source: XIOS/dev/branch_openmp/Note/rapport ESIWACE.tex.backup @ 1620

Last change on this file since 1620 was 1560, checked in by yushan, 6 years ago

report update

File size: 17.8 KB
Line 
1\documentclass[a4paper,10pt]{article}
2\usepackage[utf8]{inputenc}
3\usepackage{graphicx}
4\usepackage{listings}
5\usepackage[usenames,dvipsnames,svgnames,table]{xcolor}
6\usepackage{amsmath}
7\usepackage{url}
8\usepackage{verbatim}
9
10% Title Page
11
12\title{Developing XIOS with multi-thread : to accelerate the I/O of climate models}
13
14\author{}
15
16
17\begin{document}
18\maketitle
19
20\section{Context}
21
22The simulation models of climate systems, running on a large number of computing resources can produce an important volume of data. At this
23scale, the I/O and the post-treatment of data becomes a bottle-neck for the performance. In order to manage efficiently the data flux
24generated by the simulations, we use XIOS developed by the Institut Pierre Simon Laplace and Maison de la simulation.
25
26XIOS, a library dedicated to intense calculates, allows us to easily and efficiently manage the parallel I/O on the storage systems. XIOS
27uses the client/server scheme in which computing resources (server) are reserved exclusively for IO in order to minimize their impact on
28the performance of the climate models (client). The clients and servers are executed in parallel and communicate asynchronously. In this
29way, the I/O peaks can be smoothed out as data fluxes are send to server constantly throughout the simulation and the time spent on data
30writing on the server side can be overlapped completely by calculates on the client side.
31
32\begin{figure}[ht]
33\includegraphics[scale=0.4]{Charge1.png}
34\includegraphics[scale=0.4]{Charge2.png}
35\caption{On the left, each peak of computing power corresponds to the valley of memory bandwidth which shows that the computing resources
36are alternating between calculates and I/O. ON the right, both curves are smooth which means that the computing resources have a stable
37charge of work, either calculates or I/O.}
38\end{figure}
39
40
41XIOS works well with many climate simulation codes. For example, LMDZ\footnote{LMDZ is a general circulation model (or global climate model)
42developed since the 70s at the "Laboratoire de Météorologie Dynamique", which includes various variants for the Earth and other planets
43(Mars, Titan, Venus, Exoplanets). The 'Z' in LMDZ stands for "zoom" (and the 'LMD' is for  'Laboratoire de Météorologie Dynamique").
44\url{http://lmdz.lmd.jussieu.fr}}, NENO\footnote{Nucleus for European Modeling of the Ocean alias NEMO is a
45state-of-the-art modelling framework of ocean related engines. \url{https://www.nemo-ocean.eu}}, ORCHIDEE\footnote{the land surface
46model of the IPSL (Institut Pierre Simon Laplace) Earth System Model. \url{https://orchidee.ipsl.fr}}, and DYNAMICO\footnote{The DYNAMICO
47project develops a new dynamical core for LMD-Z, the atmospheric general circulation model (GCM) part of IPSL-CM Earth System Model.
48\url{http://www.lmd.polytechnique.fr/~dubos/DYNAMICO/}} all use XIOS as the output back end. M\'et\'eoFrance and MetOffice also choose XIOS
49to manage the I/O for their models.
50
51
52\section{Development of thread-friendly XIOS}
53
54Although XIOS copes well with many models, there is one potential optimization in XIOS which needs to be investigated: making XIOS thread-friendly.
55
56This topic comes along with the configuration of the climate models. Take LMDZ as example, it is designed with the 2-level parallelization scheme. To be more specific, LMDZ uses the domain decomposition method in which each sub-domain is associated with one MPI process. Inside of the sub-domain, the model also uses OpenMP derivatives to accelerate the computation. We can imagine that the sub-domain be divided into sub-sub-domain and is managed by threads.
57
58\begin{figure}[ht]
59\centering
60\includegraphics[scale=0.5]{domain.pdf}
61\caption{Illustration of the domain decomposition used in LMDZ.}
62\end{figure}
63
64As we know, each sub-domain, or in another word, each MPI process is a XIOS client. The data exchange between client and XIOS servers is handled by MPI communications. In order to write an output field, all threads must gather the data to the master thread who acts as MPI process in order to call MPI routines. There are two disadvantages about this method : first, we have to spend time on gathering information to the master thread which not only increases the memory use, but also implies an OpenMP barrier; second, while the master thread calls MPI routine, other threads are in the idle state thus a waster of computing resources. What we want obtain with the thread-friendly XIOS is that all threads can act like MPI processes. They can call directly the MPI routine thus no waste in memory nor in computing resources as shown in Figure \ref{fig:omp}.
