The storage networking market has been plagued by hype, over-promised benefits, broadly interpreted standards and...
even wildly diverging meanings of the same descriptive words. This is not any different from any other high-tech market; however, the consequences for the very risk-averse storage market can be and has been excruciatingly painful.
ISAs are storage initiators (for servers) or targets (storage) that can be programmed in some fashion. It is the term "programmable" that's open to interpretation depending on the vendor. Programmable can mean that the microcode (firmware) can be controlled or manipulated programmatically through an API by scripts or other programs. Or it can mean that the applications can be ported to run right on the adapter.
The former meaning can be very interesting to system vendors that may have pieces of what comes with the adapter card such as iSCSI stacks or TCP stacks. This allows them to reutilize microcode or software they have and leverage software functions in the ISA they don't, such as iWARP (RDMA) or zero copy. This makes financial sense to the system vendor where the value is in shorter time to market and less R&D. It has the potential to save them money and increase their revenue. For the user, it means little or nothing.
The other meaning can be a bit more interesting. Think of it as the ultimate offload. Instead of just offloading functions such as TCP/IP, iWARP or iSCSI a whole world of important and annoying applications can be offloaded and integrated. Take an application, such as PXE Boot or Win Boot from the SAN, where the application resides on the target storage ISA. This application would allow software iSCSI drivers on Windows or Linux to boot off the SAN and eliminate the need for internal HDDs. This means significant cost savings for the application servers and eliminates the need and cost of dual appliances on the SAN with their requirements for cabling, power, cooling and real estate.
Another offloaded application can be wide area TCP acceleration and optimization. This offloaded application has value on both the initiator and target side, again eliminating the requirements and costs of appliances, power, cooling, cables, etc. Similar benefits can be seen for encryption (in-flight over the WAN or at rest for target storage such as secondary disk or tape).
A good way to think of this type of ISA is as an integrated storage adapter and low-powered Linux or Windows server on a PCIe or PCI-X adapter card. The value to both the user and system vendor can be enormous from the reduction in dedicated appliances/servers in the storage network, fewer cables, less power, cooling, real estate, etc. This type of ISA definitely makes sense based on the application being offloaded.
Unfortunately, there are currently no standards for ISAs and there probably won't be for some time to come. Each vendor's implementation utilizes their own unique ASIC. Even the APIs are different. This means when you select a vendor (as either a vendor or user) you will have to standardize on that vendor for at least this generation of ISAs.
ISAs were developed in part as an offshoot of the iSCSI storage adapters and have their primary value on the Ethernet side of storage networks; however, there are Fibre Channel versions as well.
Be assured that there is definitely value in ISAs. The amount of value to you is a familiar story. It will depend on the application being offloaded.
About the author: Marc Staimer is president and CDS of Dragon Slayer Consulting in Beaverton, Oregon. He is widely known as one of the leading storage market analysts in the network storage and storage management industries. His consulting practice of six plus years provides consulting to the end-user and vendor communities.