of HDD and SSD. 21
Flash Storage 활용효과 - Change in SSD (1) 3D V-NAND cell 의 reliability 증가 기업향 SSD 시장적용확대 22
Flash Storage 활용효과 - Change in SSD (2) V-NAND 기반 SSD 의장점 23
Flash Storage 활용효과 - Change in SSD (3) 3D V-NAND SSD 시장전망 24
Contents IT Trend for SSD SSD Applications Hybrid Array Server Flash Flash Appliance SSD 활용효과 Cloud Computing Virtualization Big Data Virtualized Hadoop Change in SSD 산업인프라강화제안 25
Needs for Soft Skills "It's in Apple's DNA that technology alone is not enough... it's technology married with liberal arts, married with the humanities, that yields us the result that makes our heart sing. - Steve Jobs, Seattle Times, March 3, 2011 26
Needs for Soft Skills Future Work Skills 2020 Rise of smart machines & systems Extreme longevity Computational world New media ecology Superstructed organizations Globally connected world Under 6 drivers of change Sense-making Social intelligence Novel & adaptive thinking Cross-cultural competency Computational thinking New media literacy Transdisciplinarity Design mindset Cognitive load management Virtual collaboration - Institute for the Future, 2011 Close to Liberal Arts! 27
What is the Academy Doing? - Source : Gartner 28
SW 산업확대를위한인문계활용 출처 : 한국경제, 2013.03.14 가치창조를수면으로올리기위한노력 + SW 는국가기간산업 29
From Computer Science to Data Science Skills for Data scientist Soft Skills Technology Skills Communication Collaboration Leadership Creativity Discipline Passion Data Science needs better understanding of: User-computer interaction Collaborative behavior Workforce effectiveness Communication patterns Data scientists can be invaluable in generating insights, especially from "big data;" but their unique combination of technical and business skills, together with their heightened demand, makes them difficult to find or cultivate. - Source : Gartner Liberal Artist can become Data Scientist! 30
格의 R&D Open Source 관련정책 Files / sec Bandwidth (MB/s) Samsung has released F2FS for flash storage to Linux open-source group Wandering tree problem mitigated by NAT(Node Address Table) Cleaning O/H reduced by background cleaning, hot/cold separation, adaptive logging Can be configured by FTL-optimized parameters such 30 as mapping unit Compared with ext4 in FS benchmarks, almost sequential-like 20 random write performance 35 25 15 10 5 0 IOzone seq. Read seq. Write rand. Read rand. Write EXT4 30.753 17.066 5.06 4.15 F2FS 30.71 16.906 5.073 15.204 12000 10000 8000 Bonnie++ 6000 4000 < F2FS Index Structure > 2000 0 seq.create seq.stat seq.delete rand.create rand.stat rand.delete EXT4 692 1238 1370 663 1250 915 F2FS 631 7871 10832 620 7962 5992 31
Thank you! 32