65
66\begin{figure}[ht]
67\centering
68\includegraphics[scale=0.6]{omp.pdf}
69\caption{}
70\label{fig:omp}
71\end{figure}
72
73There are two ways to make XIOS thread-friendly. First of all, change the structure of XIOS which demands a lot of modification is the XIOS library. Knowing that XIOS is about 100 000 lines of code, this method will be very time consuming. What's more, the modification will be local to XIOS. If we want to optimize an other code to be thread-friendly, we have to redo the modifications. The second choice is to add an extra interface to MPI in order to manage the threads. When a thread want to call an MPI routine inside XIOS, it will first pass the interface, in which the communication information will be analyzed before the MPI routine is invoked. With this method, we only need to modify a very small part of XIOS in order to make it work. What is more interesting is that the interface we created can be adjusted to suit other MPI based libraries.
74
75
76In this project, we choose to implement an interface to handle the threads. To do so, we introduce the MPI\_endpoint which is a
77concept proposed in the last MPI Forums and several papers have already discussed the importance of such idea and have introduced the
78framework of the MPI\_endpoint \cite{Dinan:2013}\cite{Sridharan:2014}. The concept of an endpoint is shown by Figure \ref{fig:scheme}. In
79the MPI\_endpoint environment, each OpenMP thread will be associated with a unique rank (global endpoint rank), an endpoint communicator,
80and a local rank (rank inside the MPI process) which is very similar to the \verb|OMP_thread_num|. The global endpoint rank will replace the
81role of the classic MPI rank and will be used in MPI communication calls.
82
83
84\begin{figure}[ht]
85\begin{center}
86\includegraphics[scale=0.4]{scheme.png}
87\end{center}
88\caption{}
89\label{fig:scheme}
90\end{figure}
91
92%XIOS is a library dedicated to IO management of climate code. It has a client-server pattern in which clients are in charge of computations and servers manage the reading and writing of files. The communication between clients and servers are handled by MPI. However, some of the climate models (\textit{e.g.} LMDZ) nowadays use an hybrid programming policy. Within a shared memory node, OpenMP directives are used to manage message exchanges. In such configuration, XIOS can not take full advantages of the computing resources to maximize the performance. This is because XIOS can only work with MPI processes. Before each call of XIOS routines, threads of one MPI process must gather their information to the master thread who works as an MPI process. After the call, the master thread distributes the updated information among its slave threads. As result, all slave threads have to wait while the master thread calls the XIOS routines. This introduce extra synchronization into the model and leads to not optimized performance. Aware of this situation, we need to develop a new version of XIOS (EP\_XIOS) which can work with threads, or in other words, can consider threads as they were processes. To do so, we introduce the MPI endpoints. 
93
94An other important aspect about the MPI\_endpoint interface is that each endpoints has knowledge of the ranks of other endpoints in the
95same communicator. This knowledge is necessary because when executing an MPI communication, for example a point-to-point exchange, we need
96to know not only the ranks of sender/receiver threads, but also the thread number of the sender/receiver threads and the MPI ranks
97of the sender/receiver processes. This ranking information is implemented inside an map object included in the endpoint communicator
98class.
99
100
101
102
103%\newpage
104
105%The MPI\_endpoint interface we implemented lies on top of an existing MPI Implementation. It consists of wrappers to all MPI functions
106%used in XIOS.
107
108
109In XIOS, we used the ``probe'' technique to search for arrived messages and then perform the receiving action. The principle is
110that sender process executes the send operation as usual. However, to minimize the time spent on waiting incoming messages, the receiver
111process calls in the first place the \verb|MPI_Probe| function to check if a message destinate to it has been published. If yes, the
112process execute in the second place the \verb|MPI_Recv| to receive the message. If not, the receiver process can carry on with other tasks
113or repeats the \verb|MPI_Probe| and \verb|MPI_Recv| actions if the required message is in immediate need. This technique works well in
114the current version of XIOS. However, if we introduce threads into this mechanism, problems can occur: The incoming message is labeled by
115the tag and receiver's MPI rank. Because threads within a process share the MPI rank, and the message probed is always available in the
116message queue, it can lead to the problem of data race and thus the message can be received by the wrong thread.
