Indian IT Industry

Vayutuvan · Post by **Vayutuvan** » 11 Dec 2014 08:27

My (expert but not very expert) opinion is that it would pay (literally) to learn Python, R, Statitics (mathematical) and brush up on one's undergrad (if not CS it is better). Combination of that sort would always be unbeatable at any age.

Comer · Post by **Comer** » 11 Dec 2014 09:04

Thanks for the replies.
At this point am looking for diversifying skill set for the market. People are ok with my resume but wonder where would they fit me.
Singha saar, going by the description of either, I would imagine they are NoSQL dbs like dynamo,mongo etc? I always wanted to pick one from the list. Will pursue that. Right now, the layer at which I work is kind of opaque to the kind of dbs used.
matrimc, am doing some learning on python. Was wondering if scala would be useful?
rishi, though cloud is not much different from what i do, it has its own domain now.

Singha · Post by **Singha** » 11 Dec 2014 12:54

http://www.infoq.com/interviews/adrian- ... ces-devops

a interesting intview with adrian cockcroft, who was one of chief architects on netflix's microservices strategy to deliver streaming web services to thousands of parallel users. you will find pointers to interesting ideas and a book here. the book has a free pdf floating on the web btw as with most popular books.

Neshant · Post by **Neshant** » 11 Dec 2014 22:08

Singha wrote:http://www.infoq.com/interviews/adrian- ... ces-devops

a interesting intview with adrian cockcroft, who was one of chief architects on netflix's microservices strategy to deliver streaming web services to thousands of parallel users. you will find pointers to interesting ideas and a book here. the book has a free pdf floating on the web btw as with most popular books.

Without actually reading any of it, did he copy bit torrent strategy of swarm downloads from multiple sources?

Comer · Post by **Comer** » 11 Dec 2014 22:44

^^ That's actually a pretty dense interview! Am trying to fit his theories into stuff that is familiar like OLE/CORBA, RPCs and probably something equivalent to assert() etc.

And someplace he talks about how outgoing calls are blocking, that was surprising.
But like the Chaos monkey and Latency monkey. These things could be very useful

Singha · Post by **Singha** » 11 Dec 2014 23:07

no Neshant, its not peer to peer . netflix probably has one or two sites in US from where they have to serve the entire US subscriber base. and each page render could as he says need 100 smaller web services even before the movie streaming starts. we already see it in facebook and yahoo mail.....pages are split into many parts and apps go off and pull content for each part separately and somehow it all happens nicely together...well in yahoo mails case the main message area usually lags behind sometimes by seconds or sometimes fails and we try again.

Singha · Post by **Singha** » 11 Dec 2014 23:10

I have set myself a 6 month goal to attain mastery of c++(from medium level now), and learn java, python and ruby to a working functional level, apart from ofcourse reading up on all interesting stuff thats being cooked up using these basic tools.

huge tools like google kubernetes seem to entirely avoid mainstream langs and seem like a mix of Go and bash last i checked their github.

i know ppl in bay area are also learning Go!

Comer · Post by **Comer** » 11 Dec 2014 23:15

^^ Singha saar, mind if I tag along too? It is useful to have some one learn along side and keep milestones. Like running

. Anyway it is upto you!
BTW, the Go looks like Pascal

Singha · Post by **Singha** » 12 Dec 2014 00:22

i have not checked Go, like all things google it has brand aura around it..but looking at the backpak way my chrome browser works it might be underwhelming.

if you send me your outside junk id, i will create a google docs xls and share its url so we can both update progress there, share links etc. pretty much what i needed in terms of books was floating around as pdfs in weird sites...at this rate i dont know if anyone will care to write good books.....

Comer · Post by **Comer** » 12 Dec 2014 00:28

Pliss to send mail to Bee Arrr efff dot <myforumid> at geemail . We can sync up there. Good to start studying again!

Vayutuvan · Post by **Vayutuvan** » 12 Dec 2014 02:07

It is al in the libraries - be it Python (web frameworks, CMS), C++ (boost is the library lot of inards of which have made into the latest C++ standard), no idea about other langs - but I am a C affacionado (with a smattering of OpenCL) (and have programmed in languages which are almost dead now - COBOL, Algol, PL/I, Pascal, Basic Plus 2, Lisp, Prolog) with a healthy interest in Elisp (Emacs lisp), Python, and a language called Julia (which is seems great for scientoific software - similar to Python). please do not miss to learn iPython while you are learning Python. Pyhton right now is the language of choice for lots of Web frameworks, has good crypto, web, JSON, libraries and is the ideal glue language. Twisted framework is what is being used in most Scientific/Supercomputing center clusters.

