Current accomplishments:
Staffing continued to be the focus of March activities, with interviews for Level 1 DM lead search at UW, and visits to DM group in Tucson. Another promising candidate was interviewed for Level 1 DM science lead position at University of Washington. The position has informally been offered, and we're waiting to hear back whether the candidate is interested. Another offer has also been made for the SQuaRE science lead position in Tucson; we're awaiting for the response from that candidate as well.

The DM Project Scientists has begun work on the update of the Galactic star counts model for the LSST sizing model, with the help of a graduate student.

The DM Project Scientist also had a productive working two-day meeting with Eric Tollerud (AstroPy coordinator) on understanding the architecture and opportunities for collaborations with the AstroPy project. The key results have been discussed with the DM System Architect. They resulted in an increase of probability of occurrence of DM-065 risk ("LSST DM software architecture incompatible with de-facto community standards") from 25-50% to 50-75% due to divergence between DM's current C++-centric model and the pure Python model of AstroPy.

Finally, the DM Project Scientist contributed patches to EUPS and the lsstsw build and continuous integration system, resulting in speed-ups of builds of approximately ~2.3x, and also contributed materials for the presentation on the state of DM for the April AMCL meeting.

Planned activities:

In April, the DM Project Scientist expects a continued focus on completion of hiring activities, as well as increasingly more activity in updating the key science/technical baseline documents.

DM System Engineering

02C.01.02
The W15 stack release (v10_1) process started. Two release candidates were prepared in March. The release was blocked on a number of bugs that spilled into April. A full release report will be provided next month.

A new system was released for beta-testing that allows for the automation of clean-slate builds on a variety of operating systems provided by cloud back ends. It is based on Vagrant and Puppet and eventually will support a number of APIs including OpenStack, AWS, and Digital Ocean.

The beta testing version supports Digital Ocean as the most available provider for now. This allows us with a few commands to have a brand new instance of, say, RedHat 7 to be provisioned with the stack pre-requisites, and verify the user-facing experience (via newinstall.sh from our distribution server). This will allow release testing to speed up considerably and for us to be able to verify the stack’s availability on a wider range of platforms.

Work started on a Jenkins-based CI system to replace buildbot. The new system will allow for an arbitrary number of build slaves to be be instantiated on demand from a cloud or virtualised back end. A proof of concept of running the stack build script in a Jenkins slave was completed.

Simon Krughoff (SK) attended the simulations team meeting as a DM representative to keep track of simulations progress as it pertains to DM concerns. SK hosted John Parejko, Ian Sullivan, and David Reiss on UW site visits as part of the hiring process for the DM team at UW. SK also hosted Erik Tollerud (Astropy) in order to talk about the differences and similarities between how LSST and Astropy have approached similar problems: table like data, coordinates, quantities (scalar values with units), world coordinate systems.

SK also attended the Open Science Grid (OSG) all hands meeting at Northwestern. The OSG is interested in LSST workflows and how users will use OSG in the era of LSST.

Because of the multiple meetings and visitors SK made very little progress on cleaning up the diffim task.

Russell Owen (RO) worked on making SWIGed interfaces work better when passed numpy types. It turns out that numpy scalars are harder to deal with. It was decided to leave scalar interfaces alone and only pass native python types. The array interfaces were converted to accept numpy arrays.

RO ran into some issues integrating the astrometric solver into the new framework. The original plan was to bring over the solver from HSC since it is battle tested. However, it did not pass the current unit tests without major revisions. It was decided to use the version of a similar solver already existing in the LSST stack since it is better documented and tested.

SK will attend the T/CAM training in Tucson. SK will host one more candidate for the DM team job.

Yusra AlSayyad (YA) will work on making ISR tasks runnable without an an opaque input object.

SK will continue work held over from March on bringing the difference imaging pipelines up to standard.

There is some holdover work due to the issues encountered by RO. These will need to be resolved early in the next month as a new production release is expected on short timescales.

This report covers work carried out in FM5 of FY15 in the Data Release Production group (staff at Princeton plus Price and Gee working remotely).

This month was the first of the Summer 2015 development cycle. The plans developed over the preceding two months are therefore locked in to the project management system and we have commenced executing them.

Some work planned for the final days of Winter 2015 was not complete on schedule as it turned out to be more complex than anticipated. This work will be continued in Summer 2015 under the DM-1769 (“Measurement Framework Improvements”) epic. This epic has therefore increased in scope somewhat over our original planning.

Together with Krughoff (UW), Bosch & Swinbank completed a reworking of the LDM-240 Software Development Roadmap for all years through project construction. There are a number of changes since the previous version, although these are mainly for clarity and to better reflect when we expect manpower to be available to address issues, rather than any fundamental change to the plan. We expect to iterate on this in the short term to incorporate changes in other WBS elements, ensuring dependencies are mapped correctly.

