Dawning Information Industry Co Ltd Sugon Storage Cloud

Dawning Information Industry Co. , Ltd. Sugon Storage Cloud Storage Moscow, 12/2015

Data Explosive growth 2 Large capacity storage system Data management Unified management of physical resources Storage system capacity and performance dynamic extension Data security þIDC statistics, the world produces 15 PB of data every day, the amount of information data every 18 months to double þIDC statistics, the world's information data to more than 50% annual growth rate of rapid expansion

Challenge of unstructured data growth Most of the way in the file exists 支票 Demand for data sharing The throughput of the storage system is higher. 图像 电子邮件和附件 The growth of data is often TB level File level storage technology 3 手册 即时消息 Structured Data. Base 电子报表 PDF Unstructured XML 合同 演示稿 Web 页 记录 富媒体 音频和视频 发票 文档

Challenge of increasing the total capacity of storage PRC Storage Market PB forecast, 2011 -2015 PB 3000. 0 83. 4% Storage PB 2500. 0 2000. 0 90. 0% 80. 0% YOY Growth % 70. 0% 66. 4% 60. 0% 56. 7% 50. 0% 1500. 0 32. 9% 1000. 0 500. 0 4 243. 7 381. 9 2008 2009 700. 5 33. 7% 930. 9 1244. 9 30. 1% 1619. 2 2055. 6 27. 0% 2552. 9 40. 0% 30. 0% 24. 2% 20. 0% 10. 0% 2010 2011 2012 The management of storage system is very complex. Limited capacity of a single storage system Storage system, the space utilization rate is low Costs continue to rise, a linear growth trend þIDC predicts that by 2015, the total capacity of the storage device market will reach 2. 5 EB, the compound annual growth rate of about 24. 2% þIDC predicts that by 2015, the total sales of storage equipment will reach 2. 2 B$, the compound annual growth rate of about 11% US $M 2500. 0 2013 2014 2015 PRC Storage Market Customer Revenue Forecast, Customer Revenue YOY Growth % 2011 -2015 16. 9% 15. 7% 13. 8% 14. 2% 12. 5% 1500. 0 1000. 0 500. 0 1, 164. 6 1, 361. 3 1, 575. 3 1, 799. 4 2, 024. 4 20% 15% 11. 0% 2, 247. 6 10% 5% 0. 0 0% 2010 2011 2012 2013 2014 2015

Volume/Performance increase rapidly Disk capacity of small, fast data recovery, using Raid 5/6 data security is good challenged by performance and reliability. 5 2003 160 GB 2007 1 TB 2011 3 TB Single disc capacity rapid increase Disk performance is quite limited Disk capacity becomes large, Raid reconstruction time is long Second disk failure probability increase

Primary Design Principle 6 A large-scale storage system, which is constructed with high performance, high availability and low TCO, is constructed with relatively inexpensive industrial standard units. Failure as the Network Servers main component Connection of the normal consideration Using clustering architecture, modular design Hard Disks principles, and building a global unified large -scale file sharing system

Storage Requirement storage capacity • 15 TB per 7 days for 5 K Camera. • 30 TB per 7 days for 10 K Camera. • Assume that the data storage for a month 10 K Camera, A total capacity demand = 30 TB*4*1. 5=180 TB , data redundancy is 4+2: 1. Storage throughput • Each camera is 300 Kbps, 10 K road data transmission bandwidth=300 Kbps*10 K= 367 MB/s. Storage Networking • 10 Gb/1 Gb Ethernet. • Support load balancing and redundancy. 7

Solution Configuration(5 K Cams) 8 No. Name description Qty 1 Server CPU: E 5 -2640 v 3 2. 6 G 20 M 8 C*2 MEM: 32 GB Cache network interface: 1 Gb. E/10 Gb. E Power Supply: 1+1 redundant power DISK: 2 T SAS*2 10 2 o. Para(Metadata Node) CPU: 2*Intel Xeon E 5 -2600 v 3 series processor MEM: 64 GB Cache network interface: 1 Gb. E/10 Gb. E Power Supply: 1+1 redundant power DISK: 240 GB SSD *4+300 GB SAS*4 RAID Controller: raid 0/1/5/6 2 3 o. Stor(Data Node) CPU: 2*Intel Xeon E 5 -2600 v 3 series processor MEM: 32 GB Cache network interface: 1 Gb. E/10 Gb. E Power Supply: 1+1 redundant power DISK: 300 GB SAS *2+4 TB SATA *22(provide 88 TB capacity) SAS Controller: 12. 0 Gb/s SAS , RAID 0/1/10 4 Description: 1. Storage naked capacity = 88 TB * 4 = 352 TB, effective capacity = 352 TB * 0. 8 = 235 TB 2. allow any one disk/one data node failure

Solution Configuration(10 K Cams) 9 No. Name description Qty 1 Server CPU: E 5 -2640 v 3 2. 6 G 20 M 8 C*2 MEM: 32 GB Cache network interface: 1 Gb. E/10 Gb. E Power Supply: 1+1 redundant power DISK: 2 T SAS*2 20 2 o. Para(Metadata Node) CPU: 2*Intel Xeon E 5 -2600 v 3 series processor MEM: 64 GB Cache network interface: 1 Gb. E/10 Gb. E Power Supply: 1+1 redundant power DISK: 240 GB SSD *4+300 GB SAS*4 RAID Controller: raid 0/1/5/6 4 3 o. Stor(Data Node) CPU: 2*Intel Xeon E 5 -2600 v 3 series processor MEM: 32 GB Cache network interface: 1 Gb. E/10 Gb. E Power Supply: 1+1 redundant power DISK: 300 GB SAS *2+4 TB SATA *22(provide 88 TB capacity) SAS Controller: 12. 0 Gb/s SAS , RAID 0/1/10 8 Description: 1. Storage naked capacity = 88 TB * 8 = 704 TB, effective capacity = 704 TB * 0. 8 = 563 TB 2. allow any one disk/one data node failure

Product form Whole cabinet Storage System modules • Adopted the modular design, by the cabinet frame, warehouse, computing nodes, power supply, heat dissipation and management module. • Support 40 U storage node space and 4 U exchange equipment space, the largest 20 storage node can be configured(1. 5 PB). 10

Storage Node 11 Storage Node: n 2 U Storage Node n No power/fan design n reduce the point of failure Processor 2*Intel Xeon E 5 -2600 v 3 series processor Memory SAS control Disk VGA High speed network Gigabit networks System Management 8 memory slot, DDR 3 1866/1600/1333 MHz 6. 0 Gb/s SAS, RAID 0/1/10 2 *2. 5” SAS/SATA/SSD +22* 3. 5” SAS/SATA Integrated graphics controller PCI-E 3. 0 x 16 slot， Support 1*56 Gbps FDR or 2*10 GE 2*1 GBE Onboard BMC management chip, IPMI 2. 0 standard management functions

Solution Strengths 12 expansibility Easy maintainable Higher density Energy conservation Higher performance Lower Costs low noise Para. Stor help user to build EB level data sharing platform！

System Framework 13 1. Management Controller - Provide two interfaces ( Command line and GUI); - Embedded management system to monitor the whole system. 1 2 2. Index Controller - Manage all metadata and namespace - Cluster architecture, Active-Active method Para. Stor 200 3 4 3. Data Controller - Provide data storage space - Support automatic data recovery 4. Client Drive - Provide POSIX file access interface - Support Linux/Windows clients

System Architecture 14 client concurrent read-write data migration data controller …… data controller data migration data controller …… MGR data controller concurrent read-write metadata read-write Index controller management network data controller archive

Client Product Form 15 Clients (Linux/Windows) Data read-write Data controller Request data Get metadata information Index controller

NAS Cluster Product Form Linux NFS Data Controller 16 Desktop CIFS HTTPS NAS Cluster Manage System Global Shared Cache Data Controller Virtualization FTP Data Controller p Global Shared Cache: Single NAS node failure cannot lose the cached data. p Apply to virtualization application, such as VMWare, Citrix.

