White paper ATCA Compute Platforms (ACP) Use ACP to Accelerate Private Cloud Deployments for Mission Critical Workloads Rev 01
Contents 1 ATCA COMPUTE PLATFORMS... 1-3 1.1 CLOUD COMPUTING... 1-3 1.1.1 Private vs. Public Clouds... 1-3 1.1.2 Virtualization: Turnkey vs. Reference Solutions... 1-4 1.2 ATCA TECHNOLOGY... 1-4 1.2.1 Redundant 10GbE and 40GbE Helps Virtualization... 1-4 1.2.2 ATCA High Availability Features... 1-5 1.3 MISSION CRITICAL APPLICATIONS... 1-5 1.3.1 ACP Hardware Configurations... 1-6 1.3.2 ACP1400 Architecture Details... 1-6 1.3.3 ACP1400 Disk Choices... 1-7 1.3.4 ACP1400 Initial Configuration... 1-7 1.4 CONCLUSION... 1-7 ATCA Compute Platforms 1-2
1 ATCA Compute Platforms While cloud computing and virtualization can bring tremendous new values though improved sharing and efficiencies, IT departments are discovering the effort to buy, integrate and test bare metal components (computers, network and storage) procured from different suppliers can consume nearly 25% of their project timeline and development resources. The problem is compounded when virtualization is combined with embedded computing solutions, such as ATCA which boast the highest concentration of server and storage equipment per cubic foot of space. The goal of this white paper is to introduce the value of SANBlaze ATCA Compute Platforms (ACP) and how they can eliminate hardware integration efforts and reduce incoming dock to deployment times to just a few days. These single SKU configurations contain pre-tested and integrated components that target specific mission critical workloads and bring numerous advantages inherent with ATCA technology: Fully redundant cloud platform High availability Multiple levels of Fault tolerance Environmental survivability Superior Space, weight and power (Swap) characteristics 1.1 Cloud Computing IT departments worldwide are embracing cloud computing as a means to effectively utilize x86 computers which now commonly ship with two processors, each with 4, 6 and even 8 CPUs. Cloud computing is a method to interconnect all of your hardware assets (CPU, memory and storage) and manage them as a single virtualized resource pool with improved utilization efficiency. Administrators no longer choose a specific physical computer to run applications, rather they open a management tool to define a virtual computer by requesting some number of CPUs along with some quantity of memory and storage. These resources are instantly allocated from the global pool, and users can then load an OS and applications as before. However, with virtualization, administrators can dynamically grow or shrink resource allotments (elastic provisioning), migrate and load balance virtual computers, and define failover policies to hide physical component failures occurring in the data center. 1.1.1 Private vs. Public Clouds In general, two philosophies have emerged with regard to cloud computing deployments. Both offer individually unique advantages for organizations considering them. The first deployment model is private clouds and is the focus of this paper; in this arrangement, organizations own and maintain all the physical hosting hardware and thus total control over their data and its management. These organizations often cite security concerns and data privacy regulations as a primary motivation for choosing private clouds. Access points within buildings can be physically locked and restricted to employees. Externally, they can use time tested firewalls to protect unauthorized access. Others want private clouds for mobile data centers. For example, the U.S. Military chose the ACP with private cloud software to run a myriad of local applications in a virtualized environment. This mini data center functions like remote offices which are deployed in off-grid locations that can be easily moved on short notice. ATCA Compute Platforms 1-3
The second deployment model is public cloud ; in this arrangement organizations outsource ownership of physical computers and storage to 3rd party vendors such as Amazon s Elastic Compute Cloud (EC2). Subscribers pay a subscription fee to run applications over the web on virtual computers and storage hosted by Amazon. The ROI can make sense for some companies, needing some level of bandwidth. However, outsourcing in this way over the internet will add application latency, with maximum data bandwidth limited by T1/E1communication lines. Data stored off premises also merits a detailed review of data security policies. 