After the current round of LDM-240 changes have been agreed, we plan to embark on a larger scale and longer term effort to identify each entry on LDM-240 with the requirements for specific outputs specified by LSST’s Data Products Definition and use this information to produce another revision of the plan. We do not anticipate results from this before the end of the Summer 2015 cycle at the earliest.

A design for the new Footprint API was completed. This represents a substantial overhaul of the old codebase, for reasons of performance and maintainability and to enable the new functionality which will be required for upcoming work on the deblender. This design is now undergoing review; if and when it is accepted, we expected implementation to take place over the rest of this cycle. [DM-1126]

The afw_extensions_rgb package was merged into the afw package following an RFC. This reduces code duplication and means that the functionality can be entirely implemented in Python. [RFC-32, DM-2343, DM-2363]

A bug which caused a segmentation fault when attempting to write images with a certain header layout to disk. [DM-882]

A bug which caused the creation of extremely large images to result in a segmentation fault was resolved by porting and updating code from HyperSuprimeCam (HSC). [DM-527]

As described above, a number of minor tasks spilled over into this cycle from previous work and were addressed this month. In particular:

An update for the deblender which resolved an issue with incorrect substitution of noise in images was ported from the HSC stack. This issue resolved a long-standing but previously unexplained bug. [DM-1738, DM-1525]

Fixes for bugs in the RGB image generating code were also copied from HSC. [DM-2436]

After the new Footprint design has been reviewed, we expect implementation to start during April.

Work will start on measurements of shear precision in galaxy fitting. This work will build on the GREAT3 codebase to develop a library of simulated galaxy images and PSFs, then establish the optimal parameters for measuring them in the LSST stack. This provides insight into the ultimate computational burden which galaxy fitting will impose.

Work will continue on porting all the multi-band deblender work from HSC to LSST.

Measurement transformation and calibration functionality will be made available for the three major measurement types: fluxes, centroids and shapes.

Xiuqin and David worked on the first version of LDM-240 update for SUI. We feel there are more places need to be explained and defined.

Several meetings between the SUI and the Camera team on the revision of LSE-68, reaching a number of decisions on clarifications and changes to the interface for buffered data.

SUI team met at least twice weekly to discuss the SUI requirements and design issues. The minutes have been captured on confluence page.

We have continued our weekly meeting with SLAC on data access APIs. Other topics also came up, like VO TAP API, database schema browser (moved from SLAC to IPAC), data access through butler for SUI. Those topics will need further design discussion.

Firefly package has been migrated to GWT2.7, enabling us to use the new JS interop capabilities. Some simple input forms in the Firefly have been written using React framework.

Trey and Gregory went to SLAC to work with Camera team Tony Johnson and others for two days. Trey gave a demo of Firefly visualization components. He also helped the team to set up a version and they started development using it almost immediately.

SUI team hosted Robert Lupton to discuss the pipeline needs of visualization. Robert gave a demo of his current usage of ds9 and Trey gave a demo on ho wot use Firefly http APIs to show and control image display.

SUI team will take Camera team needs and pipeline needs as additional driver in designing the visualization tools.

Planned activities:
02C.05.00

Xiuqin will attend the T/CAM training.
LDM-240 road map revamp for Y16-Y20, revision.
Frossie will be visiting IPAC to discuss SDQA visualization needs and L3v3l 3 eta products.

02C.05.01

Continue the discussion of data access APIs with SLAC group.
K-T will visit to discuss butler and DM architecture with us.

02C.05.02

Continue the development on the JavaScript APIs and Python APIs for Firefly visualization components.
Update and add new the stretch algorithm in Firefly for image visualization.

02C.06.00 Science Data Archive and Application Services Management Engineering and Integration

· Added support for DDL in MetaServ (design and implementation). The Webserv now supports structures that define schema and plus additional custom fields, e.g. ucds, units. Some of that work is ahead of schedule.

02C.06.00 Science Data Archive and Application Services Management Engineering and Integration

02C.07.00 Processing Control and Site Infrastructure Management

NCSA’s management activities during the month of March included drafting requirements for three new positions in the areas of system architecture and workflows, signing the MOU with IN2P3 and appointing Don Petravick as head of the Joint Oversight Committee, and revising the milestones for FY15 in LDM-240. Additionally, the management team began considering options for better internal tracking and reporting of effort that will streamline with existing tools and processes at NCSA and the UIUC.

02C.07.01 Processing Control

In March most effort was placed on finishing activities that were not completed at the end of the last cycle (rollover from W15). More work was done on building state diagrams to show job progress throughout a nightly computing run. The goal was to improve the diagrams so that they demonstrate that the Data Management Control System (DMCS) will always be in a specific known state. Among those drafted or revised were an Archive DMCS diagram, a Base DMCS diagram, a Distributor Node diagram, a Replicator Job diagram, and a Worker Job diagram. There has been a detailed discussion and review of the components and potential states of the DMCS.

