Saturday Dec 09, 2006

"Rock tape-out just a few weeks away"

"A Rock design-completion milestone called "tape-out" for the chip is
just a few weeks away, Marc Tremblay, Sun's chief architect, said in a
meeting here Wednesday. The company is holding a contest right now: if
Sun engineers don't tape out the design by December 31, they'll all
have to wear a tie, formal attire that Tremblay suspects is lacking
from many of the designers' wardrobes."

This, if nothing else, should surely make the developers be on schedule for the "tape-out".

More at this article.

Thursday Apr 27, 2006

ZFS - production ready?

One of the nice things about ZFS is that it doesn't need fsck.

Here's a related joke, but first a question:

Q : Don't you think ZFS is production ready?

Check it out!


Tuesday Jan 24, 2006

Solaris Jumpstart basics

Solaris Jumpstart: Remotely installing Solaris OS over network, probably in many systems simultaneously.
-Possible to install an OS on many systems together, thus saving time and effort.

-OS can be installed on selected set of clients based on their configuration.

-No need to be physically present at the system(Client) to install the OS.

-The saved OS image can be patched or upgraded.

-Installation using Jumpstart is faster than it is from CD.
-Client(s) : System(s) on which the OS has to be installed.

-Boot server : Used to boot the client so that the installation process could start.

-Install server: Contains the OS image to be installed.

-Configuration server: Contains rules and profiles. Also contains sysidcfg file which provides information during installation which otherwise would've to be manually entered by the user e.g. timezone, terminal type, locale, etc.
A single system can, and often does, act as all the servers.
Assuming the OS image is in the Install server and everything is set, Jumpstart works like this-
User gets console access to the client(s) e.g. from a tip server, and issues command “boot net – install” at the 'ok' prompt (on Sparc systems) which starts the process:
-The client sends an RARP request on the subnet.

-The Boot server has in.rarpd running, so it responds with the Client's IP address.

-On getting its IP address, the client sends out a tftp request for boot kernel image.

-The Boot server responds with the boot image.

-Client sends bootp request.

-Boot server responds with the client's info from /etc/bootparams file which contains information about Install server.

-Client gets root partition, install configuration and OS image from Install server and Solaris installation begins.
Boot server and Client should be on the same subnet. Since the Client sends an RARP request in the beginning to get its IP address, the Boot server has to be on the same subnet as the Client, as RARP requests genereally don't cross a router. So, Boot server and Client should be on same side of the router. 'ifconfig -a' of both Boot server and Client gives their IP address and netmask values. AND their respective IP address and netmask to find the subnet. If it is same for both the Boot server and Client, then it is fine.
In the future, I will discuss how to set up Boot and Install servers, various files needed for Jumpstart installation, and more advanced options like custom probes.

Monday Jan 16, 2006

Security breach... that easy?

Today's InformationWeek has an interesting article about a security preassessment in an MNC. It takes the reader to breaches the hired security professionals were able to commit step-by-step, and no one in the company (apart from a few who were told about it beforehand) seems to have a clue what was going on.

While a real 'attack' of such kind wouldn't be that easy, it showed that the easiest to avoid issues were in fact the most vulnerable. e.g. Careless handling of passwords ... that company's CEO had his name as system password and his secretary wrote it down on a paper in her desk... two of the worst examples on how to maintain a password.

The full article can be read here.

Monday Jan 09, 2006

Cable, Blu-ray, and Java

Good news keeps coming these days...
At EEtimes there is a news.

"Spurred by a government mandate, cable operators committed to embrace the OpenCable Applications Platform, a middleware layer that has been the subject of debate and development for a decade."
"The decision is a coup for Sun Microsystems Inc., which will reap Java royalties on every Ocap box, and a blow for Microsoft Corp. as it seeks to embed its software in cable systems."
"In another coup for Sun, the Blu-ray Consortium voted in December to use Java software."
"The cable operators' Ocap move hands wins to Sun on several fronts. The company will reap undisclosed royalties on every Ocap box sold, and Java will become central to millions of next-generation digital TVs. That could help Sun sell more servers and software to run cable headends."

Friday Jan 06, 2006

Nifty Unix tools

Unix has always been rich in tools with the philosophy of "one tool
for one thing". Many of them work best when used in combination with
others. While there are dozens of very complex tools to do equally
complex tasks, there are some smaller but very useful ones.

I find the combination of 'more' and 'grep' most valuable in daily
usage. Does anyone agree? 'more filename | grep string' is something we
use almost everyday.

 Some tools are not as visible in normal use. 'ncheck' is one tool I
came to know about much later in my Unix usage. An example:

 To get info about a file like its permissions, owner, size, etc. we
just use ls -l. Let's get some info about vi file that we know resides
in /usr/bin/

bash-3.00# ls -l /usr/bin/vi
-r-xr-xr-x 5 root bin 194404 Jan 23 2005 /usr/bin/vi

So the file actually has 5 hard links.
What are the other 4 files that link to it?

Let's ls with -i option to get its inode number.

bash-3.00# ls -li /usr/bin/vi
89215 -r-xr-xr-x 5 root bin 194404 Jan 23 2005 /usr/bin/vi

So its inode number is 89215.

