- casino online slots

SMI-S Now On 532 Products from 28 Vendors

by Administrator on July 10, 2008

According to this email, received today, SMI-S providers are now on 90-odd more products than they were at last boast in January of 2007. 

Team SMI:

It gives me resounding pleasure to announce the addition of Data Direct Networks to the elite group of companies who certify their products for compatibility with the SMI-S interface. For details on their products please see [URL redacted]

Twenty eight companies and over five hundred and thirty two products have now completed SMI-S CTP certification demonstrating the value of the CTP brand in attracting and winning customers. To certify your products for compatibility with SMI-S please consult: [URL redacted] 


Paul von Behren – SMI Chair
Symantec Corporation  

While I am sure that this “progress” toward “unified storage management the SNIA way” is warming the cockles of certain smissy hearts, I feel compelled to ask a question or two.  Mind you, I am not part of this “elite group,” but I am interested in all aspects of storage management.

First, why do we need SMI-S if we already have web services standards from WC3 — XML/SOAP and Java standards for interconnecting and managing resources and services in a network?  Application vendors are almost universal in their adoption of the WC3 standard, as are network folks, server folks, and now, with the advent of ISE from Xiotech, storage folks.  SMI, even if fully realized, only provides the most baseline level of discovery and status monitoring of storage gear, doesn’t it?  Why build such a mechanism, when a much simpler and more robust approach might be forthcoming sooner that can manage data across an entire system (server/switch/storage)?  Prospects for realization of management nirvana via WC3 seem much higher than management via SMI, since the former is being driven by application vendors and server virtualization players (heck, even VMware supports web services integration) while the latter is driven by makers of proprietary hardware stovepipes.

Second, how does SNIA address the fact that there are vastly different levels of functionality and information provided by the various vendors who have implemented an SMI provider?  How many times have I blogged here about uneven implementations:  vendor X’s provider gives less information than an SNMP MIB, vendor y’s provider is a kluge and pretty lame compared to its API interface, vendor z charges customers an extra $30K for its provider, etc.  Seems to me that this is not a harbinger of universal management at any level other than the lowest common denominator and at the highest possible cost.

Third, why is 532 magical, noteworthy or even statistically significant?  Weren’t something like 16,000 new storage products released to the market last year alone?

Forth, how can any “adoption momentum” being touted by the smissys be sustained when so many vendors I am visiting are reducing their participation in SNIA or dropping out altogether?  In the last few days, I have met with c-level guys in several storage companies who have said that they are cutting their contribution to the organization, no longer exhibiting at SNW, and exploring TechTarget Storage Decision shows instead.


{ 1 comment… read it below or add one }

Mike July 30, 2008 at 9:38 pm

What exactly are the web services standards you’re talking about? If there is truly something better in every way than SMI-S, available today, then why aren’t the SRM vendors using it?

You could replace your “web serivices” example with any other management protocol and your argument would be the same.

Take SNMP. It’s has been around for ages, it’s more lightweight, and it’s simple enough that it can be easily implemented in the embedded software of any storage device worth its inflated pricetag. And as you mention, it’s often has more features for managing $RANDOM_STORAGE_DEVICE than SMI-S. (which is complex to implement and may even talk to the device using SNMP anyway) As soon as you start talking about web services and Java, you’ve reached the complexity of SMI-S already.

SRM vendors used to implement solutions that tried to identify and use the proprietary MIBs. It was fast and it worked. Why did they stop doing that? (Too expensive to try to support every vendors’ device? Didn’t scale? MIBs changed with every new release of every device?)

Despite all the problems with SMI-S, I think one can appreciate the problem it’s trying to solve.

Thought experiment: imagine for a moment that SMI-S was a set of standard SNMP MIBs defined by an RFC, so simple that any vendor would be laughed out of the market if they didn’t implement it. (whether that is actually feasible is entirely another discussion) Would that solve your problem with it?

Previous post:

Next post: