The Office of the CTO has been exploring how to best enable application access to RDMA for those applications requiring the ultimate in high bandwidth, low-latency communication, which includes many HPC MPI applications as well as many scale-out databases and BigData approaches. Passthrough mode is the most straightforward way to enable guest-level RDMA. With passthrough (which we call VM DirectPath I/O), a physical PCI device can be made directly visible to the guest operating system running within the virtual machine. We published a research note showing that this approach delivers very good InfiniBand latencies (under 2us) and excellent bandwidths over a wide range of message sizes. There is a downside, however: Punching through the virtual machine abstraction in this way disables several...
Posts Tagged ‘ hpc ’
2013 Predictions: When Worlds Collide
It’s time again to dust off the crystal ball, cast the yarrow sticks, read the tea leaves, and share some thoughts about where the IT world is heading over the next twelve months and beyond. Because my focus at VMware is on High Performance Computing (HPC), I’ll confine my prognostications primarily to the growing area of overlap between mainstream Enterprise Computing and HPC — places where each world can benefit from the other in important ways. Before I start, let’s get one no-brainer prediction out of the way: 2012 will be remembered for Nicira , and not Nibiru . Yes, software-defined networking ( SDN ) will trump planetary collisions and the...
Centralizing HPC Compute Resources using Private Cloud
I met recently with a group of academic CIOs and IT Directors in Denver to talk about how virtualization and private cloud technologies can help lower the barriers to adopting a centralized approach to delivering HPC resources in an academic environment. Of course, the barriers and concerns — losing control of resources, fear of not getting ones “fair share” in a shared facility, being forced into a standardized and sub-optimal software environment, etc. – are relevant outside of academia as well. Because the topic is of broad interest, I’ve published the slides in SlideRocket and added an audio track that explains why virtualization should not be viewed simply as a tax one pays to move into a cloud environment, but rather as an approach...
HPC Performance in the Cloud: Status and Future Prospects
I spoke last week at ISC Cloud 2012 in Mannheim, Germany about the performance of HPC applications in the cloud, citing results from several studies. I have summarized the talk below and my PDF slide deck is available here . For full details of ISC Cloud 2012, I recommend the coverage at HPC in the Cloud . I gave my talk to share some specific data about HPC performance in virtual environments, but I first described what new capabilities can be gained by virtualizing an HPC environment. I felt this was necessary because in HPC discussions it is often the case that...
Low Latency on vSphere at VMworld 2012
I was excited to have one of our partners, Mellanox, extend a special invitation for me to speak at their Expo booth during VMworld 2012 in San Francisco. They had allocated a 30 minute slot for a Theater Presentation and promoted my presentation via nice giveaways and postcards distributed in nearby VMworld hotels. I used this opportunity to reiterate some of the messages from the OpenFabrics Alliance Workshop in Monterey in March 2012 and to report on further progress with the Office of the CTO’s explorations into supporting RDMA at various levels in the vSphere ESXi hypervisor. In the first section of the presentation, I highlighted various use cases for RDMA...
High Performance Computing with vSphere 5.1
We just announced a pile of new vSphere features at VMworld 2012 in San Francisco this week so I thought I’d take a few moments to describe several of those capabilities that will be of particular interest to High Performance Computing customers.
Summer of vRDMA
Last summer, Josh Simons wrote a blog entry titled Summer of RDMA , summarizing work done by our intern, Adit Ranadive , who looked at RDMA performance on ESXi using VM DirectPath I/O with QDR InfiniBand cards passed through to VMs. At the same time, I had been exploring how to exploit RDMA for accelerating ESXi hypervisor services like vMotion, and some ideas around how to enable RDMA within a guest while still maintaining the ability to vMotion such VMs. In March 2012, Josh Simons and I delivered a talk about RDMA on vSphere: Status and Future Directions . In the talk, I discussed several different altenatives to expose...
Virtualization, Italian Style
I’m in Fiuggi, Italy this week teaching a four-lecture course on Virtualization and High Peformance Computing as part of the Eighth International Summer School on Advanced Computer Architecture for High Performance and Embedded Systems ( ACACES 2012 ). I have about 70 students in the class, primarily from European countries.
The Case for IT Convergence
I gave a talk recently at the HPC Advisory Council Workshop at the International Supercomputing Conference ( ISC ’12) in Hamburg. The main purpose of the presentation, which was titled Ultra-low Latency in the Cloud: How low can we go?, was to share our InfiniBand performance results with the HPC community to begin to address the question of whether cloud computing can be useful for more demanding types of applications than the throughput-oriented technical workloads that are known to run well in a virtualized cloud environment. (The short answer is we believe it can.) Before I presented our performance results, I spent the first part of the talk explaining what I see as an...
Ultra-low latency in the Cloud: How low can we go?
I’ll be speaking this weekend at the HPC Advisory Council European Conference 2012 in Hamburg, just prior to the International Supercomputing Conference ( ISC ’12 ), which runs from June 17-21st. If you will be attending either event and would like to talk about HPC and virtualization, send me an email — “simons” at the obvious place.
