I recently attended the NRENs and Grids Meeting in Dublin, kindly hosted by Trinity College. It gathered together a European audience of those involved in providing national education networks (hence the NRENs bit) and those involved in developing grid software and hardware. The JISC interest in this event was that we are currently working on a number of projects and programmes with a grid related element (such as the e-Infrastructure programme and new work that we are currently formulating under the capital programme).
The programme for the event can be found here and the slides from the presentations at the event can be found in links next to the programme item. I’ll not repeat what is on the slides in this blog entry; I’ll just point to the presentations of particular interest and comment on why I found that particular presentation interesting.
Day One – Grids
The first day focused on developments in grids. The session on eduGAIN was particularly useful in covering how eduGAIN works; it’s quite a complex system but very effective so I’d recommend using the presentation as a 101 if you’re new to it. Items of interest were that eduGAIN are going to be reviewing using Shib 2.0 and future developments also include non-web-based apps. Both of these are areas that JISC is actively involved in so it would be worth following what is being done in eduGAIN.
The next presentation looked at easing access to grids via identity federations. This was of special interest as we are currently involved in doing the same thing through the SARoNGS project. This meant we had quite a lot to share with the group and after the coffee break Jens Jensen and I did a short presentation on what we were doing under SARoNGS, receiving some useful feedback and some good contacts to share software resources and use cases. My feeling is that this is a useful area to link up with other European countries on as there are common problems that can be more quickly and effectively addressed through mutliple groups rather than one group on its own. For example, we have an issue that the SARoNGS solution is constrained by UK Federation policy on passing a unique user name and sharing information between service providers, meaning it cannot be IGTF compliant and is a little less secure. Norway has similar issues and we resolved to review what could be done in terms of a possible future change to policy that would allow a better technical solution and that would still meet the original goals of that particular aspect of the policy. I also talked with Christoph Witzig of SWITCH and there is potential to work with them on aspects of MyProxy to make interoperability easier.
Authorisation developments in grids proved to be an interesting presntation by David Kelsey as it gave an insight into future work under EGEE. The main messages were that there was a scaling back of funding for EGEE that has led to a great deal more focus on specific elements of the infrastructure that need to be tuned and that there was now an expectation from the EC of member states funding grid work. The reduction in funding has meant that the technical work on middleware has been reduced and there has been a shift to focusing on the authorisation framework and an analysis of how authorisation could be more effective. There is a broader desire to have a common policy for VOs, which would then mean that trust in them could be brokered in a similar way to the way it is in IGTF.
To wrap up the day, there was a discussion session on what we all felt would be important to address around grids. The overwhelming part of the discussion focused on levels of assurance, something we have already looked at under the ES-LoA and FAME-PERMIS projects at JISC. The overall agreement was that this is an area that needs to be addressed to allow new users onto the grid using a lower level of assurance, such as those with a federated ID as opposed to a digital certificate. It’s going to be interesting to see what happens over the next year or so as members of the group grapple with this issue. There was also some discussion on attracting more users and new users to grids. It was generally agreed that we need to lower the bar slightly for those outside the traditional disciplines that use the grid (such as particle physicists and computational chemists). Current initiatives in Europe would suggest that many have joined JISC in looking at how this could be done and have been succesful, SWITCH being one of the early ones with its IGTF compliant VASH and SLCS solution.
Day Two – Virtualisation
Virtualisation is something we have looked at previously under the NGS but the time was not quite right. Day Two showed plenty of evidence that maybe it is time to go back to this area under the new round of capital funding to see what we can do.
Cloud Computing for On Demand Resource Provisioning looked at one potential method of providing virtualised resources in a grid environment. The concept was to have a virtualised layer to separate the virtual machine from the physical location. Ignacio Martin Lorente explained how the University of Madrid was trialling using OpenNEbula to be able to do this and hence bring into use machines that had previously not been on the grid as well as allowing for burst traffic by using resources such as Amazon EC2. I won’t try to explain how the whole thing works; it’s much better explained in Ignacio’s slides. Setting up VOs on these virtualised resources can take as little as 20 seconds for a standard setup, meaning that environments can be set up and maintained easily without having to rely on being on a physical server. Ignacio finished his presentation with a look at the RESERVOIR project under the EU Framework Programme , which is a 3 year 17m euro project to get a Next Generation Infrastructure for Service Delivery. I think both of these projects have interest for JISC and it was useful to have examples of how virtualisation could work within an institution and a broader initiative to get cloud computing working across Europe.
The presentation on the Challenges of Deploying Virtualisation in a Production Grid covered pretty much what it said on the tin. Stephen Childs went through how Grid-Ireland had worked on having virtualised environments in their grid environment through open-source software called Xen. He also covered the results of a survey he carried out to look at virtualisation. The key points to come out were:
- It is important to treat a virtualised environment in a production grid in exactly the same way that you would any other production environment. Some of the virtual machines are going to be up for a long time so need patches, etc in the same way as any other physical server;
- Virtualisation is gradually gaining ground and now there is a choice of VM software from commercial to open source, it is starting to become an activity that is being engaged in across European academic institutions. However;
- This activity is currently on a trial basis as people get used to what is involved in provisioning VMs as opposed to physical servers;
- There has to be an awareness of where I/O is critical as Xen is especially weak on this at the moment, meaning a virtualised server may not be the best solution;
- There need to be solid use cases for implementing virtualisation and it must be used appropriately. The two main reasons for not using virtualisation in the survey were management issues and performance;
- A VM host does not behave in the same way as a physical host in all cases – there may be issues with compatibility even if the setup is exactly the same;
- Monitoring is still quite flaky;
Finally, Stephen outlined how Grid-Ireland has used Xen to install, effectively, ‘grid in a box’, where institutions simply needed to host the box they were given and management was carried out by Grid-Ireland. This was a neat solution for the institution but involved quite a lot of overhead for Grid-Ireland on management.
I thought this was a good presentation and Stephen is a useful person to talk with further about virtualisation (as further discussions over coffee proved). He is going to look at putting the survey into a PDF format so that the results can be shared with others.
The remaining presentations covered physical infrastructure so, whilst interesting, were not quite as relevant to what we are doing in Innovation Group.
The final discussion covered future topics and certainly one that we raised was accessing data on the grid, which we are doing quite a lot of work on under the e-Infrastructure programme .
All in all, I think this is a useful group to keep in touch with as the topics they are addressing are ones that we are either currently working on or are interested in for the future. The event provided a good opportunity to meet with others working in the same areas and share experience as well as get pointers to resources that we could use at JISC.
My thanks go to our hosts at Trinity College in Dublin, who worked very hard to make sure the event ran smoothly, with particular thanks to John Walsh for booking an excellent venue for dinner and being on hand to offer local knowledge (he even guided us back to the hotel from the restaurant!).