If you've completed a business case for purchasing a SAN, the next step is to determine which criteria to include...
By submitting your personal information, you agree that TechTarget and its partners may contact you regarding relevant content, products and special offers.
in your vendor request for proposal (RFP) to ensure you evaluate systems that fit your environment, use cases, budget and business needs.
With a relatively large list of SAN array vendors and products, this can be a somewhat overwhelming task. Without a structured approach, you may end up with an inadequate product that will most likely cause some adverse financial impact due to workarounds, reduced efficiency, or having to supplement or replace a product misfit.
This article is the second in a series that guides users through the SAN buying process, and outlines the important information needed to make an educated buying decision. The first article outlined SAN fundamentals to help you determine whether a SAN -- or a storage architecture such as DAS or NAS -- would be the best fit for your environment.
This article digs deeper into SAN essentials to describe the critical features, performance metrics and other purchasing criteria to consider when preparing a vendor RFP. A well-structured RFP that considers relevant evaluation aspects is a key factor in compiling a short list of SAN arrays and, eventually, your product of choice.
The third article will compare SAN arrays against these purchasing criteria to make sure your ultimate purchase is the right system for your organization.
Hardware: Architecting the SAN array
SAN array hardware architecture is an essential evaluation criterion, since it greatly impacts your system's availability and ability to scale. There are two main ways to architect a SAN array:
- As a monolithic single-system, multi-controller array with each controller accessing a single shared pool of storage.
- As a multi-node array that consists of loosely coupled independent nodes, each with its own processors, storage and ports.
Single-system multi-controller arrays have been around since shared storage systems came into existence and most storage systems are still architected this way.
Midrange to high-end, single-system SAN arrays typically provide two storage controllers. Some are offered as active-passive controller arrays, where one storage controller does all the work and the second only comes into the picture when the primary controller fails. Other systems on the market are architected as active-active controller arrays, where both controllers are used for data processing.
The maximum number of controllers supported directly relates to scalability and performance. High-end and very high-end SAN arrays typically offer more than two controllers; as the maximum number of supported controllers grows, the more high-end the SAN array.
Multi-node SAN arrays consist of independent storage nodes, each with its own storage controllers, disks and ports. Multiple nodes are coupled to form a single scale-out SAN array. As nodes are added, performance and capacity scale.
While some multi-node SAN arrays stripe data across all nodes to maximize performance, others work with smaller stripe sets. Similar to their single-system counterparts, the number of nodes and maximum number of supported nodes determines performance and scalability.
Disk subsystem: Drive types and features
Disk subsystem and disk options are another essential evaluation aspect. SAN arrays range from all-flash solid-state arrays, to hybrid arrays with both solid-state drives (SSDs) and hard disk drives (HDDs), to arrays with no SSD support.
A hybrid SAN array with mechanical disk drives and SSD drives usually provides the best price/performance ratio. A mix of high-capacity HDDs and a small percentage of SSDs is able to outperform traditional high-performance all-disk arrays at a better cost and with a smaller form factor while requiring less power and cooling.
You'll need to correctly size the amount of solid-state storage to optimize both price and performance. As far as large-capacity HDDs are concerned, near-line serial-attached SCSI (SAS) drives are preferred over SATA drives because SAS is becoming the preferred back-end protocol to connect to disk subsystems.
Even though SSDs are reducing the need for high-performance HDDs, 10,000 rpm or 15,000 rpm SAS drives are becoming the norm. As SAS-3 becomes more prevalent, look for 12 gigabit per second (Gbps) SAS drive support.
All-flash arrays come into play when very high performance is required. SSDs are still significantly more expensive than high-end HDDs. However, if architected correctly, an all-flash SAN array embraces writing data randomly rather than trying to get data in sequence to accommodate mechanical HDDs. As a result, these arrays can support real-time compression, deduplication, clones and snapshots, which can somewhat offset the high cost of NAND flash.
An increasingly relevant consideration is the ability to encrypt the data stored on drives. Encryption can be performed in the array back end, usually combined with an integrated encryption key manager, or via self-encrypting drives. Having the SAN array capable of performing encryption is preferable in most use cases.
For SAN arrays with multiple tiers of storage, such as hybrid SSD/HDD arrays, the ability to automatically move active data to a faster tier and inactive data to a slower tier is a factor. SAN arrays vary in how granularly data can be moved, and the more granular your data can be moved, the better.
As part of the disk subsystem evaluation, you need to review the RAID options supported by the SAN array and how data is distributed across disks and nodes. As a rule of thumb, wider stripe sets result in better performance and higher resilience.
Does the SAN array have a RAID 6 or equivalent option that enables more than one simultaneous drive failure? And what's the performance impact while a RAID is rebuilt? With increasingly larger drives, RAID rebuild times get longer and longer, and the ability of the SAN array to survive two concurrent drive failures becomes more relevant.
Solid-state storage cache
Supporting NAND flash as cache rather than, or in addition to, SSD drives is a worthwhile consideration. Using NAND flash as a cache benefits all data traversing the SAN array, not just what resides on a NAND flash tier. It also automatically keeps active data in NAND flash without additional mechanisms to shuffle data between the flash and slower disk tiers.
Some arrays offer NAND flash cache in the form of SSDs, others as PCI Express add-in cards. In some arrays, NAND flash cache only supports read transactions, while both reads and writes are supported in others. Ideally, the NAND flash cache should support reads and writes.
Performance: Benchmarking is key
Although SAN array performance is difficult to gauge as a result of workload dependency and array configuration, your evaluation wouldn't be complete without discussing IOPS, latency and throughput. To get some level of objectivity, request independent third-party performance benchmarks and, if available, understand what configurations and type of workloads the test results were obtained from.