117
118
119To solve this problem, we introduce the ``matching-probe'' technique. The idea of the method is that each thread is equipped with a local
120incoming message queue. Each time a thread calls an MPI function, for example \verb|MPI_Recv|, it calls firstly the \verb|MPI_Mprobe|
121function to query the MPI incoming message with any tag and from any source. Once a message is probed, the thread gets the handle to the
122incoming message and this specific message is erased from the MPI message queue. Then, the thread proceed the identification of the message
123to get the destination thread's local rank and store the message handle to the local queue of the target thread. The thread repeats these
124steps until the MPI incoming message queue is empty. Then the thread we perform the usual ``probe'' technique to query its local incoming
125message queue to check if the required message is available. If yes, it performs the \verb|MPI_Recv| operation. With this ``matching-probe''
126technique, we can assure that a message is probed only once and is received by the right receiver.
127 
128
129
130Another issue needs to be clarified with this technique is that: how to identify the receiver's rank? The solution to this question is to
131use the tag argument. In the MPI environment, a tag is an integer ranging from 0 to $2^{31}$ depending on the Implementation. We can explore
132the large range property of the tag to store in it information about the source and destination thread ranks. In our endpoint interface, we
133choose to limit the first 15 bits for the tag used in the classic MPI communication, the next 8 bits to store the sender thread's local
134rank, and the last 8 bits to store the receiver thread's local rank (\textit{c.f.} Figure \ref{fig:tag}). In this way, with an extra
135analysis of the tag, we can identify the local ranks of the sender and the receiver in any P2P communication. As results, when a thread
136probes a message, it knows exactly in which local queue should store the probed message.
137
138\begin{figure}[ht]
139 \centering
140 \includegraphics[scale=0.4]{tag.png}
141 \caption{}\label{fig:tag}
142\end{figure}
143
144In Figure \ref{fig:tag}, Tag contains the user defined value for a certain communication. MPI\_tag is computed in the endpoint interface
145with help of the rank map and is used in the MPI calls.
146
147\begin{figure}[ht]
148\centering
149 \includegraphics[scale = 0.4]{sendrecv.png}
150\caption{This figure shows the classic pattern of a P2P communication with the endpoint interface. Thread/endpoint rank 0 sends a message
151to thread/endpoint rank 3 with tag=1. The underlying MPI function called by the sender is indeed a send for MPI rank of 1
152and tag=65537. From the receiver's point of view, the endpoint 3 is actually receiving a message from MPI rank 0 with
153tag=65537.}
154\label{fig:sendrecv}
155\end{figure}
156
157
158
159
160With the rank map, tag extension, and the matching-probe techniques, we are now able to call any P2P communication in the endpoint
161environment. For the collective communications, we apply a step-by-step execution pattern and no special technique is required. A
162step-by-step execution pattern consists of 3 steps (not necessarily in this order and not all steps are needed): arrangement of the source
163data, execution of the MPI function by all master/root threads, distribution or arrangement of the resulting data among threads.
164
165%The most representative functions of the collective communications are \verb|MPI_Gather| and \verb|MPI_Bcast|.
166
167For example, if we want to perform a broadcast operation, only 2 steps are needed (\textit{c.f.} Figure \ref{fig:bcast}). Firstly, the root
168thread, along with the master threads of other processes, perform the classic \verb|MPI_Bcast| operation. Secondly, the root thread, and the
169master threads send data to other threads via local memory transfer.
170
171\begin{figure}[ht]
172\centering
173\includegraphics[scale=0.3]{bcast.png}
174\caption{}
175\label{fig:bcast}
176\end{figure}
177
178Figure \ref{fig:allreduce} illustrates how the \verb|MPI_Allreduce| function is proceeded in the endpoint interface. First of all, We
179perform a intra-process ``allreduce'' operation: source data is reduced from slave threads to the master thread via local memory transfer.
180Next, all master threads call the classic \verb|MPI_Allreduce| routine. Finally, all master threads send the updated reduced data to its
181slaves via local memory transfer.
