ESNet, The Energy Sciences Network

matt_c

 

Matt Cook
Head of Infrastructure and Middleware
Loughborough University
Chair of UCISA-NG

 


2014 Technology Exchange – Day 3

One of the features of conferences outside of the UK and especially prevalent in the USA is early morning sessions. It was time on day three to hit the 07:15 morning working group/camp/BoF sessions.

Unfortunately the ‘Campus Cloud Architects BoF’ was cancelled, which was really disappointing and not a good start, as I was hopeful to explore in person some of the latest concerns, trends and experiences in this area.

Industry groups have been reporting more and more interest in Cloud brokerage solutions and some companies are now recruiting for cloud broker and cloud architect roles. As cloud technology gets more mature, there is an opportunity to start brokering for the best possible service and cost for your organisation. In the sector we have seen an excellent start in this area by Janet with their agreements with Microsoft and Google for their email/applications suite.

There is a lot of development scope in this area with Microsoft Azure, AWS etc and I’m interested to explore the strategy required to position infrastructure, automation and standards to take best advantage of the emerging competition.

Perhaps this area is something that colleagues in the UCISA IG may be interested in picking up in the future?

I took advantage of the programme changes to share more details about the current UCISA activity in the ad-hoc groups using a short five-slide presentation covering these pieces of work:

• A guide to the implementation of an Information Security Management System (ISMS), launching in early 2015

• An update to the popular  ‘Exploiting and Protecting the Network’ document, launching in early 2015

• The Major Project Governance Assessment Toolkit

• UCISA Report 2013 – Strategic Challenges for IT Services.

There was a lot of interest in these areas and I had a couple of questions about integrating the planning, effort and joint working of UCISA and EDUCAUSE where there are clear overlaps and topics of interest.

The Energy Sciences Network ESnet are also interested in contributing to the Network Performance and QoS ‘community of practice workshop’  which the UCISA Networking Group are planning in January 2015 (more details coming to the UCISA NG mailing list soon).

Data Intensive Science

As an area where I have little experience, I was interested in listening to what William Johnston from ESnet had to say about large-scale data intensive science. He started by explained his view that high energy physics is seen as a prototype platform for distributed collaboration in other science fields.

He explained that as instruments get bigger, they get more expensive (in a not-quite-as-dramatic Moore’s Law relationship); therefore there are less of them which results in an increase in collaboration, globally. This shows the potential future growth of research networking bandwidth requirements.

One of the things I didn’t realise was that ESnet have extended their 100Gb full network backbone across the Atlantic into Europe, including connections in London. Their first circuit is being tested today. What does this mean for science and research in the UK?

Further details are available at:
http://es.net/news-and-publications/esnet-news/2014/esnet-extends-100g-connectivity-across-atlantic
http://www.geant.net/MediaCentreEvents/news/Pages/three-high-speed-links.aspx

William went on to talk about monitoring the network, explaining the criticality of this area. With many Infrastructure as a Service (IaaS) offerings, researchers are requesting Network as a Service; and with that the same levels of assurance and guarantees that have only previously been available with point to point links; is this going to change?

As one would expect, ESnet use perfSONAR for their assurance measurements. As I mentioned earlier, we will hopefully have representatives from ESnet and eduPERT at our Network Performance and QoS ‘community of practice workshop’ in January 2015.

Would something like perfSONAR deployed across Janet be of benefit for the community, perhaps let us know your thoughts in the blog feedback section below? I would assume it requires volunteer sites; however Janet are already looking at the possibility of network based probes for eduroam, so perhaps there is scope for a next generation of Netsight with added assurance?

ESnet also use the weathermap tool, which is also loved by colleagues at Loughborough University. It was one of the best take away messages from a Janet Networkshop Lightning talk several years ago.

The remainder of the talk was about data transfer speeds and integrity. I was surprised to hear the comment “SCP is your enemy”. Surely not? However I was approaching the problem from the wrong angle, thinking about security and not data transfer speeds and parallelisation. Look at some of the figures in the photograph below.

2014TechExDay3

 

William discussed a number of tools including GridFTP and a development from CALTECH, which stripes data across discs as part of the FTP process as well as providing up to three times CRC checking.

