Wednesday January 17, 2007
Availability Reporting using SunMC Performance Report Manager
Availability statistics can be generated using SunMC PRM Uptime reports. Here I ran some simple Uptime Status report across a few system for the last two years or so and have some interesting graphs in front of me. The timeline graphs have red areas indicating down-time and blue area representing uptime. The downtimes here included both planned (shutdown, init 0, haltsys etc.) and unplanned (power outages and other events where no shutdown command was issued) downtimes. Rest of the story is best conveyed by the picture itself...
Posted at 04:30PM Jan 17, 2007 by adikhit in General | Comments[0]
Friday January 12, 2007
Cool servers, cool graphs
Just thought I'll plot using SunMC Performance Report Manager a graph displaying system load, Overall CPU utilization (across all threads), Current consumed and the Processor Temperature. Makes an interesting graph and provides insight into capabilities of the management tool as well !![Read More]
Posted at 07:41PM Jan 12, 2007 by adikhit in General | Comments[2]
Wednesday November 15, 2006
Super Computing '06 - Day 4 (yesterday)
From a distance Cluster Management appears to be a different problem, looking closely into it one can see similarities with management of an enterprise datacenter management, except that the ingredients used will be mostly homegenous type of system and operating system in case of an HPC cluster. Though heterogenity is considered good, it doesn't appear to be a requirement at present. The building blocks today are commodity hardware components, I think the problem of heterogenity will hit HPC clusters when clusters grow over a period of time and the choice of hardware, software may be best met by some other commodity component in the market to run either the same app or a more powerful app. on the same cluster. This will obviously force existing solutions like rocks, ganglia to be ported to those newer environment. Again extensibility and integrability appear to be an afterthought with almost all present solutions out there including the best ones. I am happy for the fact that our focus at Sun is in the right direction at the word go. Yesterday was mostly spent looking around at what others are doing and is it possible for example to integrate two grids atleast from management point of view...
[Read More]
Posted at 04:40PM Nov 15, 2006 by adikhit in General | Comments[0]
Monday November 13, 2006
Super Computing '06 - Tampe Florida - day 3
The highlight of the day was a talk by Andy, on "Road to petascale computing". He covered the topic in a very simplistic way, as to how such systems can be built and the challenges that lie ahead of us during and after building such such a system, areas where improvements will help and other areas where improvements will not help as much. Later in the day we moved to Tampa Convention Center in downtown Tamp to participate in SC'06 main event. Microsoft is right at the gate :-) unfortunately there aren't many impressive things on the stall. Inside the hall the heat is really on, and literally in one corner where switches from many vendors are staged and connected, standing behind them makes one warm and comfy in a matter of moments.[Read More]
Posted at 09:41PM Nov 13, 2006 by adikhit in General | Comments[0]
Sunday November 12, 2006
Super Computing '06 - Tampa - Day 2
After hearing so many cool things on day one at HPC, I felt that we have covered it all and Sunday will be a repeat of all the same story. Turns out that was not the case, day-2 was equally interesting and was full with equal amount of energy & intensity in the participants and covering a lot of new ground. IMHO highlight of the day was presentation from the TiTech team extremely powerful content! What we do at Sun is being used in an extremely useful way.
[Read More]
Posted at 10:24PM Nov 12, 2006 by adikhit in General | Comments[0]
Saturday November 11, 2006
SuperComputing '06 - Tampa Florida
My initial thinking about HPC (High Performance Computing) was a little
different, all geek stuff ! It still is. Have been interacting
indirectly through some of our cool customer facing engineers in the
last few quarters where we were working on setting up System Manager in
some of the bigger envionments, the fact had however not sunk in till
as recent as a few weeks back. The realization was a satisfying one !
The math of large number has always fascinated me, scale teaches you
one thing among many others, which is, "designing for scalability can
not be an afterthought". The need for scalability has to be there right
from inception and, this is what Sun is about. Starting from Operating
System Solaris all the way down to the processor SPARC (and everything
in between) you can see scalability. Not content with what's there,
engineers at Sun have pushed the envelope in all direction from scaling
Processors (CMT) to connecting multiple systems (GRID) and again
everything in between. Listening to some of our senior executives and
correlating the work we are doing at System Manager and its relevance
in HPC space has been enlightneing on a Saturday (Yes its a weekend but
I'm enjoying listening to all the good stuff). I am already looking
forward towards the session on Sunday and beyond...
Posted at 08:33PM Nov 11, 2006 by adikhit in General | Comments[0]
Wednesday October 18, 2006
Understanding Report PDUs/ SNMP Exceptions in SunMC agent.log
SNMP handles errors by either reporting them as part of error-status and error-index field or can generate SNMP exceptions/ Report PDUs (Protocol Data Unit) with version 2 and 3 of the protocol. Whenever SunMC agent responds back with an exception in response to a client request it also logs the report OID (object identifier) associated with that exception in the agent.log for later viewing/analysis. This post talks about the meaning of various such OIDs returned as part of those thrown SNMP exceptions. [Read More]
Posted at 10:55AM Oct 18, 2006 by adikhit in SNMP | Comments[3]
Friday October 13, 2006
Enabling/ Disabling rows in an SNMP table
SNMP tables that support RowStatus primitive can toggle a row to
enable, disable state. It is also possible to add a new row and delete
an existing row
by manipulating value of this parameter. SunMC agents offer the
RowStatus primitive as one of the primary syntax for a node in a
table. The RowStatus primitive is implemented as part of agent
infrastructure as specified in RFC 2579.
[Read More]
Posted at 02:34PM Oct 13, 2006 by adikhit in SNMP | Comments[0]
Monday October 02, 2006
SNMPv3 in Sun MC agent
Something that came at CEC 2006, this was regarding SNMPv3 support and opening of the agents for a direct access. SNMPv3 support was added to Sun Management Center a while back with the 3.6 release, this was done with the intent of enabling direct access in a secure manner using SNMPv3usm for applications that wanted to bypass the SunMC server layer. In the past the agents supported SNMPv2usec but there are not many SNMP client stacks out there talking SNMPv2usec making the agent not so easily accessible, not so anymore since January 2005
[Read More]Posted at 10:29PM Oct 02, 2006 by adikhit in SNMP | Comments[0]
Tuesday September 26, 2006
Getting the OS version using SNMP
Many a times we need to know which OS is running on a particular
server. One can either maintain a page where the information can be
kept updated by the admins or one can write a script that can query
servers on the network which in turn can update some web-page. An
extension of this can be to convert the script into a Sun Management
Center module which can periodically update the information in a more
structured and predictable manner. A simple script is very easy once we
know which SNMP Object Identifiers to go after and that we will explore in this writeup.
[Read More]
Posted at 02:06PM Sep 26, 2006 by adikhit in SNMP | Comments[0]
Wednesday May 10, 2006
Sun Management Center
Sun Management Center Agent Modules - a better way to manage using Sun's SNMP agents.[Read More]
Posted at 01:50AM May 10, 2006 by adikhit in SNMP |