Find out if the vendor has submitted the SAN array to the Storage Performance Council (SPC) and compare SPC-1 and SPC-2 test results if available. SPC benchmarking is one of very few independent benchmarks where you can compare arrays tested the same way.
Like any benchmark, SPC results need to be viewed with the hardware configuration in mind. To give some level of specificity to performance numbers, IOPS over 1 million and latency below 1 millisecond rank at the top of the performance spectrum.
Fibre Channel (FC) was the original SAN protocol, and 16 Gbps or 8 Gbps FC is still the top choice for high-end networks.
ISCSI emerged in the SAN space about a decade ago, first in the small and medium-sized business space. With the emergence of 10 Gigabit Ethernet (10 GbE), SANs began encroaching on the high-end storage space where it now competes with FC.
Support of FC over Ethernet (FCoE) enables running FC protocols over Ethernet networks, which reduces the need to deploy and run FC networks and switches. FCoE also reduces the complexity associated with attaching FC arrays to an otherwise Ethernet-based storage infrastructure.
Fiber Connectivity support is important if you need to connect computers based on the IBM z/Architecture, the latest offspring of IBM's System/360, System/370 and System/390 series mainframes. The z/Architecture has historically been supported by very high-end SAN arrays, and is not only important for companies that run mainframes, but a good indication as to how high end a SAN array is.
Besides block-level protocol support, you should look at file system protocols such as the Microsoft-supported Server Message Block (SMB) protocol and Network File System (NFS). While some SAN arrays provide native file-system protocol support, others depend on NAS gateways for SMB and NFS access to SAN storage.
Front-end and back-end ports
Front-end ports enable servers, switches and other arrays to connect with the SAN array over supported protocols. You should look at the type of ports supported by the array -- such as 16 Gbps and 8 Gbps FC, 10 GbE and 1 GbE, and 10 FCoE -- and the maximum number of available ports.
Back-end ports connect to the disk subsystem and disk shelves. The number of ports and type of ports, such as 12 Gbps SAS, determines the maximum throughput an array can handle and directly impacts its scalability.
Availability and reliability
Besides performance, availability and reliability of the storage system should be at the top of your list. A service-level agreement (SLA) is a good indicator for how robust a SAN array is.
While 99.99% (also referred to as "four nines") availability relates to approximately 1 hour of downtime during the year, 99.999% (five nines) relates to a mere 5 minutes and 99.9999% (six nines) to 31 seconds.
You should look for at least 99.99% availability; 99.999% availability or higher is best for high-end enterprise arrays.
Determine how the vendor arrived at its SLA and, if possible, have them provide documentation. Inquire about all single points of failure and determine what hardware component replacements require downtime.
Understand if firmware upgrades can be done in a non-disruptive manner and the rollback options of firmware and software upgrades. You should aim for a SAN array with no single points of failure and where all maintenance tasks can be performed while online.
Robust error detection and system health checks should also be assessed. For instance, the array should run frequent system diagnoses and be able to repair common repairable drive issues. If there is an issue, the storage system should report back to the vendor, who should proactively take actions such as sending replacement drives. Look for a SAN array that pushes the limits on self-correcting and self-healing capabilities.
Data efficiency features
Capabilities that help to maximize the use of storage resources result in significant cost savings and should be an important aspect of your evaluation. In addition to looking at how your hybrid and all-flash arrays use data efficiency features to maximize your solid-state investment, these features should be available for all-HDD arrays.
Thin provisioning and efficient clones are widely supported and should be on your must-have list. Increasingly, primary storage systems support data deduplication and compression, and although they are not yet must-have features, both are big pluses. Regarding data deduplication, storage systems with real-time data deduplication capabilities rather than post-process deduplication are preferable. Real-time data deduplication and compression are must-haves in state-of-the-art all-flash arrays.
Quality of service (QoS) enables storage efficiency by allowing oversubscription of storage resources without impacting critical applications and data. Since SAN arrays consolidate a wide range of workloads, QoS should be on your must-have list. However, QoS capabilities vary greatly in contemporary SAN arrays.
Data protection capabilities are another critical aspect. Make sure the SAN array supports snapshots efficiently, since they can be disk-fillers.
Understand how much disk space a snapshot/subsequent snapshots use and the maximum number of snapshots supported. A SAN array should support asynchronous replication for long-distance replication, and synchronous replication to support high-availability storage area networks.
Storage management capabilities
Storage management and the ability to automate storage management tasks are an important aspect of any SAN RFP. You should be able to manage the SAN array via a command line interface that enables scripting of certain storage management tasks, and through a user-friendly management console that's preferably browser-based.
Integration with hypervisor management tools such as VMware vCenter is a big plus. Analytics tools and reporting capabilities that can comment on various storage metrics, including performance and issues, are pertinent. The ability to configure alerts and policies that trigger actions if certain thresholds or conditions are met are critical capabilities.
Integration is another crucial aspect. Request all supported VMware vSphere and vCenter integrations and supported application programming interfaces (APIs), such as vStorage APIs for Array Integration and vStorage APIs for Storage Awareness, as well as supported Microsoft Hyper-V integrations and APIs, such as Operation Data Store and Volume Shadow Copy Service. Determine which application integrations are relevant for your environment and use cases, and include those requirements in your RFP.
SAN arrays are purchased for a wide range of use cases, and those use cases and specific requirements determine how relevant certain features and RFP aspects are. While this article comments on relevant aspects, focus on what matters for the problems you are trying to solve. A good RFP should ask very specific, non-confusing questions, include all relevant aspects, and be vendor agnostic so you can arrive at an objective SAN array decision.
Why you might need to upgrade your SAN
Factors to consider when selecting a hybrid flash vs. all-flash array
Fibre Channel is still the top choice for SANs