SMI takes it place in storage annals: SNIA waves storage standard banner.
This has resulted in several management stovepipes that do not communicate particularly well, including processor complex management, network management, database management, application management and storage management.
Storage management is particularly a sticky wicket. Storage managers are faced with managing multiple management applications, most of which do not speak to each other. Locating and fixing system degradation in this environment is a challenge to say the least. For example, while tracking failing service levels on an RDMBS writing to a RAID subsystem, the storage administrator must delve through multiple storage stovepipes to find and fix the issue.
There is some level of interoperability between stovepipe management applications, but these depends on vendors exchanging management APIs--interface code that allows a vendor to somewhat integrate their package with another's. Since the companies are likely in competition, and API sharing requires endless updates and negotiation, you can imagine how well this goes over in real life.
Nevertheless, customers demand integration and put teeth in their demands by requiring interoperability levels in their RFPs. Like it or not, storage vendors have to achieve interoperable levels and are looking to automate the process.
Enter SMI and SMI-S
SMI, which is the storage piece of the broader CIM standard, provides a highly functional and interoperable interface for managing multi-vendor storage devices. SMI grew out of Bluefin, which predictably changed names when SNIA adopted it. Bluefin was the first industry standard SAN management interface. It's meant to broaden the SAN's usefulness and thereby increase adoption, both in the Fortune 1000 data centers which make up the bulk of SAN customers and, hopefully, into the mid-range market. Up until now they have largely steered away from widescale SAN management, citing interoperability, complexity and cost concerns. The storage industry hopes that the better and more widespread storage standards become, the simpler and less costly SANs will be to deploy and manage.
Building on previous CIM/WBEM development, SNIA adopted the work on storage specs and launched SMI (Storage Management Initiative) to help storage vendors adopt a sophisticated open interface. The result, a standard for managing storage networks, is SMI-S (Storage Management Initiative Specification). SMI-S provides a way for storage vendors to develop simplified and highly interoperable management applications. SNIA suggests that this will reduce the cost of storage management by providing:
* Interoperability: The ability to manage a heterogeneous storage network from a single console.
* Integration: Eliminates the need to manage each device with a separate management application.
* Choice: Allows any conforming devices in the SAN to be discovered and controlled regardless of manufacturer.
The DMTF (Distributed Management Task Force) originally developed CIM (Common Information Model) and WBEM (Web-Based Enterprise Management) standards. DMTF members turned over the storage component of CIM/WBEM, Bluefin, to SNIA which renamed it SMI-S. Ideally, the spec will standardize communication between SAN storage devices and management applications. According to Sheila Childs, SNIA Board of Directors Chairperson, implementing SMI-S as a unified management standard has the potential to radically change the technology of managing computer storage:
* Storage hardware and software vendors will be able to shift resources from API development and sharing to value-added functionality.
* SMI-enabled applications will present richer interfaces to users.
* Customers will be able to achieve lower TCO based on improved SAN interoperability and manageability.
SNIA was no stranger to CIM/WBEM. Technical Work Groups such as the DRM (Disk Resource Management) and the Fibre Channel workgroups had already worked on driving management technology developments based on CIM/WBEM technologies. But in view of the storage aspect of SMI, SNIA has named SMI-S ratification as its primary objective for 2003. Its stated goal is that all member companies' new storage networking products (arrays, switches, extenders, appliances, libraries, management software, etc.) will use the SMI interface by the end of 2005. SMI-compliant products will contain native self-describing interfaces that interpret themselves into a common storage management interface. This will make storage and interoperability development faster and less expensive for vendors, and will also allow applications such as file systems and databases to directly manage storage subsystems.
SMI-S provides single object-oriented models for each type of component in a storage network. These models define common attributes and behavior for standard features and communicate themselves using a single management protocol and transport. The transport can occur over multiple interconnects. Uniform discovery, security and durable naming services are included, and SMS-S includes a lock manager so multiple management applications can safely coexist in the network. These models and protocols are platform-independent and are based on CIM/WBEM's language MOF (managed object format). MOF precisely identifies object models, which allows compilers to automatically generate data type definitions, interface stubs and GUI constructs and insert them into management applications. Object models are extensible.
Andrea Westerinen, VP of Technology and Distributed Management Task Force (DMTF), commented that the recent CIM 2.7 was a significant occurrence. Instead of the previous large-core device file, the definitions are broken into separate settings including data input and collection, core elements, protocol inputs, storage services, and media access devices. CIM also added a database model and networking concepts. The DMTF is working to get CIM 2.8 in preliminary release this summer.
CIM is the information model whose goal is to provide common semantics across network management tools. WBEM is the vehicle for transmitting and sharing that data. In the original release of WBEM, it was largely XML schema running on top of HTTP. Now the DMTF is working to decouple the operations so they can run without specific HTTP programming.
CIM is modeling, WBEM is the communications piece. CIM includes overall designs for database and basic logical modeling, which is how products will report their inventory back. Oracle is running the database working group in the DMTF and have started with MIBs. As development continued, it became clear that the working groups needed to add more areas of control. Larry Krantz, chairperson of SNIA's Storage Management Forum said, "What we needed to do was add more of the more specific pieces. How could we control storage?" A storage-specific offshoot of CIM followed, code-named Bluefin. Bluefin defined the profiles of how vendors could use CIM to control storage devices. The new spec proved it could work, not just for storage devices and HBAs, but beyond to switches and other storage components. The expansion needed a great deal of time, expertise and widescale buy-in, so SNIA entered the picture.
SNIA renamed Bluefin to SMI, which is not only spec development but also interoperability testing. In addition to SNIA-sponsored CIM SAN conferences and Plugfests, the Interoperability Test Program (ITP), allows different vendors to test that they're interpreting SMI-S specs properly, and how interoperable they really are. The process also feeds back requirements to DMTF to help with CIM and WBEM compliance. SNIA, which addresses specific storage verticals, maintains a close working relationship with the DMTF. Krantz commented, "Think of SMI as a program to get the management interface to find, test, be interoperable, and get it out to market as fast as we possibly can."
New storage products are starting with CIM development instead of APIs. Krantz said, "API sharing will still be a major piece of it, at least in the short term. There will still be APIs, but we're going to see that decrease. I don't think it will ever go to zero, but I think it will be minimal."