How to get the other four links now that we have the inode number i.e.
find all files pointing to this inode. This seems a tough one but
another tool 'ncheck' comes to the rescue. 'ncheck' is slow but does
its work. For our example:

bash-3.00# ncheck -i 89215
89215 /usr/bin/edit
89215 /usr/bin/ex
89215 /usr/bin/vedit
89215 /usr/bin/vi
89215 /usr/bin/view

So we found all the hard links to vi using simple commands.
Pretty cool, these tools are!

Thursday Jan 05, 2006

The real story behind the Unix 'survey'

Last month, there was a 'survey' on Unix server users' vendor preferences by a group named GCG. The result of the 'study' seemingly showed Sun slipping behind IBM and HP, and was widely picked up by media without going into facts per se. Later, Paul Murphy came out with the real facts the study obviously (or deliberately) ignored.

In short, the study used something called VPI to calculate a vendor's score. Out of 197 respondents selected for the survey, let's say Sun is primary vendor for 140 of them(as the largest Unix vendor). Most of the customers have more than one vendor. For any metric, if 120 out of total 197 vote for Sun as preferred vendor, it should be 120/197 = 61% ?

OTOH, if HP has 40 customers and for the same metric as above, if 44 vote for HP it should be 44/197 = 22%.

Now here is the interesting part. According to that 'survey' HP beats Sun in this metric. Why? They calculate something called Vendor Preference Index (VPI) which is even more interesting. They calculated it by dividing the number of votes a vendor gets by number who use the vendor's Unix.
Going by this, HP gets 44/40 = 1.1 VPI and Sun gets 120/140 = lowly 0.85!

So even though Sun has 61% of respondents giving it thumbs up as compared to HP's 22%, HP actually beats Sun if you go by VPI and the 'survey'!
The calculation goes on for many metrics and, as expected, results are heavily skewed against Sun.

Detailed study of the 'survey' can be read here from where all the above numbers are quoted, and as an example of media blindly reporting the results, read this.

Wednesday Jan 04, 2006

When I shot a tiger

On new year afternoon we visited nearby Bannerghatta national park, took up the tiger safari and got to see many tigers roaming freely in the park. Used the opportunity to take some closeups. Here is one :

About OP/N1 RPE

I didn't explain the role of OP/N1 RPE in Sun in my previous post. It stands for Operating Platforms/ N1 Revenue Product Engineering, previously known as Solaris Sustaining and Engineering. As Solaris 10 is now free for anyone to use, our team has the important responsibilty of bringing revenue for Sun by providing support for Solaris. We "own" the product from the date it is released to its EOL and are responsible for any issues that come up with Solaris. This can include minor problems like improper configuration to major ones like bugs and enhancements.

Approachability and Interoperability is what my group is called in OP/N1 RPE. A & I is responsible for any issues with Patches, Packages, Installation related like Jumpstart and Live Upgrade, Libraries, Solaris Utilities, etc. With the introduction of some cool features like Zones in Solaris 10 our team is busy fixing related teething problems on Solaris 10 and Nevada.

So far, I have worked on Jumpstart, patching issues, utilities, and Zones. Still on a learning phase with these technologies and have to learn a lot.

Tuesday Jan 03, 2006

More of Ultrasparc T1

James Governor of himself saw some customers from Germany going ga-ga over Niagara.

Monday Jan 02, 2006

UltraSparc T1 aka Niagara review

Just finished reading Anandtech's detailed (but incomplete as they promise to delve deeper into it) review of Sun's new UltraSparc T1 chip. Among other things, it says : "SUN's Ultrasparc T1 is little less than a revolution in the server world. How else would you describe a 72 W, 1.2 GHz chip that is almost 3 times (in SpecWeb2005) as fast as four Xeon cores at 2.8 GHz, which consume up to 300 W? " and "The words "paradigm shift" and "disruptive" technology have been abused so many times that we don't like to use them. But in the case of the T1 CPU, it wouldn't be exaggerated to say that it is the herald of a new generation of server CPUs, and that it has disrupted the server market." and "The SUN UltraSparc T1 simply wipes the floor with the competition when it comes to performance per Watt. According to this metric, the UltraSparc T1 is 4 to 12 times better." Looks like we have a winner here! More can be found at :-)


Greetings! This is Surender Singh Samant from Sun OP/N1 RPE aka Solaris Sustaining. As a result of my new year resolution about creating a blog gone public, there was no way other than to create it ASAP (before anyone could remind me). With so much happening around it should remain updated, hopefully. A brief introduction to myself : I have completed six months at Sun recently. During undergraduation, I used SCO Unix, and at Post Graduate it was RedHat Linux. Unfortunately, Solaris was just a server operating system at that time, with few of the students even knowing it was Unix (and the most popular one at that!). At Sun I was pleasantly surprised to find Solaris 10 to be so similar to Linux from an end user's point of view. To be precise, it should be the other way around as Solaris has been there longer. With free availability of Solaris 10 and opening up of the source code, students now should definitely be more aware and eager to work on Solaris. :-)



« July 2016

No bookmarks in folder


No bookmarks in folder