“The Collective Invention” By Magritte

Your CRM applications are running in a Force.com Cloud,  Your other webapps are running in the Google AppEngine Cloud,   Your customer facing webapps are running in Amazon EC2 CloudFront and your traditional portlets are running in the Enterprise Portal. Many of these applications  Portlets access backend databases through Enterprise DataServices and a middle tier.

One of the Portlets is running very slow. Where is the problem? Even worse, one of the portlets runs slow some of the time,  and beats SLA most of the time. Where is the problem?

Production issues like these have troubled Enterprises with n-tier applications for  years. Please click on the picture below to see a detailed diagram of a multitier  Cloud Application.

Most n-tier monitoring software is not able to track individual transactions- some only  provide very high level metrics, while others provide averages for transactions. This is not enough for 80 to 90% of real life Production Support issues.  If n-tier was hard, debugging Enterprise problems that are distributed across multiple Clouds is going to be even harder.  One key problem is that the different systems are owned by different Enterprise groups, and this leads to very long conference calls where everyone blames everyone else.  Since usually the problem is in only one or two parts of a distributed architecture, it is a strategy that even the least technically savvy can pursue and win.

“DynaTrace” with its PurePath technology may offer an answer.  It offers:

“Java and .NET transaction tracing for instant assessment of business impact and rapid problem diagnostics “

It helps you find the answers you need to resolve performance problems that matter to the business quickly and reliably:

• Assess the impact of a problem (WHAT?)
• Isolate the component causing the problem (WHERE?)
• Identify the root-cause of the problem (WHY?)

DynaTrace will work with all of the components described above.  The diagram above is highly simplified.  A typical Enterprise has  30,000 to 50,000 workloads, each consisting of a multitude of applications and services.  The numbers bring perspective to the analysis.   There are many application monitoring tools which monitor, but none can deliver value in terms of diagnosing  troublespots.

  (It is unclear if it can find trouble spots
 inside a Force.com application.

But it should work well with the webservice
 calls that the application makes. )

]]> http://soarealworld.wordpress.com/2009/12/24/who-came-up-with-this-collective-invention/feed/ 0 New Century Frameworks "The Collective Invention" by Magritte MultiCloud http://soarealworld.wordpress.com/2009/12/09/john-chambers-talks-tough-on-data-centers/ http://soarealworld.wordpress.com/2009/12/09/john-chambers-talks-tough-on-data-centers/#comments Wed, 09 Dec 2009 14:42:09 +0000 soarealworld http://soarealworld.wordpress.com/2009/12/09/john-chambers-talks-tough-on-data-centers/ ]]>

Cisco has landed many deals worth $20 million or more. This makes a lot of sense. The real value of UCS is in the ease of management, and ease of management is crucial when you have large number of servers.

"Cisco may not have the necessary breadth to be a true soup-to-nuts player. IBM and HP, for example, sell their own storage gear. "  I tend to agree. I have not heard much about their consulting partners and the offerings that they are creating.

I see a very big market for Remote Desktops in middle income countries.  There is concern over license compliance in these countries, and using unlicensed software may be rampant- but users are not necessarily satisfied with the experience. If Desktop software with a per CPU license, were to be used,  economics of Remote Desktop can be extremely compelling. 

Microsoft, unfortunately, has a per client device policy. This means every netbook or client device that uses the software on the server must be licensed. This reduces the value of having server side computing.

Oracle may agree to put mySQL in a separate business unit. This business unit may have a separate board, according to a news item in The Register.

Meanwhile the MariaDB fork is getting momentum.

Is this what customers actually want?

“As Ellison said in October, “TJ Watson’s IBM was the greatest company in the history of enterprise in America because its combination of hardware and software was running most of the enterprises on the planet. We think with the combination of Sun technology and Oracle technology we can succeed and beat IBM. That’s our goal.”

Do we want a return to the monolithic hardware/software combination with limited choices?

How will affect choice and price?

I just noticed that Ltech has announced an Integration between Google and Microsoft Sharepoint.  It is not clear but it should probably work with Sharepoint hosted in the Azure Cloud.

I will study it in detail and post about whether it is a live-live bidirectional integration, or just a way of migrating content.  I see great potential for a live-live bidirectional or on-demand synch between content in a proprietary CMS and content being displayed in the Google Apps Cloud.

This is a quick review of  this interesting company.

I first read about them in the Register:       “Compellent – the billion-dollar storage company?   Screw the recession up”

Count on El Reg to make good predictions-

Compellent is now  proposing to raise funds through an IPO.

3.34 million shares at $19.25 per share.  That should raise some cool cash to raise additional money for future sales and marketing efforts.

I think “Thin Provisioning” is the key feature that is going to drive additional sales. It is reviewed in detail here.

Just saw this in the register:

Cisco UCS price appears to be around 10,500 dollars for 192 GB of memory using 48 slots.

This is a very good price for a lot of memory, and may make this a system of choice for memory intensive applications like Remote Desktops, Memcache appliances etc..

Amazon EC2 has discovered the power of main memory databases- It wants to offer to 10 million transactions per day for under $20 using Times10 main memory database. My question: Why do I need to run it on Amazon to get such a low price? Over 3 years, $20 per day adds up to $20,000. A fractional share of a Times10 in-memory database on a new Dell or Cisco UCS server can deliver similar performance- with more control over the environment, live-live disaster recovery, persistence to a permanent Oracle store etc..

I can see this as useful as a temporary solution, but on an ongoing basis it is going to cost a lot more.

http://tinyurl.com/ykj48kw

http://www.networkworld.com/news/2009/100909-cisco-rack-servers.html?hpg1=bn

“UCS C250 M1 Extended Memory Rack-Mount server is two rack units, and has 384GB of memory and eight disk drives. It costs $10,339 and will be available in December.”

Phil Wainewright writes that: “Cloud is no place for amateurs”. He makes the case that recent outages at IBM and T-mobile run facilities proves that big companies overestimate their competence at SaaS and Clouds.

I do not see this as an issue of big versus small, I think it is a matter of how reliable and rugged a system you are able to construct. A small company or an Enterprise could create a reliable cloud with the right kind of fault tolerance built in.

http://blogs.zdnet.com/SAAS/?p=899&tag=nl.e539

Emotionally, incidents like these strengthen the argument for internal on-premises Clouds constructed using SOA Layer and VMWare based fault tolerance.

http://blogs.zdnet.com/SAAS/?p=899&tag=nl.e539