1. PanDA & Networking Kaushik De Univ. of Texas at Arlington ANSE Workshop, CalTech May 6, 2013

2. Introduction  Background  PanDA is a distributed computing workload management system  PanDA relies heavily on networking for data transfers  ANSE provides opportunity for PanDA integration with networking  Note: data transfers are done asynchronously within PanDA – by DQ2 in ATLAS, PhEDEx in CMS, pandamover  PhEDEx scope is part of ANSE  DQ2 is not part of ANSE  Pandamover may be within ANSE scope? TBD  Ambitious goals for PanDA in ANSE  Direct integration of networking with workflow – never attempted before for large scale automated systems like PanDA  Would prefer to make it agnostic to the choice of data transfer system – thus providing higher level of workflow optimizations May 6, 2013Kaushik De

3. Concept: Network as a Resource  PanDA as workload manager  PanDA automatically chooses job execution site  Multi-level decision tree – task brokerage, job brokerage, dispatcher  Also manages predictive workflows – at task definition, PD2P  Current scale – one million jobs completed daily at ~hundred sites  Site selection is based on processing and storage requirements  Can we use network information in this decision?  Can we go even further – network provisioning?  Network as resource  Optimal site selection should take network capability into account  We do this already – but rather crudely using job completion metrics  Network as a resource should be managed (i.e. provisioning)  We also do this crudely – mostly through timeouts, self throttling May 6, 2013Kaushik De

4. Current Status  Three parallel efforts to integrate networking in PanDA  US ATLAS funded – primarily to improve integration with FAX  ASCR funded – bigPanDA project, taking PanDA beyond LHC  ANSE funded – this project  We are coordinating the three efforts to maximize results  US ATLAS FAX status  Preliminary work done by UC team (HC testing), UTA team (pilots)  Need help from ANSE/ASCR to move forward – US ATLAS team is busy with many other higher priority items to get task finished  ASCR status  Next Generation Workload Management and Analysis System for Big Data, PANDA integration with networking, DOE funded (BNL, U Texas Arlington)  Work has started with BNL team (Dantong Yu) May 6, 2013Kaushik De

5. ANSE PanDA Status  Personnel  Artem Petrosyan and Danila Oleynik hired at UTA  Arriving at UTA tomorrow May 7  Funding - 0.5 FTE ANSE, 1.5 FTE ASCR  Both will be at UTA for ~2 years  There is commitment from Dubna for 1-2 more years  Nominally, Artem will be paid by ANSE – but they work as team  Plan of work  For discussion in this meeting May 6, 2013Kaushik De

6. Proposed ANSE PanDA Use Cases 1)Use network information for FAX brokerage 2)Use network information for job assignment  Improve flow of ‘activated’ jobs  Better accounting of ‘transferring’ jobs 3)Use network information for PD2P 4)Use network information for site selection 5)Use network information for cloud selection 6)Provision circuits for PD2P transfers 7)Provision circuits for input transfers 8)Provision circuits for output transfers May 6, 2013Kaushik De

7. Proposed ANSE PanDA Use Cases 1)Use network information for FAX brokerage 2)Use network information for job assignment  Improve flow of ‘activated’ jobs  Better accounting of ‘transferring’ jobs 3)Use network information for PD2P 4)Use network information for site selection 5)Use network information for cloud selection 6)Provision circuits for PD2P transfers 7)Provision circuits for input transfers 8)Provision circuits for output transfers May 6, 2013Kaushik De

8. FAX Integration with PanDA  We have developed detailed plans for integrating FAX with PanDA over the past year  Networking plays an important role in Federated storage  This time we are paying attention to networking up front  The most interesting use case – network information used for brokering of distributed analysis jobs to FAX enabled sites  This is first real use case for using external network information in PanDA May 6, 2013Kaushik De

9. FAX for Distributed Analysis  Plan of action  Currently, DA jobs are brokered to sites which have input datasets  This can limit and slow the execution of DA jobs  Use FAX to relax constraint on locality of data  Use cost metric generated with Hammercloud tests initially – treat as ‘typical cost’ of data transfer between two sites  Brokerage will use concept of ‘nearby’ sites  Calculate weight based on usual brokerage criteria (availability of CPU…) plus network transfer cost  Jobs will be sent to site with best weight – not necessarily the site with local data or available CPU’s  Cost metric already available – soon to be tested in brokerage Kaushik DeMay 6, 2013

10. HC BASED WAN TESTS  HC submits 1 job/day to all of the “client” nodes. Client node is the one using the data. It is an ANALY queue.  All the “server” sites have one and the same dataset. Server sites are the ones delivering data.  Each job, each half an hour, in parallel:  Pings of all of the “server” sites.  Copies a file from a site (xrdcp/dccp)  Reads the file from a root script  Uploads all the results to Oracle DB at CERN  Result are shown at: http://ivukotic.web.cern.ch/ivukotic/WAN/index.asphttp://ivukotic.web.cern.ch/ivukotic/WAN/index.asp  Results are also given in JSON format to SSB: http://dashb-atlas- ssb.cern.ch/dashboard/request.py/siteview#currentView=Network+Measur ements&highlight=falsehttp://dashb-atlas- ssb.cern.ch/dashboard/request.py/siteview#currentView=Network+Measur ements&highlight=false 20/03/2016 ILIJA VUKOTIC IVUKOTIC@CERN.CH 10

11. HC BASED WAN TESTS 20/03/2016 ILIJA VUKOTIC IVUKOTIC@CERN.CH 11 Color coded conn. quality Weight is the time to transfer file

