editor Derrick Harris spoke with Songnian Zhou,
co-founder and CEO of Platform Computing, about the company's role in
the Grid and HPC markets -- as Platform has a strong presence in both.
This is evidenced by the company's recently announced partnerships with
Microsoft, IBM and HP, and Zhou goes into detail as to why each of
these is important not only to the companies involved, but to the
First, I'd like to discuss the three
announcements Platform made this week. Can you give a brief recap of
the partnerships the company struck with HP, IBM and Microsoft around
cluster/HPC resource management?
Absolutely, I'd be happy to provide a recap. At SC05 Platform made three announcements:
- We announced an integration with Microsoft on their forthcoming Windows Cluster Compute Server.
- We announced an expanded relationship with IBM through an integration with their industry leading Blue Gene /L supercomputer.
- We announced that HP will now be offering Platform Rocks as part of their Linux cluster management solutions.
What makes these announcements noteworthy?
In the case of Microsoft there are a number of things.
Firstly, there's been significant interest in what their plans are for
high performance computing (HPC). By working with Platform as
their partner of choice for Grid computing, Microsoft has the ability
to seamlessly integrate within enterprise IT environments managed by
Platform LSF. This integration will allow LSF to seamlessly schedule
and pass jobs to and from Windows CCS. The relationship will evolve as
we work together on a Web services standard for job scheduling. In
addition, Microsoft is going after the lower end of the market, which
opens up new opportunities for Platform.
In the case of IBM, we are at the opposite end of the HPC spectrum with
our integration to IBM's Blue Gene /L. Powering many of the
world's top supercomputers, Blue Gene /L is a powerful and scalable HPC
system. Through integrating with LSF, Blue Gene users now have the
ability to concurrently process multiple jobs of variable sizes, which
in turn improves utilization and decreases turnaround time. In
addition, Platform LSF also extends advanced workload management across
With HP, what we're seeing is an expansion of our relationship to
incorporate Platform Rocks -- our industry-leading cluster management
solution based on NPACI Rocks from SDSC. HP will now offer this
solution on their Xeon-based Cluster Platform 3000.
What do these partnerships with major IT vendors mean to Platform? What do you gain?
These partnerships come natural to Platform. When we
announced Platform Enterprise Grid Orchestrator (EGO) in August, we
took a major step in expanding Grid technology adoption into enterprise
IT. Mainstream enterprise is much different than HPC, and we realize to
be successful we can't go it alone. \ Over the last six months, we have
expanded our partner program and broadened relationships with the
industry's leading IT providers such as IBM, Dell, HP, Microsoft,
Intel, SAS and VMWare, with many others to come. By building and
expanding these relationships, we have significantly strengthened our
ability to deliver solutions to enterprise IT.
On the other hand, what do these companies gain from partnering with Platform?
As these companies look to move into
adaptive/utility/on-demand computing solutions, Grid is a key enabling
technology in the delivery of these solutions. In most cases, it's more
efficient to partner with an industry leader in this area than develop
a homegrown solution. Platform Computing has 13 years of experience,
1,700 large enterprise customers and a reputation as a leader and
pioneer in Grid computing. These factors make us the right choice to
work with in this area.
Moving away from standard HPC and toward Grid computing, I'm
wondering what Platform is up to with its Grid business. Can we expect
some customer announcements around Platform EGO in the near future?
Absolutely, 2006 is poised to be a banner year for
Platform Computing. The customer interest in enterprise Grid is
escalating and as we bring to market Platform EGO-based products in
early 2006, we expect to make a number of new customer and partnership
announcements. Our timing is spot on and Platform EGO is a unique
solution that differentiates us in the marketplace.
Speaking of customers, are you seeing any encouraging, or disconcerting, trends in adoption of Grid computing?
Yes, we're seeing a number of encouraging signs in the
adoption of Grid computing. The HPC market remains strong and continues
to grow, but we see a new opportunity emerging on the horizon in the
enterprise. With many large businesses enjoying healthy ROI from
Grid computing, we're now seeing their enterprise IT departments
getting in on the action. These initial steps are being driven by a
number of factors in the organization. In some cases, we're seeing
server consolidation and the need to improve utilization as a driver.
In other cases, virtualization is the jumping off point for enterprise
Grid technology. Lastly, were seeing many large applications players
integrate Grid technology within their products -- SAS is one such
example, embedding Platform EGO technology into their recently
announced Grid Manager product. All of these signs are
encouraging and point to ways enterprises are beginning to leverage
Grid technology within the data center.
What are your feelings on the importance of widespread
standards adoption? Do you feel they are necessary in order to
establish widespread adoption of Grid technologies?
The adoption of industry standards is key for enterprises
to leverage the benefits of a heterogeneous IT environment.
Platform is committed to the support of industry standards and have
architected our products to take advantage of them whenever possible.
In the case of Platform EGO, we will take it one step further by making
a standards-based toolkit, heavily based on Web services, freely
available to the developer community so they can easily customize
integrations within their environment. We are also an active
participant in several of the key standards bodies for Grid.
The Platform Web site says that the Platform Globus Toolkit is built on GT3, are there plans to switch over to GT4?
At this point we're not seeing demand from our users to
make this change. When this changes, we will look at making the switch
to GT4. Platform has a longstanding relationship with the Globus
consortium and will continue to support our customers that use the
I've heard on numerous occasions discussions about the
differences between cluster and Grid computing. Seeing as how Platform
deals in both of these areas, and seeing as how SC really brings these
(along with other HPC) technologies together, can you give your two
cents on the differences and similarities between clusters and Grids?
That's a great question and one I get regularly. A cluster
is usually a group of servers inter-connected to look like a single
server environment. A cluster can be large scale and heterogeneous, but
is a single management domain. A Grid is when you take a group of
clusters and link them together seamlessly for widespread sharing. To
draw an analogy, if the CPUs and nodes were leaves, then a cluster
would be a tree, whereas a Grid would be the forest.
And, speaking of SC, are you looking forward to this year's
incarnation? What is it about SC that draws so many attendees from
pretty much everywhere on the planet?
SC is a unique event in the IT industry. It's the
industry's premier event in high performance computing and really a
landmark activity in assessing the state of the industry each year. At
no other HPC event do we get such a great collection of customers,
partners and industry experts. I am especially looking forward to this
year's event. My prediction is that we will start to see more
people from traditional enterprise IT in attendance this year. With HPC
and Grid beginning to expand into the data center, there's a new
audience that's eager to learn about the benefits of this technology
and how they can leverage it.
Is there anything else you'd like to add?
Just that I look forward to seeing everyone at the show and encourage them to stop by the Platform (#2124) booth to say hello.