iSCSI-based storage arrays have not captured a substantial percentage of the storage market, but there are a number of compelling products out there. A healthy level of competition in this space, coupled with the relatively aggressive demeanor of the competing vendors, should give IT managers a distinct edge when shopping for new storage.
When creating your RFP (request for proposal), be sure to talk to users and developers to get a firm idea of the different storage and performance needs in the organization.
For more sample RFPs, go to go.eWEEK.com/rfp.
The following questions are designed to serve as a foundation for an RFP for iSCSI storage arrays.
Scalability
What is the maximum capacity of your storage system?
List both raw and usable capacities.
• For a single storage shelf?
• At full capacity (maximum number of shelves populated with the largest-capacity drives)?
* What is the maximum throughput (in megabytes per second) supported on a single controller?
* How many controller units does the product support?
* Do you offer clustering capabilities?
Startups EqualLogic and LeftHand Networks are selling iSCSI systems that can be clustered to add storage capacity and performance.
* Does the product have redundant controllers?
* How much cache does each controller have?
List maximum value.
* Do the controllers have battery-backed cache?
RAID
The level of RAID implemented should depend on the type of application being stored.
* What levels of RAID does your product support?
Check all that apply.
• RAID 0
• RAID 10
• RAID 5
• RAID 1
• Dual-parity RAID (RAID 6)
* If your storage system does not currently support dual-parity RAID, when do you expect that such support will be added?
The issue of dual-parity RAID has risen in importance lately because of the growing popularity of SATA (Serial ATA) hard drives. SATA drives are generally less reliable than Fibre Channel and SCSI drives, and dual-parity RAID adds a level of security.
More important, however, is the issue of RAID rebuild times. SATAs high data capacity goes from being a strength to a weakness in a disk-failure scenario: Large SATA drives (with current capacities of up to 750GB each) can take hours to rebuild if a drive within a RAID dies. Dual-parity RAID adds resilience—with it, three SATA drives would have to die before data loss would occur.
* Does your array have the ability to change RAID levels non-destructively?
This ability is becoming more common, and it allows IT managers to tune their storage to changing application needs. For example, if RAID 5 turns out to be too slow, an IT manager may want to switch to a RAID 10 configuration without bringing down applications and reformatting data volumes.
Connectivity
* In addition to iSCSI, how does your solution connect to servers?
• Fibre Channel?
• NAS (network-attached storage)?
Many iSCSI storage systems on the market today support Fibre Channel as well as NAS. Fibre Channel is recommended for performance-intensive applications. iSCSI is much cheaper to implement and should be good enough for common applications such as e-mail.
TKTK
Tiered Storage
* What hard drive types do you support?
Check all that apply.
• Fibre Channel
• SCSI
• SATA
• SAS (Serial Attached SCSI)
Fibre Channel hard drives provide the highest performance; SATA provides high capacity at a low price; and SCSI and SAS sit in the middle. An added benefit of SAS systems is that they often have the ability to work with SATA drives.
* Does your array support multiple drive types?
* If yes, can drive types be mixed within the array? (For example, can you have Fibre Channel drives and SATA drives in the same shelf?)
With ILM (information lifecycle management) growing in importance, more IT managers have been asking for systems that have both expensive and inexpensive disks. In these hybrid systems, data can be moved to cheaper storage as it gets older, without leaving the array (no need to buy two different arrays). We are starting to see new storage systems that have the ability to move data among storage tiers within a storage system (for example, moving old data from Fibre Channel disks to SATA disks).
Power and Heat
* Does your solution have redundant power supplies?
* What are the power requirements?
• For a single controller
• For a disk shelf
• For management node (if necessary)
• For a fully populated system (maximum number of controllers and disk shelves)
* How much heat does the array release?
• For a single controller
• For a disk shelf
• For management node (if necessary)
• For a fully populated system (maximum number of controllers and disk shelves)
Interoperability
* What server operating system platform(s) does your array support? Please list.
* On what operating system(s) does your management console run? Please list.
Data Protection
Most arrays today come with a number of different technologies for protecting the data stored on them. IT managers should rate the responses they receive in this section according to the needs of their environment.
* Does your array have asynchronous or synchronous replication capabilities?
* What network protocol do you use for replication (for example, iSCSI, Fibre Channel or proprietary)?
* Does your product support snapshots?
* How many? List per-volume and per-array values.
* Does your product support cloning?
Cloning is a feature that has become more common during the last year. An array with cloning can mount the image of a snapshot. For example, after taking a snapshot of a database volume, an IT manager can use the snapshot to make a clone of the volume that can be assigned to a different server. Cloning is a great tool for development environments and can come in handy for testing patches before rolling them out to production servers.
Pricing and Support
* Provide base configuration pricing.
* Provide pricing for a unit outfitted with maximum options.
* Provide pricing for support contract options and annual renewal fees.
Customer References
* Please list customer references.