此簡報為 Will 保哥 於 2015/6/25 (四) 接受 SQL PASS Taiwan 邀請演講的內容。
現場錄影: http://www.microsoftvirtualacademy.com/training-courses/sql-server-realase-management?mtag=MVP4015686
[ Will 保哥的部落格 - The Will Will Web ]
http://blog.miniasp.com
[ Will 保哥的技術交流中心 ] (Facebook 粉絲專頁)
https://www.facebook.com/will.fans
[ Will 保哥的噗浪 ]
http://www.plurk.com/willh/invite
[ Will 保哥的推特 ]
https://twitter.com/Will_Huang
[ Will 保哥的 G+ 頁面 ]
http://gplus.to/willh
本簡報是 Will 保哥在 2016/6/24 於 CTJS 中台灣 JavaScript Conference 的演講簡報
[ 相關連結 ]
本次演講的 Live Demo 原始碼
https://github.com/doggy8088/ctjs2016-ng2demo
The Will Will Web記載著 Will 在網路世界的學習心得與技術分享
http://blog.miniasp.com/
Will 保哥的技術交流中心 (臉書粉絲專頁)
http://www.facebook.com/will.fans
Will 保哥的噗浪
http://www.plurk.com/willh/invite
Will 保哥的推特
https://twitter.com/Will_Huang
This document discusses MediaV's implementation of reliable advertising systems using Docker and Mesos. It describes how the system handles over 10 billion impressions with huge computing needs through containerization. It addresses common problems encountered like debugging, network performance, storage, service discovery, scheduling, and data loading. Frameworks like Marathon and Chronos are used on Mesos for orchestration and batch jobs. Health checks, port resources, and Dockerfile reviews are important practices.
HKIX Upgrade to 100Gbps-Based Two-Tier ArchitectureMichael Zhang
HKIX upgraded to a two-tier 100Gbps architecture to support continued traffic growth and new connections. This involved migrating to a new facility with more ports and space for 100GbE interfaces. The upgrade helps ensure HKIX can reliably serve as a critical internet infrastructure in Hong Kong and support keeping intra-Asia traffic within the region.
此簡報為 Will 保哥 於 2015/6/25 (四) 接受 SQL PASS Taiwan 邀請演講的內容。
現場錄影: http://www.microsoftvirtualacademy.com/training-courses/sql-server-realase-management?mtag=MVP4015686
[ Will 保哥的部落格 - The Will Will Web ]
http://blog.miniasp.com
[ Will 保哥的技術交流中心 ] (Facebook 粉絲專頁)
https://www.facebook.com/will.fans
[ Will 保哥的噗浪 ]
http://www.plurk.com/willh/invite
[ Will 保哥的推特 ]
https://twitter.com/Will_Huang
[ Will 保哥的 G+ 頁面 ]
http://gplus.to/willh
本簡報是 Will 保哥在 2016/6/24 於 CTJS 中台灣 JavaScript Conference 的演講簡報
[ 相關連結 ]
本次演講的 Live Demo 原始碼
https://github.com/doggy8088/ctjs2016-ng2demo
The Will Will Web記載著 Will 在網路世界的學習心得與技術分享
http://blog.miniasp.com/
Will 保哥的技術交流中心 (臉書粉絲專頁)
http://www.facebook.com/will.fans
Will 保哥的噗浪
http://www.plurk.com/willh/invite
Will 保哥的推特
https://twitter.com/Will_Huang
This document discusses MediaV's implementation of reliable advertising systems using Docker and Mesos. It describes how the system handles over 10 billion impressions with huge computing needs through containerization. It addresses common problems encountered like debugging, network performance, storage, service discovery, scheduling, and data loading. Frameworks like Marathon and Chronos are used on Mesos for orchestration and batch jobs. Health checks, port resources, and Dockerfile reviews are important practices.
HKIX Upgrade to 100Gbps-Based Two-Tier ArchitectureMichael Zhang
HKIX upgraded to a two-tier 100Gbps architecture to support continued traffic growth and new connections. This involved migrating to a new facility with more ports and space for 100GbE interfaces. The upgrade helps ensure HKIX can reliably serve as a critical internet infrastructure in Hong Kong and support keeping intra-Asia traffic within the region.
Fastsocket is a software that improves the scalability and performance of socket-based applications on multicore systems. It addresses kernel inefficiencies like synchronization overhead that consume over 90% of CPU cycles. Fastsocket introduces techniques like receive flow delivery, local listen/established tables, and a fastsocket-aware VFS to partition resources and process connections locally on each CPU core. In production at SINA, Fastsocket improved HTTP load balancing throughput by 45% on a 16-core system. Future work aims to further optimize performance through techniques like improved interrupt handling and system call batching.
Spark SQL is a module for structured data processing on Spark. It integrates relational processing with Spark's functional programming API and allows SQL queries to be executed over data sources via the Spark execution engine. Spark SQL includes components like a SQL parser, a Catalyst optimizer, and Spark execution engines for queries. It supports HiveQL queries, SQL queries, and APIs in Scala, Java, and Python.
CUDA 6.0 provides performance improvements and new features for several CUDA libraries and tools. Key updates include up to 2x faster kernel launches, new cuFFT and cuBLAS features for multi-GPU support, up to 700 GFLOPS performance from cuFFT, over 3 TFLOPS from cuBLAS, and 5x faster cuSPARSE performance compared to MKL. New features also improve the performance of cuRAND, NPP, and Thrust.
The document discusses network integration considerations for Hadoop data centers. It addresses traffic types, job patterns, network attributes, architecture, availability, capacity, flexibility, management and visibility. It provides examples of buffer usage on switches and recommendations for dual 1GbE or 10GbE NIC configuration for Hadoop servers.
Hadoop Hardware @Twitter: Size does matter.Michael Zhang
This document discusses Twitter's experience scaling their Hadoop clusters and evaluating different hardware configurations. It describes how Twitter developed the "Twitter Hadoop Server" (THS) to optimize for different workloads like backups, processing, and cold storage. The THS uses single-socket servers with fewer disks optimized for cost and density. Testing showed the THS outperformed baseline dual-socket servers on processing benchmarks while providing higher storage density and lower costs per node. The document concludes that for large clusters, specialized hardware can improve performance and efficiency compared to a single size-fits-all approach.
Q con shanghai2013-[ben lavender]-[long-distance relationships with robots]Michael Zhang
The document discusses long distance relationships with robots and maintaining open communication. It describes how the author started their relationship with a robot named Hubot by installing it on GitHub and customizing it with scripts. The author emphasizes values like passion, dedication, responsibility and communication to make the relationship work despite distances of thousands of kilometers. It provides tips for deploying robots, monitoring their operations, and using them to help with work communications and status updates.
Q con shanghai2013-[jains krums]-[real-time-delivery-archiecture]Michael Zhang
The document describes Twitter's real-time delivery architecture. It aims to evolve beyond being just a web stack by isolating responsibilities into components like routing, presentation, logic, storage and retrieval. The architecture uses various services and technologies like the Write API, Ingester, Blender, Hadoop, Redis, and Earlybird to ingest tweets in real-time, cache timelines and search indexes, and deliver tweets to users via pull and push methods. The goal is to improve site speed, reliability, and developer innovation speed.
Q con shanghai2013-[黄舒泉]-[intel it openstack practice]Michael Zhang
This document summarizes an Intel IT presentation on their OpenStack practice. It discusses Intel's contributions to OpenStack projects, their converged OpenStack and IT platform, and their solutions for continuous delivery, deployment, Tempest testing automation, and improving the speed of OpenStack snapshots.
Q con shanghai2013-罗婷-performance methodologyMichael Zhang
This document summarizes a presentation on performance methodology at Salesforce given at QCon Beijing 2014. It discusses:
- The importance of performance for user experience, decreasing costs, and serving more customers.
- Salesforce's dedicated performance team is organized by areas like UI, mobile, platforms, and infrastructure.
- Key performance metrics include response time, throughput, CPU/memory utilization, and database metrics.
- Performance is tested proactively via feature and regression tests, and passively via production analysis. Automated testing uses tools like JMeter and internal frameworks. Profilers like Yourkit and HeapAudit help identify causes.