One needs also to learn some domain specific languages like R which can integrate well with Python. For visualization, TikZ for Latex and graphviz for web presentation graphics. If one wants to develop image processing plug-ins for GIMP, then it is done using scheme (better lisp).

As always, Java and JS are important for business web applications.
Things have gone completely out of control at this point, IMHO. This is especially so with the Template Metaprogramming. I find that kind of metaprogramming kills both understanding and maintenance. I hope it is short lived (but probably not).

SQL is always a requirement in MIS/business applications/products.

Vayutuvan · Post by **Vayutuvan** » 12 Dec 2014 03:46

About boost, you get all the data structures, several algorithms, container data structures, partial bindings - currying - for pthreads purposes or otherwise, lambda functions (anonymous in place functions with parameters - much better than just #define macros), Qt integration etc. The library also has GPGPU wrappers, Python wrappers with which one can Python enable old dusty deck programs/applications. It is a completely no strings attached library with source code. There are also lexer and compiler tools as well as gramar definition and all kinds of tools to support DSL development. The source code can be used anyway you want - rip out what you want and use it any where, even commercial if you want.

Also forgot to say, for portable GUIs (Win, Linux, Mac) Qt is widely used. So that is a good skill to learn as well.

Singha · Post by **Singha** » 12 Dec 2014 04:41

>> Template Metaprogramming. I find that kind of metaprogramming kills both understanding and maintenance. I hope it is short lived (but probably not).

boost is a massive library and the kind of templates and static fns (classwide fns) they use are definitely hard to understand for someone starting off into c++. unfortunately most of the big shot object gurus love stuff like templates, mutators and stuff .... so there is no option but to understand.

Vayutuvan · Post by **Vayutuvan** » 12 Dec 2014 05:00

singha and saravana: May be design patterns is one way to go (but then you guys/gals might have already mastered that stuff). It was an entertaining read but very hard to connect to real life unless one has already done some OO programming <scatalogical>, It is very heavy reading in the abstract (i.e. if one is not on any definite project/aplication). One of the authors was an advisor to a few my batch mates. Several dropped out with an MS (but excellent programmers and are happily coding their way through life - in US, Israel etc.). This prof. is a very nice guy and was active and helpful on internal departmental Usenet groups. What was surprising to me at that time was that he never got a tenure at my Alma Mater - remained an Adjunct prof. for a long time (and probably retired now).

SaiK · Post by **SaiK** » 12 Dec 2014 22:53

design patterns is the way to go if the design in question has different interpretations to the solution. however, it is important to understand what the design is doing, solving what problem, and is the right design for the solution is what matters. no standard design pattern applies for specific problems.

this is my suggestion, don't go after design patterns, but design for the specific problem you are creating. then see if it maps to a pattern, and document that. during that process, you might end up refactoring or changing to accept the pattern approach or perhaps end up with a modified pattern for the rest of the team to follow.

bottom: don't chase behind patterns for designer's benefits. design behind for users benefits. that is what adds value and increase your kb-wala to survive technical debts.

Vayutuvan · Post by **Vayutuvan** » 12 Dec 2014 23:00

Saik: Probably they should listen to you thn moi. I can't architect out of a paper bag - not with the new tech/frameworks etc.

Comer · Post by **Comer** » 12 Dec 2014 23:13

matrimc saar, design patterns is the place I hit a road block/mental block. I am ok with using the classical data structures in C and understand the concepts in OOPS. But since I have no big experience in designing large projects using OOPS I can't really relate to them. And thanks for all the suggestions.
I have a thinking like SaiK too. Figure out what is the solution for this problem and try to fit into any known solutions. But I can understand where the patterns come from. If you have an amazing library like what java has(third party or otherwise) it is easy to talk in the language of design patterns so that it is easy to get a class off the shelf and more importantly everyone in the team speaks the same language. Sometimes I need to sync up with teams which use java. Once they get to the design level, it might as well be greek to me.
After several years thinking using procedural languages, it needs brain reorienting to start thinking in OOPS. In college it was much easier. One other mental block, using sql like langs leads to loss of control

Vikas · Post by **Vikas** » 12 Dec 2014 23:16

Does anyone has any contacts in a company called MetricStream HQ'ed at Palo Alto and Bangalore.
I need to get my resume in but not thru generic job opening page of the company. Never have received response if I go thru Job openings section of a company.

Looks like it is a small compact company (was a start up earlier) which is working in area of my domain specialty.

Singha · Post by **Singha** » 13 Dec 2014 00:27

Mesosphere Turns Data Center into One Huge Computer
In June, venture capitalist Vinod Khosla said at GigaOm’s Structure conference in San Francisco that the most important opportunity in the business of IT was getting rid of all the IT people.