1.1.2 Virtualization: Turnkey vs. Reference Solutions There are several very good commercial virtualization software packages; the most popular four being VSphere by VMware, Hyper-V by Microsoft, Xen by Citrix, and Oracle Virtual Machine (OVM). SANBlaze chose VMware for its ACP platform because it enjoys a 65% share of the market and has won favorable licensing agreements within several branches of the U.S. Military. The ACP is a turnkey, integrated solution and represents the least setup complexity and arguably the most value in terms of cutting deployment tasks. The platform is assembled with components known to interoperate with VMware, and have earned VMReady certifications. This means fewer resources (and time) to integrate, configure and test new systems. By contrast, reference solutions can provide more flexibility in terms of hardware selection, but often require more involvement by integration teams, and more coordination with suppliers. 1.2 ATCA Technology The SANBlaze ACP platform is constructed with ATCA components. ATCA is an open standards specification that defines a moderately rugged blade computing environment with literally hundreds of suppliers, 10 years in the making. ATCA blades aren t anemic, struggling to fit enterprise class computing into spaces too small to accommodate the technology. Rather, these blades feature Intel s newest 64-bit XEON CPUs, terabytes of storage, 10 to 40 Gb Ethernet networks and modern operating systems. To create a cloud hosting platform, system designers combine multiple compute, network or storage blades to create scalable platforms with as few as two blades or groups of many dozens. ATCA was originally conceived to fulfill the computing needs of the US and European telecommunications Industry, replacing monolithic proprietary systems with reusable COTS products that enable faster development, easier technology refresh and to foster a competitive supplier environment to keep costs down. ATCA platforms offer redundancy, remote monitoring and consume up to 50% less space, weight and power (SWaP) than an equivalent set of commercial desktops or rack servers. Lower power yields double dividends: less electricity to run ATCA equipment, which means less thermal run-off and thus less electricity to refrigerate the data center. Lastly, it s worth noting that embedded ATCA computers are designed using components having 7+ year life cycles. This distinction means fewer qualification cycles than deployments using commercial rack servers with shelf lives only 18-24 months long. 1.2.1 Redundant 10GbE and 40GbE Helps Virtualization An ATCA chassis includes a backplane to accommodate two 10Gb (or 40Gb) Ethernet switches and two 1Gb switches, each wired in a star fabric configuration. This architecture provides full redundancy for each server slot, yielding two independent fabric ports for data movement, and two 1Gb ports for data management. This arrangement benefits virtualization in two ways: ATCA Compute Platforms 1-4
1. Dual connections ensure server and storage remain online even if an entire switch component is being serviced or suddenly fails. Path redundancy is crucial to virtualized environments that rely on the network to pool physical resources. 2. 10Gb links are better suited to support IP storage (iscsi and NAS) and sophisticated load balance features offered by VMware such as vmotion and Storage vmotion. 1.2.2 ATCA High Availability Features ATCA designers included several attributes to ensure the technology delivers the high availability features needed to host mission critical applications. We ve already mentioned the redundant high speed (10GbE) networking backplanes. ATCA platforms also incorporate redundant power supplies, redundant fans and a sophisticated redundant shelf management (ShM) system that monitors temperature and voltage of every member component. The shelf manager pair contains mini computers that coordinate with each other using I2C serial communications to control blade power, spin fans faster if temperature rises, and manage live component hot swap. The shelf managers support long distance remote communications using secure console shells (ssh) over standard Ethernet lines. All of this contributes to the platform reliability of the SANBlaze ACP, and its ability to operate normally at 40 C, and tolerate 72 hours periods at 55 C. 1.3 Mission Critical Applications The family of ACP products from SANBlaze delivers a foundation for hosting your mission critical applications. By definition, a mission critical application is any application that is critical to running your organization or business. Offline applications prevent employees from doing their jobs: selling, order entry, manufacturing, designing, receiving emails or even making phone calls (VoIP). Hardware fails on occasion; disk drives