Number of luminosity blocks from the start of of a data-taking run until a valid beam spot has been determined and made available to the High Level Trigger processors during the 2011/12 proton-proton run. One luminosity block corresponds to about one minute of data-taking. A beam spot update is performed on the HLT farm if the position changes by more than 10% of the estimated luminous width; or the estimated luminous width changes by more than 10%; or the errors on these quantities decrease by more than 50%.

Time spent by the ca. 7600 Level-2 High Level Trigger applications loading the beam spot information from the conditions database. The time is measured between the reception of the update notification and the successful loading of the beam spot parameters from the conditions database. This extremely short update time is achieved by means of a database proxy hierarchy deployed on the High Level Trigger farm. The update time is significantly less than the average event processing time of 60ms and thus can be performed while data-taking is ongoing.

Number of beam spot updates to the High Level Trigger farm vs. luminosity block number for the 2011/12 proton-proton running period. The inset shows that the majority of beam spot updates are performed during the beginning of the fill. One luminosity block corresponds to about one minute of data-taking. A beam spot update is performed on the HLT farm if the position changes by more than 10% of the estimated luminous width; or the estimated luminous width changes by more than 10%; or the errors on these quantities decrease by more than 50%.

Number of beam spot updates to the High Level Trigger farm for the 2011/12 proton-proton running period. Only runs with a duration of at least one hour are considered. Entries with very low number of beam spot updates per run are due to run re-starts during an ongoing LHC fill. A beam spot update is performed on the HLT farm if the position changes by more than 10% of the estimated luminous width; or the estimated luminous width changes by more than 10%; or the errors on these quantities decrease by more than 50%.

Time-evolution of the luminous centroid position in x measured in the High Level Trigger (HLT) during one LHC fill. The position is shown relative to the first valid beam spot and one luminosity block corresponds to about one minute of data-taking. The black dots are the current beam spot position measured in five minute intervals. If the position changes by more than 10% of the estimated luminous width; or the estimated luminous width changes by more than 10%; or the errors on these quantities decrease by more than 50% a beam spot update on the HLT farm is performed (red squares). The update is performed with a delay of about two luminosity blocks with respect to the current measurement. The green band indicates one of the update criteria, i.e. if the position changes by more than 10% of the estimated luminous width.

Time-evolution of the luminous centroid position in y measured in the High Level Trigger (HLT) during one LHC fill. The position is shown relative to the first valid beam spot and one luminosity block corresponds to about one minute of data-taking. The black dots are the current beam spot position measured in five minute intervals. If the position changes by more than 10% of the estimated luminous width; or the estimated luminous width changes by more than 10%; or the errors on these quantities decrease by more than 50% a beam spot update on the HLT farm is performed (red squares). The update is performed with a delay of about two luminosity blocks with respect to the current measurement. The green band indicates one of the update criteria, i.e. if the position changes by more than 10% of the estimated luminous width. The dip in the y-position at the beginning of the fill is currently under investigation and is possibly related to thermal effects.

Time-evolution of the luminous centroid position in z measured in the High Level Trigger (HLT) during one LHC fill. The position is shown relative to the first valid beam spot and one luminosity block corresponds to about one minute of data-taking. The black dots are the current beam spot position measured in five minute intervals. If the position changes by more than 10% of the estimated luminous width; or the estimated luminous width changes by more than 10%; or the errors on these quantities decrease by more than 50% a beam spot update on the HLT farm is performed (red squares). The update is performed with a delay of about two luminosity blocks with respect to the current measurement. The green band indicates one of the update criteria, i.e. if the position changes by more than 10% of the estimated luminous width.

Time-evolution of the estimated luminous width in x by measuring the width of the vertex distribution in the High Level Trigger (HLT) and applying a first-order resolution correction. One luminosity block corresponds to about one minute of data-taking. The black dots are the current estimated luminous width measured in five minute intervals. If the position changes by more than 10% of the estimated luminous width; or the estimated luminous width changes by more than 10%; or the errors on these quantities decrease by more than 50% a beam spot update on the HLT farm is performed (red squares). The update is performed with a delay of about two luminosity blocks with respect to the current measurement. The green band indicates one of the update criteria, i.e. if the estimated luminous width changes by more than 10% of the nominal estimated luminous width. The initial decrease of the luminous width is due to the bootstrapping procedure of the resolution correction. The sub-sequent increase of the width is caused by the increase in beam emittance during the fill.

Time-evolution of the estimated luminous width in y by measuring the width of the vertex distribution in the High Level Trigger (HLT) and applying a first-order resolution correction. One luminosity block corresponds to about one minute of data-taking. The black dots are the current estimated luminous width measured in five minute intervals. If the position changes by more than 10% of the estimated luminous width; or the estimated luminous width changes by more than 10%; or the errors on these quantities decrease by more than 50% a beam spot update on the HLT farm is performed (red squares). The update is performed with a delay of about two luminosity blocks with respect to the current measurement. The green band indicates one of the update criteria, i.e. if the estimated luminous width changes by more than 10% of the nominal estimated luminous width. The initial decrease of the luminous width is due to the bootstrapping procedure of the resolution correction. The sub-sequent increase of the width is caused by the increase in beam emittance during the fill.

