Infrastructure simplification requires an understanding of a diverse range of products and technologies, each with its own specialist jargon. Below we explain the more common terms you’re likely to encounter.
A relatively new development for industry standard servers, the main benefit of a 64-bit processor is the ability to address more memory than with 32-bit systems. This can be of benefit for server virtualisation, but is of most value for database applications where whole databases can be held in RAM rather than on much slower disk storage.
Effectively servers on cards, blades slot vertically into a custom chassis in an equipment rack. Up to four processors per blade can be accommodated, plus varying amounts of on-board memory and disk storage. There’s no wasted space with a blade server and energy/cooling requirements are minimised. However, the small format does limit disk capacity and external storage is usually required.
Refers to the replacement of older, under-powered servers, with a smaller number of newer, more powerful, systems. However, it can also refer to the replacement of distributed storage system with a SAN and to the simplification of the LAN/WAN infrastructure.
Dual-core and multi-core processors
The latest trend in microprocessor design is to build two processing cores onto a single chip. These dual-core processors provide extra performance and, because they conform to industry standards, can still be used in standard motherboards; the separate cores appear as independent SMP processors to the operating system. Multi-core designs with four or more processing cores per chip are also promised for the future.
Explicitly Parallel Instruction Computing (EPIC)
Intel’s 64-bit Itanium-2 processor is based on a new architecture called EPIC. Operating system software and any applications must be complied specifically for use with EPIC, while a built-in emulator is used to run x86 software. However, this doesn’t always provide the same level of performance as a native x86 processor like the Intel Xeon or AMD Opteron.
When specifying a server for consolidation it’s important to allow for sufficient proceeding power, memory and storage capacity to cope with the applications of all the servers being replaced. You should also allow a margin for growth and to handle unexpected peaks in demand. This margin is referred to as 'headroom'.
Refers primarily to servers based on x86-compatible processors, typically from Intel or AMD including 64-bit chips like AMD's Opteron and Intel's Xeon EM64T. The term also implies support for other industry standards in the areas of memory, storage and connectivity, the end result -- and the acid test -- being the ability to install and run Windows, Linux and other industry-standard operating systems unchanged.
A recently introduced protocol for use on storage area networks, iSCSI enables storage to be accessed remotely using the SCSI block access protocol over an IP-based Ethernet network. It's a lot easier and cheaper to implement compared to a traditional SAN, which requires custom Fibre Channel network infrastructure.
In order to match virtual disks to physical storage devices on a storage area network, most vendors map the virtual resources to the Logical Unit Numbers (LUNs) used to identify SCSI storage devices. As a common industry standard, this approach delivers cross-platform compatibility between different SAN vendors and devices.
The more comprehensive network and systems management products are often referred to as 'frameworks'. Instead of using individual agents to, for example, collect inventory data, distribute software, support remote troubleshooting and so on, just one extensible agent is deployed. The central management application, console and extensible agents together make up the management framework.
Network Attached Storage (NAS)
A self-contained file server, ready-configured to plug into the network. The storage is made available using standard network file sharing protocols. Performance is lower than for a SAN, where block access protocols are used to share storage, but an NAS appliance is typically cheaper and easier to deploy and easier to manage (management is usually via browser interface).
RAID protected storage
Short for Redundant Array of Inexpensive Disks, RAID requires multiple hard disks (an array) across which data is logically spread as a whole, rather than written to one disk at a time. Enhanced redundancy and performance are the key benefits. Also, with a suitable RAID controller, individual disks can fail and be replaced with no overall loss of data.
Reliability, availability and serviceability (RAS)
Options like redundant power supplies and fans, error-correcting memory, and RAID-protected storage are typical reliability and availability features, designed to maximise server uptime. That’s of particular importance when replacing multiple servers with one consolidated system -- as, too, are things like tool-free access, hot-swap disks, hot-swap PCI cards and other serviceability features.
The setup and deployment of a new server. With server virtualisation and high-density blade server it’s possible to provision new servers very rapidly in response to changing demand, typically, using imaging technology to copy a standard operating system image onto the new host.
Also referred to as partitioning, server virtualisation is where physical hardware resources are shared between two or more virtual servers running on the host system like ordinary software applications. Virtual servers appear to have their own processors, memory, network connectivity and storage. They can run a variety of operating systems and can be run concurrently on the same host.
Simple Network Management Protocol (SNMP)
An industry standard protocol that allows network devices, host systems and applications to be managed and monitored from a central console. The SNMP targets usually require constant polling, but some can generate their own alerts (traps) when attention is required.
Where virtualisation is employed it is usually possible to take so-called 'snapshots' of virtual server configurations and virtual disk contents. It’s then possible to undo any later changes and restore these virtual resources to their state at the time the snapshots were taken. Snapshots can also be used to run backups while the virtual servers and disks are in use and being updated.
Storage Area Network (SAN)
On a SAN, disk storage is remote from the servers that use it, connected by either a Fibre Channel or, more recently, ordinary Ethernet networking infrastructure. The disks access the storage using the same block access protocols as for locally attached storage, but can be shared by multiple remote hosts running a mix of operating systems.
You can logically divide up the disk storage on a SAN into a series of 'virtual' disks. This makes for better utilisation of the available capacity and for faster deployment with, for example, no need to power down host servers to attach new disks. Virtual disks can also be dynamically re-sized and backups taken with no impact on either performance or availability.
Symmetrical Multi Processing (SMP)
An SMP server spreads its processing load across two or more processors to enhance overall performance. Industry standard servers typically have 2, 4 or 8 processors, although more can be supported. SMP is of most value for applications that consume a lot of processing power, such as server virtualisation.