《Power平台与x86对比分析报告》由会员分享，可在线阅读，更多相关《Power平台与x86对比分析报告（36页珍藏版）》请在金锄头文库上搜索。
1、,“一般用途” vs “量身定制”,Power平台与x86对比分析报告,1. 2. 3. 4. 3,最佳性能 更低的费用 更好的服务,有些事情Power可以做，但X86做不到 “一般用途” vs “量身定制” 最有价值,架构设计 最好的性价比 效率更高的密集业务 更好的RAS技术,4,POWER “量身定制”的架构,x86 设计初衷是针对多用途多业务的市场环境 -智能手机，电子图书馆，桌面服务器 设计目标: 通过增加处理器内核数量来提高服务器性 能 优势： 只受限于CPU数量,Power 设计初衷的“目的性” 很强 用于满足全新或现有的企业级业务负载 (Analytics, Big Data,
2、 OLTP, ERP, SAP, etc.) 设计目标: 通过提供单核处理器的性能而减少整体业务负载的开销 优势： 集成处理器微架构、虚拟机管理程序、操作系统、IBM中间件和IBM存储,5, ,The Ivy Bridge CPU 技术是精简版的Sandy Bridge CPU 技术，架构基本保持不变 Ivy Bridge isIntels 至强和安腾Tick-Tock计划 “tick” 部份 根据Intel发布周期而定： - 处于 “tick“ 周期内的处理器架构相比之前处理器架构只会有少部份提高 相对于现有的32纳米技术，22纳米技术主要改内了晶体管技术，而性能处于相同水平 -更多的晶体管
3、 - 更多的内核, 更多的CPU槽位 - 更多的L2及L3缓存 - 更多的 $ 问题: 1. 22纳米技术= 更高的性能? 2. 每台服务器核心越多 = 性能越好?,Ivy Bridge 是新一代的具有突破性变革的架构么？,RPE2* per Core,Ivy Bridge 架构下每核心性能并没有提高,6,2069,2049,0,500,1500 1000,2000,2500,2 槽位 HP 服务器 2283,Sandy Bridge-EP 2.9 GHz 16-c,Ivy Bridge-EP 2.7 GHz 24-c,Ivy Bridge-EX 2.8 GHz 30-c,Sandy Brid
4、ge EP 2.9 GHz 16 cores,The number shown is best in each category (sockets and number of cores) *Gartner RPE2 Details: ,RPE2 值来源于以 下六种测试结果,SAP SD Two-Tier,TPC-C, TPC-H, SPECjbb2006 and,two SPEC CPU2006 components,Ivy Bridge EP 2.7 GHz 24 cores,Ivy Bridge EX 2.8 GHz 30 cores,Core,Core,Core,Core,Power
5、Core,Power Core,4 Cores = 4 软件许可,2 Cores = 2 软件许可,不应该关心一台服务器有多少处理器核心 而应该关心每个核心产生的性能 每核心更高的性能=在完成相同数量的工作下，更少的软件许可 更少的软件许可费用 = 更低价格的工作量,注意: 多数软件及中间件是根据核心数量收费,7,POWER8 处理器是 量身定制 所以性能优异,8,Sandy Bridge EP,POWER 7+ Systems,POWER8,Clock rates SMT options Max Threads / sock Max L1 Data Cache Max L2 Cache Ma
6、x L3 Cache Max L4 Cache Memory Bandwidth,1.83.6GHz 1,2* 16 32KB 256 KB 20 MB 0 31.4-51.2 GB/s,1.7-3.7GHz 1, 2* 24 32KB* 256 KB 30 MB 0 42.6-59.7 GB/s,1.9-3.4 GHz 1, 2* 30 32KB* 256 KB 37.5 MB 0 68-85* GB/s,3.14.4 GHz 1, 2, 4 32 32KB 256 KB 80 MB 0 100 180 GB/sec,3.0-4.1 GHz 1, 2, 4, 8 96 64KB 512 KB
7、 96 MB 128 MB 230 - 410 GB/sec,* Intel calls this Hyper-Threading Technology (No HT and with HT) *32KB running in “Non-RAS mode” Only 16KB in RAS mode *85GB running in “Non-RAS mode” = dual-device error NOT supported,Ivy Bridge EX E7-88xx v2,Ivy Bridge EP E5-26xx v2,1. 2. 3. 4. 3,最佳性能 更低的费用 更好的服务,有些
8、事情Power可以做，但X86做不到 “一般用途” vs “量身定制” 最有价值,架构设计 最好的性价比 效率更高的密集业务 更好的RAS技术,AIX,$3.09,12,183,800,每秒用户交互数量,DB2,AIX,6 VMs,DB2,85,939,WebSphere on platform,Database off platform 2S/24 Core Ivy Bridge-EP (2.7 GHz) This is an IBM internal study designed to replicate a typical IBM customer workload usage in t
9、he marketplace. The results were obtained under laboratory conditions, and not in an actual customer environment. IBMs internal workload studies are not benchmark applications, nor are they based on any benchmark standard. As such, customer applications, differences in the stack deployed, and other
10、systems variations or testing conditions may produce different results and may vary based on actual configuration, applications, specific queries and other variables in a production environment. Prices, where applicable, are based on published US list prices for both IBM and competitor, and the cost
11、 calculation compares the cost per request for the 3yr life of the machine. 3 year total cost of acquisition comparisons are based on similar expected hardware, software, service VIOS using only 0.05 cores,L4 cache 激活,P8 - 24 cores throughput 183800 +11% 165064,针对于WEB应用负载，L4缓存提高了整体性能,L4 = 128MB 测试平台
12、: Pre-GA POWER8 Firmware level 1409B Measured using 8 LPARs ( *3 cores each) using SRIOV ports AIX 7.1 TL3 SP3, WAS 8.5.5.1 JDK7, FB 3.6,16,17,CPU Processing,IOPS (Input Output Operations per Second),CPU 闲置 I/O 繁忙,最好的OLTP性能要求对两种类型的资源进行优化 CPU 繁忙 I/O 闲置,高效率使用 CPU技术及缓 存技术,高效率使用SSD 硬盘或使用IBM FlashSystem存
13、储,System Cost 3 year TCA $3,835,481,System Cost 3 year TCA $2,094,115,Intel-based Processors Rack 32 database cores 2 Server Nodes,2.7x Faster 80% Lower cost,per Trade per second Exadata X3-2 Quarter Rack,基于POWER8平台的DB2 BLU性能是基于 Oracle DB on Exadata 的 2.7x 倍，并且成本降低了80% OLTP (Brokerage) Workload,19,3
14、 Storage Nodes Based on IBM internal tests comparing the IBM system with a comparably priced, comparably tuned competitor configuration (version available as of 01/01/2013) executing a materially identical online transaction processing workload in a controlled laboratory environment. Testsmeasured T
15、radeCompletion throughput rates to execute identical SQL query workloads. More Trade Completions is indicated by higher Trade Completions/second. Competitor configuration: Unit (usable uncompressed capacity = 9.5TB) including competitor recommended software options and features. Results may not be t
16、ypical and will vary based on actual workload, configuration, applications, queries and other variables in a production environment. Users of this document should verify the applicable data for their specific environment.,1,543 $1,357,Trade Completions per second (TPS) per TPS,561 $6,837,Trade Completions per second (TPS) per TPS,DB2 AESE v10.5 Power S824 with 24 cores AIX 7.1, 64-bit FlashSystem 840,1. 2. 3. 4. 3,最佳性能 更低的费用 更好的服务,有些事情Power可以做，但X86做不到 “一般用途” vs “量身定制”