182
183\begin{figure}[ht]
184\centering
185\includegraphics[scale=0.3]{allreduce.png}
186\caption{}
187\label{fig:allreduce}
188\end{figure}
189
190Other MPI routines, such as \verb|MPI_Wait|, \verb|MPI_Intercomm_create| \textit{etc.}, can be found in the technique report of the
191endpoint interface.
192
193\section{The multi-threaded XIOS and performance results}
194
195The development of endpoint interface for thread-friendly XIOS library took about one year and a half. The main difficulty is the
196co-existence of MPI processes and OpenMP threads. All MPI classes must be redefined in the endpoint interface along with all the routines.
197The development is now available on the forge server: \url{http://forge.ipsl.jussieu.fr/ioserver/browser/XIOS/dev/branch_openmp}. One
198technique report is also available in which one can find more detail about how endpoint works and how the routines are implemented
199\cite{ep:2018}. We must note that the thread-friendly XIOS library is still in the phase of optimization. It will be released in the
200future with a stable version.
201
202All the functionalities of XIOS is reserved in its thread-friendly version. Single threaded code can work successfully with the new
203version of XIOS. For multi-threaded models, some modifications are needed in order to work with the multi-threaded XIOS library. Detail can
204be found in our technique report \cite{ep:2018}.
205
206Even though the multi-threaded XIOS library is not fully accomplished and further optimization in ongoing. We have already done some tests
207to see the potential of the endpoint framework. We take LMDZ as the target model and have tested with several work-flow charges.
208
209\subsection{LMDZ work-flow}
210
211In the LMDZ work-flow, we have a daily output file. We have up to 413 two-dimension variables and 187 three-dimension variables. According
212to user's need, we can change the ``output\_level'' key argument in the xml file to select the desired variables to be written.
213
214In our tests, we choose to set ``output\_level=2'' for a light output, and ``output\_level=11'' for a full output.
215
216\subsection{CMIP6 work-flow}
217
218\begin{comment}
219\section{Performance of LMDZ using EP\_XIOS}
220
221With the new version of XIOS, we are now capable of taking full advantages of the computing resources allocated by a simulation model when
222calling XIOS functions. All threads, can participate in XIOS as if they are MPI processes. We have tested the EP\_XIOS in LMDZ and the
223performance results are very encouraging.
224
225In our tests, we used 12 client processor with 8 threads each (96 XIOS clients in total), and one single-thread server processor. We have 2
226output densities. The light output gives mainly 2 dimensional fields while the heavy output records more 3D fields. We also have differente
227simulation duration settings: 1 day, 5 days, 15 days, and 31 days.
228
229\begin{figure}[ht]
230 \centering
231 \includegraphics[scale = 0.6]{LMDZ_perf.png}
232 \caption{Speedup obtained by using EP in LMDZ simulations.}
233\end{figure}
234
235In this figure, we show the speedup which is computed by $\displaystyle{\frac{time_{XIOS}}{time_{EP\_XIOS}}}$. The blue bars
236represent speedup of the XIOS file output and the red bars the speedup of LMDZ: calculates + XIOS file output. In all experimens,
237we can observe a speedup which represents a gain in performance. One important conclusion we can get from this result is that, more dense
238the output is, more efficient is the EP\_XIOS. With 8 threads per process, we can reach a speedup in XIOS upto 6, and a speedup of 1.5 in
239LMDZ which represents a decrease of the total execution time to 68\% ($\approx 1/1.5$). This observation confirmes steadily the importance
240of using EP in XIOS. 
241
242The reason why LMDZ does not show much speedup, is because the model is calcutation dominant: time spent on calculation is much longer than
243that on the file output. For example, if 30\% of the execution time is spent on the output, then with a speepup of 6, we can obtain a
244decrease in time of 25\%. Even the 25\% may seems to be small, it is still a gain in performance with existing computing resources.
245
246\section{Performance of EP\_XIOS}
247
248workfloz\_cmip6
249light output
25024*8+2
25130s - 52s
25232 days
253histmth with daily output
254
255\end{comment}
256
257
258\section{Future works for XIOS}
259
260
261\bibliographystyle{plain}
262\bibliography{reference}
263
264\end{document}         
Note: See TracBrowser for help on using the repository browser.