Simplifying and refactoring the Event Services API, another activity worked on extensively in the W15 cycle, was an additional focus of effort in March. After integrating and testing changes from the API code review (which were mostly minor and aesthetic), this activity was completed. The completion of this refactoring allowed for work on the Apache logging framework to progress. Unit tests were created in preparation for writing the DM message appender class.

NCSA team members participated in the LSST Technical Operations Working Group (TOWG). They developed use cases to help with specifications for the buildings in La Serena, justifying the required office space based on the necessary level of staff.

02C.07.02 Infrastructure Services

In March, the new Information Security Officer (ISO) began researching the project’s existing requirements and security policies. He reviewed SCADA documents, the data management design, the operations plan, and the observatory site requirements to gain a high-level understanding of the project’s systems and architecture. He investigated identity management systems that could be implemented for users to access scientific data. One such system in Shibboleth.

Work began on infrastructure configuration management. This will simplify managing the various servers and VMs at NCSA. Initial work was done prototyping with Puppet, configuration management tool.

After an internal discussion with the NCSA team about information classification, a document was developed describing storage structure and policies for the LSST DM developers environment. It identifies storage categories based on the data classification category (LPM-122), responsible manager of the data in that category, disk quota, purge policy, and backup and archive policies.

A major area of focus for this cycle is file systems and management. Since the LSST team at NCSA still lacks the ability to procure goods, we have been working closely with the Innovative Systems Lab (ISL) at NCSA, an organizational vehicle used to investigate pre-production technologies at NCSA. In March we evaluated and commented on an ISL proposal to investigate the CEPH file system for its properties as an alternative to the LSST baseline file system GPFS.

Research began on file management technologies, specifically iRODs functionality. This work was accomplished using resources on the NCSA ISL OpenStack. After setting up and configuring servers and host clients, testing was done on data replication. The configurations were also installed in Docker.

One activity related to understanding commercial file systems between data centers was not planned to begin until May, but was worked on in March due to the timely visit of vendors from DataDirect Network (DDN), a provider of big data storage. The NCSA team discussed our system needs and DDN presented on a product called WOS. This is a commercial object store which has interfaces for NFS, GPFS and Swift. Notes from this visit are captured in Jira ticket DM-2230.

02C.07.03 Environment and Tools

No activities are planned in this area for the S15 cycle.

02C.07.04 Site Infrastructure

Discussions began between NCSA and the qserv and SUI teams about the requirements for setting up a qserv prototype. We established “short-term” and “long-term” system needs (VMs, memory, disk space, ports, etc.). Implementation of this prototype is not scheduled to begin until July, though it will require NCSA to purchase new equipment and will thus be highly dependent on finalizing the pending procurement contract.

The base data center requirements were revised with updates to square footage, average rack weights, and power requirements. These revisions were circulated around with the Chilean site team.

No activities dependent on purchasing new hardware have been planned until mid-summer due to pending procurement contract.

Several LOE tasks emerged during March and were addressed. Two user accounts were created and one was disabled since the user has left the project. Some issues with the existing systems at NCSA were fixed, including a Crashplan failure, a GNU library issue, and a drive failure on one of the LSST machines at NCSA (lsst-dbdev2). The NCSA infrastructure team deployed NAGIOS Puppet agent checks and active bandwidth network tests that would maximize bandwidth from a user perspective. Some packet loss was detected, indicating we are not getting the bandwidth we would like.

At the end of the month several users reported slowness with the SSD on one of the development machines; this is being investigated.

·	Synchronized metaserv schema structure with the front-end of the form for collecting user input data. The work on the form is done by Tony, the person we are adding to the group. He is now ready to do the “final” implementation.
02C.06.02.01 Data Access Client Framework
·	No updates here, nothing was scheduled for March here
02C.06.02.02 Web Services
·	Fine-tuned Data Access Interfaces (DM-1916) – the latest version reflects the latest thinking and decisions made at Data Access meetings
·	Ported dbserv and metaserv to the db module – the web service now uses the db module (better code reuse)
·	Implemented basic db pool to avoid problems with mysql timeout
·	As a result, the WebServ is now ready for testing by the IPAC team, however the format of results returned is still not “final”, so after basic testing, they stopped and are waiting for the “proper” formatting of the results.
02C.06.02.03 Query Services
·	Built Qserv Release 2015_03. Release notes are pasted below. Few highlights:
·	have new worker management system – we can now manage worker nodes through ssh
·	have much better unit test for one of the core complex components: query
02C.06.02.04 Image Services
·	Implemented image stitching (DM-2013)
·	Discussions with Robert Lupton about advanced features of image cutout service