MPIfR Correlator Report (November 2007) _______________________________________ MPIfR/BKG MKIV CORRELATOR =========================== GENERAL Current correlator capabilities are listed at: http://www.mpifr-bonn.mpg.de/EVN/MK4CORstatus CORRELATOR HARDWARE > The correlator is being rebuilt to make space for a cluster for running the DiFX software correlator. All tape drives except one will be decommissioned before the end of the year. > In a second stage in 2008 the box which housed the Honeywell tape drives will be enlarged to make space for two to three more racks for the compute cluster. > Present playback status: 8 Mark 5As; (6 of these can be upgraded to 5B, mainboard upgrades underway ) 4 Mark 5Bs (3 operational on the correlator, 1 for eVLBI) 1 tape drive for thick and thin tape > Medicina repaired 2 SUIMs (Station Unit Interface Module). Haystack sent 4 spare SUIMs, but more SUIMs have failed in the meantime. SUs are becoming more and more fragile. > As a replacement for the old correlator control computers (HP-UX machines) Linux servers and a 20TB Raid have been installed. > For eVLBI a P2P (point to point) connection at 10 Gb from the institute to Dwingeloo to be shared with Lofar will probably be realised in December 2007. A 1 Gb line to Geant is available. > 1 server with 8 TB is used for Internet transfers of VLBI data. A second server with potentially more disk space will be purchased before the end of 2007. CORRELATION SOFTWARE > In a joint project Haystack and Bonn have been porting correlator software to Linux. A first release from Haystack was installed in April. The software includes Mark 5B support. Some of the correlator programs are still executed on an HP-UX system. The biggest complication in running such a mixed system is the different byte orders of the different processors. The stability of the software is satisfactory. SOFTWARE CORRELATOR > At a DiFX meeting in Bonn a "DiFX users group" was formed. Members are Swinburne, Curtin, NRAO, MPIfR, Metsähovi. A development plan was agreed and tasks were distributed. > MPIfR has hired an astronomer who will devote 50% of his time to the development of DiFX. > MPIfR has started using the DiFX SVN repository for installing the latest versions of the software correlator. > A first 10 station VLBA observation is being correlated. Its output parameters (short integration time, number of lags) are such that it cannot be correlated on any hardware correlator. > A proposal for a compute cluster for the software correlator was successful. A medium sized cluster will be installed in 2 steps. Towards the end of 2008 it is expected that the "soft correlation power" will surpass that of the hardware correlator. OPERATIONS > Bonn has declared to the IVS that no more tape observations will be correlated after 2007. A major problem at the correlator is that the tape drives will not work properly after they have been idle for weeks. > A first observation with 10 stations total in one pass was correlated with the new software and hardware with the following combination of playbacks: 8 Mark 5As, 1 thick tape (MK III format), 1 Mark 5B > The transfer of K5-recorded data from Tsukuba including conversion to Mark 5A is done on a regular basis for geodesy. > Production correlation with Mark 5B playback is working well. So far data from Westford and Badary are delivered in 5B-format regularly. First recordings from Parkes, Mopra, ATCA and Ceduna were received in Mark 5B format. E-VLBI > Regular near-realtime correlation of UT1 measurements are done on each Monday morning. Data from Tsukuba, Ny Alesund and Wettzell are transferred to the correlator via Internet. UT1 is determined in much less than 24 hours. > The following data transfers are performed regularly now: R1 sessions: Onsala, Tsukuba (incl. conversion), Ny Alesund EUROP sessions: Onsala, Metsähovi T sessions: Onsala, Metsähovi Tests will be done soon with Wettzell, Tigo, and Medicina. OTHER > Effelsberg is equipped with one Mark 5A and one Mark 5B unit, both will have gigabit Ethernet. The Mark 5B has been connected to the MK IV terminal with available VLBA samplers and a VSIC board. This allows to record Mark 5B format for tests at the JIVE correlator. > MPIfR took part in joint exploratory VLBI observations with Haystack at 4Gbit/s in spring of 2007. Due to limitations in the compatibility of the DBBC and the DBE, only 1 Mark 5B+ unit could be used at Pico Veleta which limited the maximum data rate on the European sided to 2 Gbit/s. Due to a cable problem only 1 Gbit/s could finally be recorded. > The fibre connection between Effelsberg and the institute is ready and presently under test. > A 2nd generation A/D converter has been developed for the DBBC2. The board can be used for different hardware applications by adding a small daughter board. Production is starting now. > A VSI to/from 10 GBE converter board for the DBBC is being developed. It can be connected to both the A/D converter and the FPGA boards. > A project for implementing RFI suppression on the DBBC is being investigated. > MPIfR will cooperate with Italian groups for programming the DBBC hardware for different applications. > We plan to equip Effelsberg with a Mark 5C unit in 2008. W.Alef