A Closer Look at the PureApplication System
On April 11, 2012, IBM announced the PureSystems platform. The first two members of this new family are IBM PureFlex System and IBM PureApplication System. The PureFlex System is a combination of server, storage and networking hardware with a management infrastructure that provides single pane of glass management for all the compute nodes in the system. You can read more about it in my article, “A Closer Look at the PureFlex System.”
PureFlex can be seen as an infrastructure system while PureApplication is a platform system. PureApplication System goes beyond just the integration by design for the platform, adding on expert application patterns that ensure scalability and best-of-breed implementations.
Multiple industry surveys show that more than 70 percent of IT budgets are spent on operations and maintenance, leaving very little money and personnel time to implement new technologies and applications. Additionally, according to IBM, more than two-thirds of companies go over schedule on their project deployments. IBM PureApplication System is a member of the family of new expert integrated systems from IBM and is designed to help alleviate these problems by combining an integrated system with a simplified experience and built-in expertise.
PureApplication System uses a purpose-built management system with a unified console. When you log into the PureApplication System, you choose between the workload and system console tabs, depending on the actions you want to take. From the console, you can deploy application environments, monitor virtual servers and environments, apply maintenance, gather data for chargeback, manage licensing and entitlements, monitor hardware utilization and resource consumption.
PureApplication System comes pre-built in four configurations and smaller configurations can be upgraded to the next level as the organization needs increase. These configurations are:
W1500-96 W1500-192 W1500-384 W1500-608
Cores 96 192 384 608
Memory 1.5TB 3.1TB 6.1TB 9.7TB
SSD Storage 6.4TB 6.4TB 6.4TB 6.4TB
HDD Storage 48TB 48TB 48TB 48TB
The system is shipped fully assembled and cabled within the rack and upgrades can be completed without powering down the installed machine. The storage consists of 54.4TB (unformatted) that’s split between SSDs and hard drives. The system monitors disk usage and optimizes placement of data across the disks so that hot data is stored on the faster SSDs. This helps prevent hot spots and improves I/O performance and is done automatically.
The system is designed for redundancy with a redundant management system, redundant power supplies and storage in multiple RAID arrays. Compute nodes run stateless so it’s easy to failover work to another compute node should there be a hardware failure on a compute node. The base hardware consists of the rack, a Storwize V7000 disk subsystem with two controllers, two expansions and a total of 16 x 400 GB SSDs and 80 x 600 GB HDDs to provide the listed disk space. The compute nodes go into a Flex chassis that holds 14 x half-width compute nodes, 2 x 10Gb Ethernet switches and 2 x 16Gb fiber channel switches. The nodes themselves consist of 2.6 Ghz eight-core Intel processors (16 cores per node) with 256 KB of L2 per core and 4 MB eDRAM L3 per core. Additionally, each compute node has 2 x 4-port 10 GbE and 2 x 2 port 8 Gb/s fibre channel interfaces. A full 608-core system consists of three chassis with 38 compute nodes, redundant PureApplication System Management nodes and redundant VM management nodes—and only consumes approximately 30 KW of power.
comments powered by