Storage Informer
Storage Informer

Tag: Source

Software installation is only the beginning – Ranking and Relevancy

by on Jul.07, 2010, under Storage

Enterprise Search: Software installation is only the beginning – Ranking and Relevancy

EMC logo

One of the most common misconceptions I hear almost on a weekly basis is that search is just a install and go piece of software. The simple fact is this just isn’t the case if you want to maximize you return on investment and improve your content’s findability, configuration is essential.

At it’s highest level the success of a search engine is it’s ability to return relevant results and ranked in a reasonable order. To do this a search expert typically will create what is known as a ranking and relevancy model. If your site or application has any degree of personalization/user types, multiple models may be required. (Some search engines attempt to do this automatically with varying degrees of success).

So I’ve got the ranking and relevancy model defined, I can leave it now right? Alas No. As you add new content sources to the search engine or the content changes, as should the model. So the model should be regularly maint ained. I’d recommend the following:

  • First month since go live (Weekly updates)
  • First 3 months since go live (Every 30 days)
  • First year, at least review the model every 90 days
  • And update the model(s) if any new content source is added

Sometimes the process of maintaining the model is an hour of work, other times it could be in the 8-16 hours range, it really depends on how much the content has changed since you last updated it.

So what’s the value of this and why should your company spend the money in maintaining the ranking and relevancy model?

Put simply it’s the difference between good search results and bad ones and from a business point of view it all ends up as  return on investment. If you don’t update the model regularly, the hundreds of thousands, perhaps millions of dollars you invested will not generate a significant return. You users will complain about search, IT will look bad for a “failed” p roject and the board will wonder what the money was spent on.

To give you an example I hear from a new potential client who says “We have vendor a, but it doesn’t work after we spend $xxxxxxx of dollars, so now we want vendor b”. To which I always reply when was the last time the search engine was looked at and the ranking and relevancy refined. The answer I always get is “not since we installed it x years ago”.

So trust me, maintain your search engine, you won’t regret it.

Update your feed preferences

URL: 69732d6f6e6c792d7468652d626567696e6e696e672d72616e6b696e672d616e642d72656c6576616e63792e61737078&i=70726f78793a65383437633065372d333864392d343563302d623539332d3536373437333033653038383a3137343633

Leave a Comment :, , , , , , , more...

IWJT, Expo 2010 and Shanghai, China

by on Jul.01, 2010, under Storage

IWJT, Expo 2010 and Shanghai, China

I’ve recently returned from the International Workshop on Junction Technology in Shanghai China. Shanghai is a fascinating city (or mega-city). The Chinese say that Xian is the city of the past, Beijing is the city of the present and Shanghai is the city of the future. I certainly agree with that characterization of Shanghai! From the LED-dominated night skyline to the new high-speed maglev train – Shanghai truly is the city of the future.

 (Picture from left to right is Sadanand Deshpande of IBM,  Raj Jammy of Sematech, Kelin Kuhn of Intel and Yun Wang of Ultratech).


The International Workshop on Junction Technology (IWJT) is a critical conference for the implant and annealing community. Why do we care? Well, in a modern CMOS device, one of the most critical parameters for improving performance is the parasitic resistance from the channel to the contact (see Figure, left). This parasitic resistance is the sum of Racc (the resistance in the source-drain extensions) + Rspreading (the resistance as the current fans out into the epi + Repi (the bulk epi resistance) + Rinterface (the resistance between the epi and the silide) + Rsilicide + Rccontact. IWJT primarily focuses on improving three of these resistance components, Racc, Rinterface and Rsilicide.


Racc is the resistance of the source-drain extension.  The percentage impact of Rext has been increasing each generation (see Figure, left). Rext is a very sensitive function of the implant and anneal technology. The goal is to achieve the possible junction depth (XJ) with the lowest possible resistance (Racc).   This is achieved by improvements in implant and anneal technology.





One of the highlights of the conference was the vigorous discussion on molecular implants (see Figure, left). The key idea with molecular implants is to use a large (or very large) molecule to improve the amorphization, reduce the penetration and increase the dose of the implant into the substrate. One molecular implant (BF2) has been around for a while – but the industry is increasingly moving to much larger molecules such as decaborane, octadecaborane, and carborane. While a number of talks discussed molecular implants, the keynote by Anthony Renau, the modeling study by Takaaki Aoki and the invited paper by Wade Krull all focused significant discussion on the benefits and challenges of molecular implants. 

Another of the highlights of the conference was the focused discussion on annealing technology. The key idea with modern annealing technology is to activate the implant with as little subsequent movement of atoms as possible (see Figure, left). While a number of papers discussed anneal technology, the invited papers by Lee on long-mS flash anneals, and Wang on laser annealing all focused significant discussion on the challenges facing anneal technologies in the coming generations.


While I was in Shanghai, I had the opportunity to visit EXPO 2010 for an afternoon and evening. While I can ramble on indefinitely about the various displays and pavilions, my major geek take-away was that Shanghai must have bought out the market on LED technology. After dark, EVERYTHING seemed to be lit up with complex programmed light shows.  It was virtually impossible to take a bad picture.




Stay tuned.   My next blog (mid-July) is on VLSI!

Comments (0) (closed)


Leave a Comment :, , , , , , more...

Riverbed Accelerates Pathway to The Cloud with New Release of Riverbed Optimization System

by on Jun.06, 2010, under Storage

Riverbed Accelerates Pathway to The Cloud with New Release of Riverbed Optimization System

Riverbed Technology (NASDAQ: RVBD), the IT performance company, announced today the release of version 6.1 of the Riverbed Optimization System (RiOS), which powers Riverbed’s award-winning wide area network (WAN) optimization solutions. With this release, Riverbed accelerates organizations’ pathway to the cloud by providing enhanced acceleration for cloud services and enterprise applications. These enhancements provide organizations more flexibility when determining the best approach for consolidating IT resources to public, private and hybrid cloud environments and ensure that organizations obtain maximum performance regardless of their network architecture. Full Story…


Leave a Comment :, , , , , , , more...

Looking for something?

Use the form below to search the site:

Still not finding what you're looking for? Drop a comment on a post or contact us so we can take care of it!

Visit our friends!

A few highly recommended friends...