Thoughtput

Super Structure

The GigaOM team hit the conference scene with a bang by hosting the Structure08 conference today. Here's a few things I enjoyed.

• The speaker roster and attendee list were top notch.
• The panels covered pertinent topics on the challenges of building web scale applications.
• It was a one day, one track conference. Densely packed and too the point.
• Coverage of the infrastructure supporting the boom in cloud computing applications is long overdue, as evidenced by the sell out crowd.

They have put practically the entire conference online. Best place to start might be the list of Live Coverage of Structure08.

Beyond the Component View

There is more news on the SSD front as SearchStorage reports that EMC will introduce another direct-attached memory solution by placing SSD drives within their Clariion arrays. This is similar to the announcement earlier this year that they would make SSD drives available in their high-end Symmetrix arrays. Back then we wrote a series of blog posts about that announcement including:

• Video: Using Memory to Speed Storage Performance - Jan 23, 2008 5:41:29 PM
• Quenching Data Centers' Thirst for Memory - Jan 17, 2008 5:05:44 PM
• The Flashtastic DMX-4 - Jan 14, 2008 10:30:21 PM

Just about every major vendor has announced some type of solution to add more memory to their systems. This is an inevitable result of growing application needs and the demand for greater performance. Sun, NetApp and others have all made noise about incorporating more memory into their server and storage systems.

Recently we added another video to our blog on Caching as a Network Service to showcase some of the deployment differences of a number of these solutions.

The great news is that more options for deploying memory in the data center are becoming available, and the component choices only assist in helping customers design effective solutions.

However, the wave of memory-based solutions goes far beyond the introduction of another type of SSD component. It is important that editors and analysts covering this new category of solutions keep an architectural perspective about how memory is being deployed, compared to the just looking at the component view.

The tricky part is that SSDs have been around for over 30 years. Since then there have been a whole wave of new memory-based solutions (many which make use of SSDs) that can reside within servers, within the network, or within the subsystem. In each of these areas, the deployment model be quite different. Sometimes memory will be deployed as a persistent media, often it will be deployed to assist and complement a persistent disk-based infrastructure.

Whichever the case, it is critical that we graduate from component-level coverage to system-level coverage.Our series on Memory in the Data Center is a great resource to get started.

Cloud Computing Show - Episode 4

Here is The Cloud Computing Show - Episode 4.

Guests:
Bert Armijo, Senior Vice President of Sales, Marketing and Product Management, 3tera
John Pozadzides, Vice President and Chief Marketing Officer, Layered Technologies

Hosts:
Gary Orenstein, Vice President of Marketing, Gear6
Steve Norall, Chief Technology Officer, TechValidate

Show notes:

Opening and Introductions

Hot Topics in the Cloud

    -EMC World (Mozy, Hulk & Maui)
    -why do people feel a need to trash the terminology? (see Scattered Clouds)
    -does cloud computing signal the end of the SMB market?

Company Introduction and Q&A - 3TERA

Company Introduction and Q&A - LAYERED TECHNOLOGIES

Overview and Q&A on Collaboration between 3tera and Layered Technologies

Cloud Computing Show - Episode 3

We caught up with Robin Harris from StorageMojo to talk a bit about Cloud Computing, what has to happen for greater adoption of cloud-based services, and what might not make it into the cloud.

As always, Robin has an insightful take on the industry, and the short podcast is definitely worth a listen. Enjoy!

Click here for the Cloud Computing Show - Episode 3

Using NFS for Virtualization

Our article on NFS virtualization benefits was just published in CommNews.

Running virtual machines on network file systems (NFS) provides a number of architectural advantages, starting with the fact that virtual machines use files to store their image information. Since network-attached storage (NAS) systems are built from the ground up for file management, the administrative time and resources to oversee a large number of virtual machine files is less than what is required to manually assign and place those files on individual logical unit numbers (LUNs).

Click here for the complete article.

Special thanks to Nick Triantos for providing some assistance with this article. Nick is the author of Storage Nuts & Bolts Blog. 

Freescale Frees Memory

In a sign that there is money to be made in the memory market, Freescale Semiconductor has spun out a division to an eager group of venture capitalists.

No doubt this will stir up another round of discussion about all of the different types of media and memory types coming to market. For more on the media front, see our series Memory in the Data Center - Part I.

The new company, EverSpin Technologies, will focus on MRAM which is different than DRAM. According to the NY Times article MRAM uses less power and is considered more stable.

It is helpful to see all of the memory technologies coming to market...a sure sign indication that our latency thresholds are getting even more stringent as users demand instant access to ever larger pools of information.

Video: Caching as a Network Service

The need for data center, memory-based caching solutions is rising. Today's application infrastructure simply requires too much information too quickly for conventional disk-only systems to keep up. There must be a way to deliver low-latency, rapid file access without a wholesale forklift upgrade of existing storage systems.

While caching is a near-universally accepted solution, there are often questions of how and where that caching investment should reside. In this short video (3m 42s) we outline Caching as a Network Service...and why we see network deployments as the most efficient and effective means to enhance existing and new data center solutions.

Enjoy! Click here to see the video directly on YouTube.

Google I/O, HP Goes 2.0, and VMware Targets Performance

Several interesting announcements popped up last week. Most notable was the flurry of stories around the Google I/O developers conference. We now know that AppEngine will include specific APIs to assist with image manipulation and rendering, along with caching.

We have long followed Google's innovative memory approach(es), including Google's Use of More Memory, More on Google's Storage Performance Gap and Google's Storage Performance Gap.

A couple of comments from the InfoWorld article highlighted:

...The company at its Google I/O conference in San Francisco also will expand access to the service and add APIs for caching and image manipulation.

...Google's caching API for App Engine will make it faster for developers to render applications. (source: InfoWorld)

This is further validation of the importance of memory and caching for data intensive applications across the board, but particularly in the rendering arena.

Elsewhere, HP announced that they were merging their high-performance computing and web computing groups to form a Scalable Computing Initiative. It appears that just about every major systems vendor now has a cloud computing or web 2.0 focused effort. We've talked about other offerings from EMC and IBM on our Cloud Computing Roundtable Podcasts Episode 1 and Episode 2.

Finally, VMware announced the acquisition of B-hive Networks...

With this acquisition, VMware will leverage the B-hive team and technology to offer proactive performance management and service level reporting for applications running within VMware virtual machines - on both servers and desktops.
(source: VMware)

We covered some of the potential concerns around performance with virtualization and captured a few notes in our post Overcoming Virtualization I/O Pitfalls.