NETWORK WORLD NEWSLETTER: MIKE KARP ON STORAGE IN THE ENTERPRISE
EMC's SMARTS approach to the enterprise
By Mike Karp
Today, we finish our overview of where EMC is going. Last time,
we were describing the SMARTS technology that EMC acquired to
give it root-cause analysis capability across the enterprise.
You will recall that for storage, SMARTS uses data collected by
ControlCenter (ControlCenter gets its info using both SMI-S and
ControlCenter normalizes the data and then passes it up the line
to the SMARTS InCharge software. At some later point, the data
will be munged together with data from other areas of the
enterprise to give managers a comprehensive enterprise overview.
Many companies are today using pieces of InCharge, but from the
EMC presentation I saw, it was impossible to tell to what
InCharge is still a work in progress, of course. Data collection
depends on "adapters" (ControlCenter is the storage adapter),
and in cases in which adapters have not yet been built, no
information gets into the system. Thus, the speed with which
these adapters can be brought online will play an important role
in determining what SMARTS can be smart about. If you are
looking for value in understanding the cross-domain
relationships that this software potentially reports on, find
out about the schedules for those adapters that will be most
meaningful for your sites before you make any decisions.
Based on what I heard during the meeting at EMC, InCharge is not
rules- (or, if you prefer, "policy-") based, but rather relies
on looking at the network connections and then interpreting
events by looking at the symptoms that the events have spawned.
At the heart of all this is a good old-fashioned table look-up
that correlates problems and symptoms. InCharge first interprets
those symptoms based on the likelihood of certain problems being
the cause; then comes the nifty part (and, presumably, the
SMARTS intellectual property), in which probabilities are
weighed to select a likely root cause.
Probabilistic approaches such as this are somewhat rare and,
because they rely on statistical analysis, are likely to have
inherent shortcomings in cases in which the causes of a
particular set of symptoms are statistical outriders (that is,
they represent an unusual case). In other words, the rarer the
root cause, the less likely it is that the root cause can be
identified. This may not be much of a problem in established
networks, but may be much more problematic in more volatile
Hopefully the SMARTS correlation technology can account for this
- after all, SMARTS has been working on it for quite some time.
At this point the software has no heuristics for learning on its
own, however it may be a good bet that something like that is
in-plan down the road.
Suffice it to say that all this is complicated stuff and not a
subject for the faint of heart. But if you are serious about
keeping an eye on one of the most dynamic players in the storage
area, the SMARTS technology is certainly one you should track.
It, along with the other technologies referenced earlier in this
series about EMC (plus EMC's VMware and its consulting practice)
mark EMC as a company readying itself to emerge as a major
player across the enterprise.
Copyright Network World, Inc., 2005