How the Post-PC era changed IT Ubuntu for next gen datacenters
2 likes1,330 views
The document discusses how the rise of ubiquitous computing devices in the "Post-PC Era" has impacted server infrastructure needs. It describes how there are now an incredible number of internet-connected devices generating many client requests, requiring scalable and distributed systems to handle big data and workloads like social networks and online gaming. It then introduces ARM processors as an option for power-efficient servers that could enable increased server density in datacenters and replace single-server setups in small businesses with low-cost "micro-servers". The rest of the document discusses Ubuntu's support for ARM servers and potential workloads like web and cloud applications that could be suitable targets.
How the Post-PC era changed IT Ubuntu for next gen datacenters
- 1. How the Post-PC era changed IT Ubuntu for next gen datacenters Giuseppe Paterno' EMEA OEM Technical Manager gpaterno@canonical.com www.canonical.com October 2011
- 2. The Post-PC Era Welcome to the Post-PC Era An era where the personal computing is moving from the desktop form factor to more ubiquitous computing such as netbooks, smartphones, Tablet, Smart Television and in-vehicle entertainment systems. 2
- 3. Post-PC impact on Server Side ● Incredible number of devices connected to the network ● So many clients, so many requests ● High demand for scalable-mission critical systems and architectures ● Distributed non-relational data, also known as “big-data” ● Reduced needs for huge server(s) in small business 3
- 4. What all these devices have in common? 4
- 5. The ARM Processor ● ARM is a 32-bit RISC architecture (coming to 64-bit) ● Supports multi-core (up to 4 cores) ● Available from many vendors: FreeScale, nVidia, TI, Marvell, .... ● Advantages: low power consumption, low cost ● Largely used in client devices.... ● ... but ... 5
- 6. ARM processors on servers ... power consumption and assemply costs sounds interesting in the server side. In particular: 1) In small and medium business they can replace singleservers with “micro-servers” such as smart-NAS 2) In large datacenter, where cooling and power are important costs, ARM allows more server density. 6
- 7. Drivers for ARM Servers in the data centre Data center trends: ● ● ● Power = problem - $26B/yr to power and cool volume servers Facilities max'd on power means that there is a CapEx hurdle before deploying more servers Web and Cloud workloads are emerging as distinct addressable market ARM SoCs fast evolving for server workloads ● ARM processors already deliver multicore 2GHz+ performance ● Virtualisation, 64 bit support coming over next 2 years ● Lower power is just the beginning. Re-architecting of the server subsystem , fabric and systems management to further increase the TCO advantages. Potentially large TCO Advantages (OpEx and CapEx) ● ● 7 CapEx driven by lower acquistion costs and higher density at existing data centers OpEx drops based on power and cooling
- 8. Ubuntu 11.10 and ARM Ubuntu Server 11.10 is the first general-purpose, serverfocused operating system that supports the ARM architecture. Canonical has done significant work to support the architecture with the core OS and with essential high-volume server applications such as the LAMP stack. The ARM architecture is considered to have the potential to improve the performance per watt density of data centers dramatically. 8
- 9. Ubuntu on ARM – Since 2008 1st Native Builds Ubuntu 9.04 Ubuntu 9.10 Ubuntu 10.04 Ubuntu 10.10 Power Mgmt Neon Optimisation ARM v7 JIT acceleration Flash 10 SMP support Thumb2 sppt Optimisation Device tree ARM v6 First device: Sharp Netwalker Webkit-based browser Toolchain Improvements Supporting QT TI OMAP 3 Marvell Amanda support Freescale IMX51 support + TI OMAP 4 ARM & Canonical collaborate to accelerate Cortex A8 & A9 support 2008 H2 9 2009 H1 2009 Builds 1st NativeH2 Linaro based 2010 H1 2010 H2
- 10. Micro-Server for SMB deployment With clouds being deployed, there's less need for a real server in Small and Medium Business. Most of the need is an advanced file-storage solution and some generic purposes functions: ● Web server ● PBX (Asterisk) ● Small (local) databases ● Back-up 10
- 11. Datacenter: target workloads Web servers Web servers App interfaces Web hosting Application Servers Offline Analytics Storage Servers Memcached appliances Search indexing Object Storage MapReduce Swift Finance modelling Gluster Content delivery ● Workloads are common LAMP stack workloads ● Workloads that parallel-lize well are obvious targets ● 11 Systems Management becomes even more important when you look at the node density possible with ARM boards in blades
- 12. ARM for scalable Web 2.0 The perfect workload for ARM in the Web 2.0 architecture, where millions of clients are trying to access the same resource (such as Facebook, twitter, ...) ● ● ● 12 Cassandra is an open source distributed database management system that is designed to handle very large amounts of data spread out across many commodity servers while providing a highly available service with no single point of failure MapReduce is a software framework introduced by Google to support distributed computing on large data sets on clusters of computers. A number of clones can implement the same framework, such as Hadoop Web Servers and Application Servers that are able to serve million of pages in a flexible ways
- 13. ARM and Big-Data ● ● ● 13 Big Data generally refers to database applications that manage Non-Relational or Unstructured data. Differences occur in the consistency model, the data model and therefore workload suitability Big Data applications are used extensively by companies with huge amounts of unstructured data (Facebook, Google, Yahoo etc..) but increasingly more traditional enterprises (Banks, Power companies, Telcos, Govts) are seeing the value in being able to process huge amounts of this type of data quickly.
- 14. Why Ubuntu Server on ARM? Perfect match on target workloads ● ● ARM experience ● ● ● Ideal business model ● ● 14 Ubuntu is strongest in the areas that ARM will break into – Webservers, Hadoop & Cloud workloads. Popularity with developers will accelerate the very patching of FLOSS apps and ISV apps that is needed to see growing support for ARM in the ecosystem. Canonical have worked for three years with ARM and ARM partners on optimising for Cortex A8 and A9. Ubuntu is already natively compiled for Cortex A9 with Thumb2 optimisation and support for NEON switched on Ubuntu fully supports all Linaro optimisations and is the leading adopter of work from Linaro. Canonical charges for Services not for bits. Customer do not pay for security patches or updates. Hyper-scale users (target market) strongly desire this model. This model also lends itself very neatly to Cloud deployments
- 15. Timeline Ubuntu Server 12.04 LTS Ubuntu Server Edition 11.10 April 2012 Optimisation/test 11.10 Ubuntu Server Editioncycles November – March 2012 Ubuntu Server Edition 11.10 Ubuntu Server Edition 11.10 October 2011 Development begins Ubuntu Server Edition 11.10 June 2011 Team Server Edition place Ubuntu selected and in11.10 June 2011 Partners onboard Ubuntu Server Edition 11.10 May 2011 Development Cycle 11.10 Alphas: June, July, August Beta: September Release: October 15
- 16. Example case studies Render Farms Weta Digital ● ● Ubuntu used to render Avatar and King Kong VDI Infrastructure Guateng Province, SA ● 35,000 cores in 5,000 HP Blades with 104TB RAM ● File and print servers Internal file and print servers running on Ubuntu 16 Government Cloud Centralised datacentre providing infrastructure to 2M pupils in 1,200 schools 1500+ Dell Servers Hosting Rackspace ● Over 50,000 Ubuntu Servers British government team responsible for showing better IT patterns by using new technologies . Infrastructure built out on Ubuntu • Web 2.0 Web 2.0, social networks, online gaming, hosting service providers, public cloud providers, and startups prefer Ubuntu
- 17. This will be your next datacenter 17
- 18. Questions please Thank you Giuseppe Paterno' EMEA OEM Technical Manager gpaterno@canonical.com www.canonical.com