Time-evolution of the estimated luminous width in z by measuring the width of the vertex distribution in the High Level Trigger (HLT) and applying a first-order resolution correction. One luminosity block corresponds to about one minute of data-taking. The black dots are the current estimated luminous width measured in five minute intervals. If the position changes by more than 10% of the estimated luminous width; or the estimated luminous width changes by more than 10%; or the errors on these quantities decrease by more than 50% a beam spot update on the HLT farm is performed (red squares). The update is performed with a delay of about two luminosity blocks with respect to the current measurement. The green band indicates one of the update criteria, i.e. if the estimated luminous width changes by more than 10% of the nominal estimated luminous width.

Transverse horizontal distribution of primary vertices reconstructed online in the High Level Trigger. The distribution corresponds to 1 minute of data taking. The mean of the distribution reflects the luminous centroid position, while its width shows the uncorrected luminous size, which at this stage is still dominated by the intrinsic resolution.

Transverse vertical distribution of primary vertices reconstructed online in the High Level Trigger. The distribution corresponds to 1 minute of data taking. The mean of the distribution reflects the luminous centroid position, while its width shows the uncorrected luminous size, which at this stage is still dominated by the intrinsic resolution.

Longitudinal distribution of primary vertices reconstructed online in the High Level Trigger. The distribution corresponds to 1 minute of data taking. The mean of the distribution reflects the luminous centroid position, while its width shows the luminous length, which is much larger than the intrinsic resolution.

The transverse distribution of primary vertices reconstructed online in the High Level Trigger. The distribution corresponds to 1 minute of data taking and contains well over 100\,000 vertices.

The transverse distribution of primary vertices reconstructed online in the High Level Trigger. The distribution contains over 70 million vertices collected during one LHC fill.

The horizontal and vertical luminous tilt angles with respect to the ATLAS coordinate system measured online by the High Level Trigger. Transverse vertex distributions are divided into bins along the z-axis. The error bars indicate the error on the mean and are mostly too small to be visible. The slope is extracted with a linear fit versus z to measure the horizontal and vertical tilt angles.

Transverse luminous centroid position measured online by the High Level Trigger using only track transverse impact parameter and azimuth angle. The transverse impact parameter d0 versus azimuth angle phi distribution measures the transverse luminous centroid position through a sinusoidal fit. The distribution corresponds to 1 minute of data taking. Point-by-point error bars are too small to be visible. This method is used as a cross-check of the vertex method and is in excellent agreement with it.

The displacements in x and y between each pair of separately reconstructed half vertices of a split primary vertex as a function of the number of tracks, measured online in the High Level Trigger. To measure the instrinsic vertex resolution, primary vertices are split, and the two resulting sets of tracks are fitted separately. The displacement measures the resolution of the vertex reconstruction as a function of the number of tracks per vertex, and is used to compute a correction that is applied in real time. The distributions illustrate the two competing features of this measurement: the resolution improves substantially the higher the number of tracks in the vertex, while the statistical precision is much higher for the lower track multiplicities.

The displacements in z between each pair of separately reconstructed half vertices of a split primary vertex as a function of the number of tracks, measured online in the High Level Trigger. To measure the instrinsic vertex resolution, primary vertices are split, and the two resulting sets of tracks are fitted separately. The displacement measures the resolution of the vertex reconstruction as a function of the number of tracks per vertex, and is used to compute a correction that is applied in real time. The distributions illustrate the two competing features of this measurement: the resolution improves substantially the higher the number of tracks in the vertex, while the statistical precision is much higher for the lower track multiplicities.

The observed width of the primary vertex distribution, the measured resolution, and the resolution-corrected width, all as a function of the number of tracks per vertex. The observed width decreases with the track multiplicity as the resolution improves, while the corrected width stays essentially flat. The split vertex method is limited by the need for a corresponding sample of events with twice as many tracks per vertex. The method works well down to a certain multiplicity but then breaks down when the resolution becomes approximately twice the true width. The same behavior was observed for fills with larger beam sizes.

The width of the longitudinal primary vertex distribution along with the measured resolution as a function of the number of tracks per vertex. The z resolution has a similar dependence on the number of tracks as the resolution in the transverse plane. However, in contrast to the transverse plane, in the longitudinal direction the resolution-correction is entirely negligible.

Time-variation of the luminous centroid position in x, y, and z measured in the High Level Trigger every five minutes, for six separate LHC fills recorded over the span of four days.

Time-evolution of the luminous size in x, y, and z measured in the High Level Trigger every five minutes, for six separate LHC fills recorded over the span of four days. The effect of the transverse emittance blow-up is clearly visible during each fill and is on the order of 15\% on the horizontal width and 10\% on the vertical width for a 10 hour fill. Similarly, the luminous length increase from longitudinal emittance blow-up is on the order of 10\% for a 10 hour long fill.

The luminous centroid position in x, y, and z measured in the High Level Trigger for each of the 1024 colliding bunch pairs separately. Distinct structures are visible, in particular in the vertical position of the interaction-point, with variations of up to 5\,$\mu$m and repeating patterns across the injected bunch trains.