The opportunity is in building a data center OS that will automate resource management much like a computer OS does.

New investor Khosla Ventures led the recent round with additional investments from Andreessen Horowitz, Fuel Capital, SV Angel, and others. The company raised a $10.5 million series A in June, and its total funding is now approximately $50 million
…..
Mesosphere stands to make a major impact in the data center. The company formally launched its Data Center Operating System (DCOS) into early access earlier this week and plans public launch in the first half of 2015. It also recently raised more than $36 million in a Series B funding round, which the company will use to build out its engineering and sales teams even further.

….
DCOS is built on Apache Mesos, the tool famous for helping Twitter get a handle on data center operations and killing the “fail whale.” Mesos abstracts CPU, memory, storage, and other compute resources, creating virtual pools.
….

What’s in DCOS
The DCOS consists of a distributed systems kernel with enterprise-grade security, based on Mesos. It includes a set of core system services, including a distributed init system (Marathon), distributed cron (Chronos), service discovery (DNS), storage (HDFS), and others, all capable of launching containers at scale.

SaiK · Post by **SaiK** » 13 Dec 2014 00:45

let them go and do a dcos, a clever IT guy will hack in, screw them to hell.. there would be none to help them recover after that. sony is already hit its $hit on the fan.

Singha · Post by **Singha** » 13 Dec 2014 05:23

more dhoti shiver via a friend over lunch:

one of the founders of nicira(now with vmware) came to brocade(storage networking area) last week for a talk. he boldly stated none of the traditional networking cos will remain in business in 10 years.

he showed some pix of how a IBM server looked in 50s(huge) and now(tiny), and then a pic of netz CRS1 router(big) and a sw controller(pure sw, can run on any server). he said the churn that server industry went through in 60-70-80s, the networking industry is starting to go through now...

---
vmware NSX (really nicira acquisition) is a totally sw defined datacenter control system on any hw they claim. they called it SDDC - sw defined data center.

SaiK · Post by **SaiK** » 13 Dec 2014 05:54

yup.. SDN is the way to go.. chacha network has put in about $3b for SDN. what is interesting is, its north bound api to manage networks, and completely provisionable over network. i can on demand ask for 100 node network configuration, and then change it to 400 node network dynamic config, with guaranteed SLA on bandwidth and response times. nec was ahead in the game of controllers on this (1 year back info)

Singha · Post by **Singha** » 13 Dec 2014 06:25

one mysterious co which seems to have deep connections everywhere is "NTT data"....i see it in india, i see in SV, iirc have even seen it in D.C. somewhere.

Vayutuvan · Post by **Vayutuvan** » 13 Dec 2014 07:51

Isn't NTT Nippon Telephone and Telecom?

Singha · Post by **Singha** » 13 Dec 2014 10:14

Yes. But why are they everywhere?

rajpa · Post by **rajpa** » 13 Dec 2014 10:40

NTT operates a body shop in a lot of countries.

ArmenT · Post by **ArmenT** » 13 Dec 2014 13:31

saravana wrote:But I can understand where the patterns come from. If you have an amazing library like what java has(third party or otherwise) it is easy to talk in the language of design patterns so that it is easy to get a class off the shelf and more importantly everyone in the team speaks the same language. Sometimes I need to sync up with teams which use java. Once they get to the design level, it might as well be greek to me.

Design patterns are actually a different bird than a language with a lot of libraries like Java. The whole idea with design patterns is that given a particular class of problem, the pattern comes up with a general purpose solution that may take a little longer to implement initially, but avoids painting yourself into a corner, should the requirements of the problem change.

One of the first things you learn in design patterns is to give preference to composition over inheritance. So, all that stuff about inheritance being the best thing since sliced bread that your lecturer taught you in class, comes crashing down when they start talking about real world issues and how inheritance can sometimes get in the way. After reading one of those books, I understood for the first time why interfaces (a.k.a abstract class methods) should be declared.

By the way, some of the base java library classes are examples of how to do things the wrong way. For instance, look at the Observable class

. Because of the fact it is a class that has to be inherited from, rather than if it was an abstract interface, it becomes near useless for most practical code. Another one that is badly busted is the standard DateTime class. Remind me to b*tch about how it is badly designed sometime.

Javee · Post by **Javee** » 13 Dec 2014 13:32

Singha wrote:one mysterious co which seems to have deep connections everywhere is "NTT data"....i see it in india, i see in SV, iirc have even seen it in D.C. somewhere.