12. Read Testing (Hammercloud) May 6, 2013Kaushik De Source: Rob Gardner

13. Details of FAX DA Plan  Performance data kept in AGIS (A Grid Information System)  Schema has been defined for Hammecloud (HC)  Auto-update, validate and monitor this information  Alternative sources could be perfSONAR etc – schema is flexible  Information in schedConfigDB  Data is massaged and stored in special table in pandaDB (Oracle)  Schema is fixed - irrespective of source(s)  Update must be reliable – information used for real-time decisions  PanDA brokerage uses the information to calculate weight – single weight incorporating all knowledge of CPU, storage, and networking  First use case of network as resource May 6, 2013Kaushik De

14. Proposed ANSE PanDA Use Cases 1)Use network information for FAX brokerage 2)Use network information for job assignment  Improve flow of ‘activated’ jobs  Better accounting of ‘transferring’ jobs 3)Use network information for PD2P 4)Use network information for site selection 5)Use network information for cloud selection 6)Provision circuits for PD2P transfers 7)Provision circuits for input transfers 8)Provision circuits for output transfers May 6, 2013Kaushik De

15. Job States  Panda jobs go through a succession of steps tracked in DB  Defined  Assigned  Activated  Running  Holding  Transferring  Finished/failed May 6, 2013Kaushik De

16. Assigned Jobs  Assigned -> Activated workflow  Group of jobs are assigned to a site by PanDA brokerage  For missing input files, data transfer is requested asynchronously  PanDA waits for “transfer completed” callback from DDM system to activate jobs for execution  Network data transfer plays crucial role in this workflow  Can network technology help assigned->activated transition?  Number of assigned jobs depend on number of running jobs – can we use network status to adjust rate up/down?  Jobs are reassigned if transfer times out (fixed duration) – can knowledge of network status be used to set variable timeout?  Can we use network provisioning in this step? Per block? May 6, 2013Kaushik De

17. Transferring Jobs  Transferring state  After job execution is completed, asynchronous data transfer is requested from DDM  Callback is required for successful job completion  How can network technology help?  New jobs are not sent if too many transferring jobs  Can we make this limit variable using network information?  Fixed timeout is set for transferring - delays completion of tasks  Can network status information help – variable timeout?  Can we use provisioning? Per job, per block, or tune based on average rate? May 6, 2013Kaushik De

18. Proposed ANSE PanDA Use Cases 1)Use network information for FAX brokerage 2)Use network information for job assignment  Improve flow of ‘activated’ jobs  Better accounting of ‘transferring’ jobs 3)Use network information for PD2P 4)Use network information for site selection 5)Use network information for cloud selection 6)Provision circuits for PD2P transfers 7)Provision circuits for input transfers 8)Provision circuits for output transfers May 6, 2013Kaushik De

19. PD2P – How LHC Model Changed  PD2P = PanDA Dynamic Data Placement  PD2P is used to distribute data for user analysis  For production PanDA schedules all data flows  Initial ATLAS computing model assumed pre-placed data distribution for user analysis – PanDA sent jobs to data  Soon after LHC data started, we implemented PD2P  Asynchronous usage based data placement  Repeated use of data → make additional copies  Backlog in processing → make additional copies  Rebrokerage of queued jobs → use new data location  Deletion service removes less used data  Basically, T1/T2 storage used as cache for user analysis  This is perfect for network integration  Use network status information for site selection  Provisioning - usually large datasets are transferred, known volume May 6, 2013Kaushik De

20. Proposed ANSE PanDA Use Cases 1)Use network information for FAX brokerage 2)Use network information for job assignment  Improve flow of ‘activated’ jobs  Better accounting of ‘transferring’ jobs 3)Use network information for PD2P 4)Use network information for site selection 5)Use network information for cloud selection 6)Provision circuits for PD2P transfers 7)Provision circuits for input transfers 8)Provision circuits for output transfers May 6, 2013Kaushik De

21. ATLAS Computing Model May 6, 2013Kaushik De Tier1 site Tier2 site Cloud Tier2 site Tier2D site Task job  11 Clouds 10 T1s + 1 T0 (CERN) Cloud = T1 + T2s + T2Ds (except CERN) T2D = multi-cloud T2 sites  2-16 T2s in each Cloud job Task  Cloud Task brokerage Jobs  Sites Job brokerage job

22. Task Brokerage  Matchmaking per cloud is based on:  Free disk space in T1 SE, MoU share of T1  Availability of input dataset (a set of files)  The amount of CPU resources = the number of running jobs in the cloud (static information system is not used)  Downtime at T1  Already queued tasks with equal or higher priorities  High priority task can jump over low priority tasks  Can knowledge of network help  Can we consider availability of network as a resource, like we consider storage and CPU resources?  What kind of information is useful? May 6, 2013Kaushik De

23. Job Brokerage  Brokerage policies define job assignment to sites  IO intensive or TAPE read -> prefer T1  CPU intensive -> T1+T2s  Flexible: clouds may allow IO heavy jobs at T2s with low weight  Matchmaking per site in a cloud  Software availability  Free disk space in SE, Scratch disk size on Worker Node (WN), Memory size on WN  Occupancy = the number of running jobs / the number of queued jobs, and downtime  Locality (cache hits) of input files  Can we add network information to the matchmaking? May 6, 2013Kaushik De

24. Job Dispatcher  High performance/high throughput module  Send matching job to CE upon pilot request  REST non-blocking communication  Different from brokerage, which is asynchronous  Matching of jobs based on  Data locality  Memory and disk space  Highest priority job is dispatched  At this point networking is not as important  Is this true – we still have to transfer output  Can we initiate provisioning? May 6, 2013Kaushik De

25. Summary  Many different parts of PanDA can benefit from better integration with networking  Need to work out details of information to be collected  Need to work out details of circuit provisioning  Developers starting soon (actually tomorrow)!  It would be useful if we have well defined work plan for them, based on discussions here May 6, 2013Kaushik De