Interestingly the last point was about data integrity, which is critical for the field of data intensive science. William referenced the paper Stone and Partridge, 2000 “When The CRC and TCP Checksum Disagree”.

During the break, I had a bit of a Google to find any UK user or interest groups for Research Computing and HPC. I found the HPC SIG, if you know of any others, please pop them in the blog comments to share.

Connecting 40Gb Hosts

Whilst in the ‘big data’ mindset, there was an interesting session where colleagues from Fermi Labs, ESnet and CALTECH shared best practice infrastructure configuration to support high-speed data transfer.

There was some very interesting visual modelling, which demonstrated the affinity the network card has with a particular processor socket and core. The difference between optimising for data transfer is significant 37Gbps vs 26Gbps max on a 40Gbps link.

It was a packed session with many colleagues standing at the back; there is certainly an art to tweaking infrastructure to perform in the best possible manner. It was also interesting to hear there are three 100Gb network cards in development and testing.

Pushing the Boundaries of the Traditional Classroom

There was a bit of a clash in the programme, so I didn’t get to spend a lot of time in this session, but it was interesting to see what Indiana University had done with their ‘Collaboration Café’.

It led me to wonder what the key limitation of adopting more of these learner centric classroom designs is? Is it financial or is it resistance from academic colleagues in the same way as there was/is resistance to lecture capture and VLE environments?

UCISA are working along with SCHOMS and AUDE on an update to Learning Space design principals. This document should be really useful, especially as the final point from the presentation was all about the removal of wires.

At Loughborough we are trialling the Epson projector series that use the Epson EasyMP software and iProjection App. What wireless projectors and/or screens are you using? Let us know in the blog feedback section below?

Other Thoughts

The other talks I attended through the day continued on the research and big data theme. It included hearing about the PetaBytes (PB) of data required by some of the medical research being undertaken as part of the ICTBioMed platform. One of the speakers commented that biology is becoming more like computer science by the day; confirming again that multidisciplinary research is a firm requirement for a lot of modern applied research.

Some examples of digital biology given were: DNA Sequencing, Gene Expression Analysis, Protein Profiling and Protein to Protein interactions.

A number of the speakers came in via videoconference; it was interesting to see the mix of success and failure of this bold move. It seems strange that we still struggle to co-ordinate a remote video connection with the technology we have at our disposal in 2014.

Another speaker also made reference to the worldwide nature of large research groups and collaborations and said this collaboration technology was essential.

Video Collaboration

For the final session of the day, I was interested to see what the future held for video based collaboration in a session with speakers from: Internet2, Pexip, Evogh, Blue Jeans and Vidyo. I didn’t manage to ask Robb from Blue Jeans more about the removal of the Skype interface API that was so disappointing, however during the panel he mentioned that they had a Google Hangouts bridge to standards based systems available.

There were some interesting remarks from Hakon Dahle who is CTO at Pexip based in Oslo (but was previously CTO at Tandberg and Cisco).

Hakon described their distributed architecture, where it is possible to start small and grow appropriately with options to add capacity on demand in an agile manner.

Latency was still an issue with global video conferencing and there was a panel debate about the pros/cons of transcoding increasing latency vs accessibility and interoperability.

“Transcoding is a necessary evil”; especially with new protocols like WebRTC etc!

There were very positive comments about WebRTC and how it will make video more accessible and will make face to face communications easier; however there is already a divide with Google VP9 protocols being favoured by some players in the market especially when delivering very high resolution 4K streams.

Hakon explained that WebRTC seemed the most promising technology to allow direct person to person video calls and will bring about a lot of new use cases and that the new use case element is the most exciting in terms of innovation.

Learning Points

• How do we best position our infrastructure to take advantage of emerging Cloud competition?
• How do we collaborate more with colleagues from Internet2, ESnet and EDUCAUSE? Is this something UCISA and Janet/Jisc can facilitate?
• Future growth potential of research data transfer requirements
• Are we best serving our research communities, what more can we do?
• Global nature of research and therefore the communication requirements.

Matt Cook

Leave a Reply

Your email address will not be published. Required fields are marked *