Product Features 17 High performance Large capacity High reliability Para. Stor 200 High usability High scalability Low TCO

Large Capacity 18 Para. Stor 200 Index controller cluster Data controller cluster Files at a 10 billion level EB level storage capacity

Ultra-high IO Performance 19 Parallel cluster architecture design to satisfy high concurrent IO requirement 2. 5 GBps Aggregated bandwidth, which is the sum of data controller bandwidth, increases linearly with capacity Full active index cluster improves the processing ability of massive small files Strip optimization provides high single stream IO bandwidth

High reliability——Node fault tolerance 20 /home/appl/data/web/important_big_spreadsheet. xls /home/appl/data/web/big_architecture_drawing. ppt /appl /home/appl/data/web/unstructured_big_video. mpg /data Logical layer /web Index controller 0 Switch 1 1 2 3 A B Ca Db Ec Fd 4 Ge 5 Hf I J K L 6 Switch 2 7 8 9 10 11 g h i j k l Physical layer 0 1 2 A B C

High utilization data protection solution ：N+M D 0 D 1 21 Node 2 Node 3 Node 4 Node 5 D 0 D 1 P 0 D 2 D 3 D 4 D 5 D 6 P 1 D 7 D 2 D 3 D 4 D 5 D 6 D 7 N+M: B, N represents the number of data blocks, M represents the number of fault-tolerance disks , B represents the number of fault-tolerance nodes. N+M is a higher utilization solution on the premise of data reliability!

High scalability 22 LAN Parallel data I/O After expansion Data controller 1 2 1 2 3 4 3 4 Data controller Initial state Data controller 1 2 1 2 3 4 3 4 Data controller Add Data Controller Data migration automatically

Type of Interface Access interfaces • • Private Linux, Windows kernel access interfaces Standard NFS, CIFS interfaces POSIX API Map. Reduce programming interface REST programming interface SOAP programming interface SNMP interface Network interface • 20 Gb/40 Gb/56 Gb IB • 10 Gb/1 Gb Ethernet • Support load balancing and redundancy 23

Rich Access Interface 24 Internet application server FTP storage servers ●●● NFS, CIFS, REST …… I/O node 应用节点 VFS o. App SNMP I/O node 应用节点 VFS o. App Para. Stor Management and monitoring I/O node 应用节点 ●●● VFS o. App

Usability 25 Para. Stor web management interface Include Monitoring systems 、 System management 、 Advanced management Easy to management and monitoring storage system software and hardware resources.

Low TCO Low purchase costs 26 • Adopt X 86 architecture hardware • Software and hardware integration storage system Low management costs • Automated troubleshooting • Administrators rarely involved • 5 years after-sale services for both hardware and software Buy on demand • Capacity and performance can be dynamically expanded • No need to spend much one-time fund

Typical Application Field HPC þClimate þOil Field þGene Research þMaterial Internet þOnline Video þOnline Music þSocial Network þOnline Store 27 Radio/TV þIPTV þVideo Editing þRender Farm þConsumer Behaviors Cloud Storage þOnline Storage Space þOnline Storage Rent þCloud Backup & DR þHuge files Sharing

Changchun Traffic Police � 流量统计 28 � 逆行 Scale： 6*o. Stor � 高速行人 � 抛撒物检测 Para. Stor Applicants： n Traffic video data storage n Law enforcement recorder data n Virtualization storage platform

THANKS ADDRESS： Sugon Building, No. 36 Zhongguancun Software Park, No. 8 Dongbeiwang West Road Haidian District, 100094, Beijing, P. R. China TELEPHONE： 86 -10 -56308000 Weibo：http: //weibo. com/zksugon WWW ：http: //www. sugon. com