Despite the potential expense, however, the need for SAN services is growing rapidly. To whit:
Earlier this month IDC announced figures on the 4th quarter 2009 shipments of disk storage systems. Despite the economic downturn, shipment of storage system capacity was up 33.4% year-over-year at 3,304 petabytes. Thats over 3 exabytes.
Compare that to figures for 2Q 2007, when growth was at over 50% and the capacity shipped a mere 1.3 exabytes. Its pretty clear that data storage continues to grow at exponential rates, regardless of the economic climate.
Data (and more importantly information) is the core asset of many businesses, and so as data quantities grow, the need to secure that data becomes ever more important. Add to this the requirement to ensure that data is adequately managed from a regulatory perspective and its clear that cost effective and reliable data storage is a must.
What is the best approach to deploying SAN services within your organization? And how should you choose the right vendor and technology to suit your needs?
SAN Services Overview
Todays SANs and SAN Services
Fibre Channel Switches and Directors
Host Bus Adaptors
Choosing a SAN Services Vendor
SAN Services Selection Process
San Services Summary
It's worth taking some time to look at why SANs and SAN services technology arose in the first place.
In the mid 1990's IT saw the rise of departmental servers, countering the mainframes centralized model. Distributed computing empowered individual departments within an organization to select and purchase their own computing resources and to be freed from the restrictions of central IT.
However, the servers moved back into the data center as it became obvious that they were best physically managed in a secure central location. At the same time, we also saw the rise of companies like EMC offering dedicated storage arrays with advanced features such as RAID and data replication. These new storage devices used physically bulky and restrictive cabling to connect arrays to servers.
Storage connectivity presented a real problem in server deployments for a number of reasons. SCSI, the de-facto connection protocol, was limited by distance. Typically this was 25 meters, which when internal cabling within cabinets is included, made it difficult to connect many servers to the same storage array.
In fact, storage and servers were effectively coupled to each other. The SCSI connectors werent small and cables were bulky. A solution was needed to overcome the physical limitations of using standard SCSI interfaces and so Storage Area Networks were born. They resolved the physical hardware restrictions by offering:
Enhanced Connectivity - SAN services technology removes the distance restrictions of SCSI, enabling device-to-device connections of up to 10 kilometers, depending on the cabling type and optics used. Servers were no longer required to be adjacent to the storage.
Scalability - SAN services technology offers significant scalability over deployment of direct attached servers by increasing the number of servers that could be attached to a single SAN storage array. Today a single storage array can support hundreds of server connections.
Consolidation - SAN services technology consolidates storage from many servers into a shared infrastructure.
Todays storage area networks consist of a number of key components.
These are highly available devices that act as the data storage equivalent of the network switch. They connect hosts to storage and can be networked together to form fabrics. Switches and Directors (also known as director-class switches) are differentiated from each other by their level of availability.
Typically, directors have totally redundant and hot swappable components, including power supplies, fans and dual service processors.
HBAs connect the host servers into a storage fabric. They provide the physical interface for moving data across the SAN from the host to storage. In some respects HBAs can be thought of as analogous to Network Interface Cards or SCSI controllers. As storage and IP networks start to converge, HBAs will be replaced by CNAs (Converged Network Adaptors).
Modern storage arrays are categorized by the level of availability they provide and grouped as either Enterprise or Modular.
Enterprise-class storage devices provide very high levels of availability and redundancy of components. Failure of a single component will have little or no impact on performance and availability. Modular devices offer high levels of availability but performance of these devices is typically impacted by a component failure. Apart from availability, cost is a big differentiating factor between Enterprise and Modular arrays, with modular devices being significantly cheaper.
SANs today can be connected using a number of protocols. These specify the data transmission format used to communicate between storage and host. Devices can be connected using Fibre Channel (usually carried over fibre-optic cables) or two IP based protocols, iSCSI (Internet Small Computer System Interface) and NAS (Network Attached Storage). Both of these latter protocols are cheaper to implement, as they dont require relatively expensive Host Bus Adaptors in each server.
The best choice of protocols is a subject of much debate and the emergence of a new protocol called Fibre Channel over Ethernet (FCoE) is adding further fuel to the fire. FCoE offers the benefit of combining both Fibre Channel and IP traffic onto a single unified interface or Converged Network Adaptor (CNA). The use of FCoE is not widespread as standards have only recently been ratified and device support is limited.
There are a huge range of storage vendors and products to choose from in todays market. So selecting a storage vendor and product may seem like a daunting prospect, especially considering the high costs involved.
Ultimately, before embarking on a purchase, organizations should take a step back and look at the SAN service they want to provide to their customers. The first step is to develop a Service Catalogue, describing the service customers can expect. This will be a list of storage offerings describing features of the provided storage.
The storage catalogue should avoid mentioning specific hardware but instead should focus on more abstract details, such as price, performance, availability and service levels for a fixed volume capacity. The reason for keeping the storage catalogue hardware agnostic is to enable technology to be replaced in the future without requiring the catalogue to be rewritten. Over time, technology improves and the same service can usually be delivered for a lower cost.
In addition to the storage catalogue, thought should be given to additional features the SAN service may deliver. These can include:
RAID protection - All storage arrays provide multiple levels of RAID protection, from RAID-1/10 to RAID 5 & 6. Each RAID level has a different performance characteristic (for example RAID-1 is more suited to random write data) and cost overhead.
Remote replication - Most storage devices provide the ability to replicate data to another storage array, typically situated in another data center. Replication can be synchronous or asynchronous depending on whether updates are confirmed in both devices before acknowledging to the host that the data has been written.
Local replication - Local replicas of data (also known as snapshots or clones) provide point-in-time copies of data and can be used for recovery, backup or other business related processing. Clones are full copies of data while snapshots usually track only changed data.
Thin Provisioning - This is a technology that ensures only data written by the host is stored on disk. It can reduce storage capacities by as much as 50% on some environments.
De-duplication - This feature identifies common blocks of data and keeps only a single copy on disk. It can result in significant storage savings, but has the tradeoff of impacts on performance and availability.
Once the requirements of the SAN have been established, the most appropriate vendor and technology can be chosen.
For medium to large enterprises with a significant SAN services capacity requirement, the best approach is to issue a tender document to prospective vendors. The tender document will outline the requirements, set the parameters against which each vendor must respond and the required format of the responses.
Although this is a formal approach it helps to normalize the structure of responses to make vendor and product comparisons easier. In addition to requesting information on technology, the tender document should also require the vendor to provide information on service levels and agreements and performance guarantees.
Making the final decision can seem daunting and it is worth considering involving an independent storage consultant to assist with the whole process of delivering a SAN service. The consultant can provide advice and direction in product selection, history and background on the vendors. They can also manage the tender and selection process for the customer.
To summarize, implementing a SAN service within your organization enables data to be stored in a more secure and cost effective manner. The process of SAN implementation should start by establishing requirements and service offerings. The service catalogue this creates enables more informed and appropriate technology selection. The selection process should ensure the vendor is committed to providing the right solution by implementing performance and service guarantees as part of a vendor agreement.
One final thing to consider: implementing your first SAN may only be the start of an ongoing cycle of technology refresh.
Chris Evans is a consultant with Brookend, an independent SAN and SAN services consultancy based in the UK.