NTT is like tier-2 desi company, IIRC they purchased Intellidata or something and do a lot of body shopping.

negi · Post by **negi** » 13 Dec 2014 15:10

NTT's major acquisition iirc was Keane Inc. mostly a IT services company.

negi · Post by **negi** » 13 Dec 2014 15:20

GD who are the major players that have major offerings which control the first 3 OSI layers ? Doesn't Netz rule the roost there too ? Software/Virtual Network guys cannot disrupt that segment right ? Someone still has to make a switch which controls the physical movement of data/packets. My brother used to work for Ciena they specialize in optical switches over a period of time their team has reduced and most of the chaps are now in Netz/JNPR but the company is still chugging along ok.

Singha · Post by **Singha** » 13 Dec 2014 19:59

Hw teams are facing huge cuts everywhere incl netz the board design and asic fpga types.
Brocade let go most hw last yr and kept a few guys to glue together reference designs from merchant vendors like Broadcom.

Coming to optical it looks safe. Data centers need even fatter pipes to move around stuff and workloads vs the smaller enterprise server farms of yore.

In pure play routers nd switches Huawei types are offering lots of options at 10% margin something which no other vendor can live on..not with bau area costs and honest accounting.
None know how much real profit cheenis make as they have govt loan support and guangxi.
But they have made vast strides in portfolio nd topline.

Likes of vmw nd next gen boys say hw is irrelevant..just buy cheapest disparate hw nd let sw brain control everything like vmw nsx vision.open apis like openflow and ovsdb permit control nd data plane pgming by third party sw outside the boxes.

Netz is saying let me sell u both sw apic controller and hw nexus9k for end to end control nd visibility into hw layer.
This is aci project. Application centric infra. But apic gui in browser is like 20 yrs behind in look n feel...typical netz failing asking guys born into cli to go gui. Fresh ideas nd apple vmw gui people needed there

Singha · Post by **Singha** » 13 Dec 2014 20:02

As with netz, arista alalu Ericson et al are in same boat.
Hype machine is screaming hw is passé like jsf says agility is old hat

vishnua · Post by **vishnua** » 14 Dec 2014 05:33

negi wrote:NTT's major acquisition iirc was Keane Inc. mostly a IT services company.

Yes. Caritor, desi tier 2 company bought Keane but the kept the name Keane which is unusual but somehow all the Caritor guys in mgmt (Desi) by Keane( US white) got pushed out.

After that NTT bought Keane and got all the clients along with it. NTT also has quiet big data centers. Guess, NTT, "diversified" into SW services business.

Keane was miserable company to work for based on the feedback i received back in 2002 when it was just Keane.

SaiK · Post by **SaiK** » 14 Dec 2014 21:34

i would choose an 80-20 offer over kean offer.

Sridhar K · Post by **Sridhar K** » 15 Dec 2014 13:28

Workforce optimisation on at TCS; hiring plans on track

http://www.moneycontrol.com/news/busine ... ef_article

Javee · Post by **Javee** » 15 Dec 2014 23:25

I guess TCS is cutting down the bloated middle, the other 3 will follow suit as the market dynamics change.

Singha · Post by **Singha** » 19 Dec 2014 07:08

all mujahids wanting to get in tune with algorithm heavy interviews, there is a free part1 and part2 algorithms course in Coursera by tim roughgarden from Jan19 - May. it covers dynamic programming also in part2. but its best we start from part1 itself.

so get the login, register for the course and then click here to see preview of the lectures. I am astonished by the karatsuba multiplication take on the simple multiply thing last night.

https://class.coursera.org/algo-004/lecture

clearly any work on scaling and lord knows everything is expected to scale to insane extents these days needs deep knowledge of how to structure for it and find the right algorithms. this course might help set us thinking in that direction in more structured way than usual seat of pants approach.

vivek_v · Post by **vivek_v** » 19 Dec 2014 07:25

Probably OT but If one actually wants to learn algorithms in any depth then Robert Sedgewick's algorithms in Java would be a good book to start since it has lesser amount of math. If one really wants to push into Bigdata analytics, Supervised/unsupervised ML (and can digest the math) then books like Intro to algorithms by CLRS + some Linear algebra would be required. The reason is that while knowing what algorithms to be used where is essential, it is equally important to be able to create derivatives and mix and match of existing algorithms for which proofs and ability to create new time and space complexities would be essential.

I would have to be honest in saying graduates from Khan universities with their rigorous Design and Analysis of algorithms courses are better placed for the upcoming areas than most of our IT conscripts. It is terrifying to think on how most of current IT would look like in the future.

Singha · Post by **Singha** » 19 Dec 2014 08:33

One od the intro lectures in link above lists book refs all of which are legally free pdf.
I did study discrete math in second yr of engg but we were never taught any real world use case.
So forgot it long ago.

Comer · Post by **Comer** » 19 Dec 2014 08:53

Singha, signed up for that, thanks. If you are interested in Python, have a look here;
http://www.pythonlearn.com/