can suddenly stop working, computers can crash and refuse to reboot. If these applications are important, it is best to execute on a hardware platform with built-in redundancy and split second failover responses. ATCA Platforms are designed for hands-off operation, and intended to thrive in harsher environmental conditions outlined in a set guidelines called NEBS (Network Equipment-Building System). NEBS compliance requires that equipment must survive conditions such as the severe vibration caused by earthquakes, noisy power supplies, and elevated temperatures common in central office/ deployments. By meeting these tougher standards, ATCA is ideal for Mil-Aero applications that need equipment survivability. For example, ACP platforms, properly shock mounted, survived military barge testing that inflicted 60G shocks to the equipment. ATCA Compute Platforms 1-5
1.3.1 ACP Hardware Configurations To suite different budgets, the ACP is available in two configurations, a 6 slot and a 14 slot chassis. Both offer similar redundancy capabilities but differ in the capacity of CPU, memory and storage they can accommodate. For comparison purposes, the chart below shows two sample content configurations. Users can easily trade compute slots for storage slots to meet specific application needs. Family Slots CPUs Memory (DDR3) RAID storage Switch ACP1400 14 120 1.92 TB 12TB 2@10Gb ACP6000 6 48 0.768 TB 6 TB 1@10Gb Initially, users could purchase a 14 slot with fewer blades to start, and later add compute or storage blades as application needs dictate. Each new blade is immediately available to the cluster of adjacent nodes and resources via redundant 10GbE links. 1.3.2 ACP1400 Architecture Details ATCA blade resources are interconnected with dual star Ethernet fabrics as shown. The RAID disks are pre-configured in double redundant manner; RAID-5 (single parity) is used on each blade to survive local disk failures and RAID-1 (mirror) is used between two blades to ensure storage availability even if a partner blade fails entirely. VMware ESXi hypervisor (v5.1) is loaded on each of the compute blades, along with a preconfigured VM (virtual machine) running vcenter. ATCA Chassis ATCA2000 RAID 12 TB 10Gb Switch 10Gb Switch External Switch(s) ATCA2000 RAID 12 TB The pictured unit contains 7 CPU blades, redundant RAID blades (each with 6TB), a pair of redundant switches and 3 spare slots. These spare slots can be used to expand either computers or storage. In this configuration users can randomly remove any component, and the platform continues to run. The entire pool of physical resources (storage and CPUs) is presented to VMware for configuration assignment. ATCA Compute Platforms 1-6
Laptop ACP1400 1.3.3 ACP1400 Disk Choices The ATCA storage blade includes a SAS controller device from LSI Corporation that performs hardware RAID 0,1,5,6 functions. This controller supports SAS, SATA and SSD disk which are user serviceable through slots that exist on the front and back of the unit. Storage is added in slot increments, with up to 10 SSD each. Though more expensive, SSD can provide 5-10X more performance than a rotating SAS drive, which often justifies their cost for some applications. The blade also supports hybrid configurations that mix some number of rotating disks with solid state disks. The SSD can be configured to create large read/write cache, or dedicated to separate datastores and assigned to specific VM s needing ultra fast storage. 1.3.4 ACP1400 Initial Configuration Users can begin final configuration tasks using any client machine (laptop) immediately after powering up the platform. The laptop communicates with a vcenter application pre-loaded on the platform. At this point the user enters license information purchased for this hardware configuration, defines virtual machines (VM), and begins to install OS (Windows/Linux) and application. Applications run no differently on a VM than they do on physical hardware. 1.4 Conclusion The SANBlaze ACP is an ideal platform for hosting VMware virtualization software with just right provisioning to extract the maximum performance and efficiency out of your hardware investment. Its features are tailored for private cloud applications that require high availability and high performance hardware infrastructure. The ACP, based on ATCA technology provides an unmatched level of reliability with dual 10Gb (or 40Gb) Ethernet fabrics, native environmental monitoring, and redundant power and cooling. This whitepaper outlines the many reasons the SANBlaze ATCA Compute Platforms (ACP) are the right choice for hosting your mission critical applications. ATCA Compute Platforms 1-7