Hi All, Thank you Mohammad for your elaboration on the issues!
I have written most of the multi-gem5 patch so let me add some more clarifications and answer to your concerns. My comments are inline below. Thanks, - Gabor On 27/06/2015 10:20, "gem5-dev on behalf of Mohammad Alian" <gem5-dev-boun...@gem5.org<mailto:gem5-dev-boun...@gem5.org> on behalf of al...@wisc.edu<mailto:al...@wisc.edu>> wrote: Hi All, Curtis-Thank you for listing some of the differences. I was waiting for the completed multi-gem5 patch before I send my review. Please see my inline response below. I’ve addressed the concerns that you’ve raised. Also, I’ve added a bit more to the comparison. -* Synchronization. pd-gem5 implements this in Python (not a problem in itself; aesthetically this is nice, but...). The issue is that pd-gem5's data packets and barrier messages travel over different sockets. Since pd-gem5 could see data packets passing synchronization barriers, it could create an inconsistent checkpoint. multi-gem5's synchronization is implemented in C++ using sync events, but more importantly, the messages queue up in the same stream and so cannot have the issue just described. (Event ordering is often crucial in snapshot protocols.) Therefore we feel that multi-gem5 is a more robust solution in this respect. Each packet in pd-gem5 has a time-stamp. So even if data packets pass synchronization barriers (in another word data packets arrive early at the destination node), destination node process packets based on their timestamp. Actually allowing data packets to pass sync barriers is a nice feature that can reduce the likelihood of late packet reception. Ordering of data messages that flow over pd-gem5 nodes is also preserved in pd-gem5 implementation. This seems to be a misunderstanding. Maybe the wording was not precise before. The problem isn’t a data packet that “passing” a sync barrier but the other way around, a sync barrier that can pass a data packet (e.g. while the data packet is waiting in the host operating system socket layer). If that happens, the packet will arrive later than it was supposed to and it may miss the computed receive tick. For instance, let’s assume that the quantum coincides with the simulated Ether link delay. (This is the optimal choice of quantum to minimize the number of sync barriers.) If a data packet is sent right at the beginning of a quantum then this packet must arrive at the destination gem5 process within the same quantum in order not to miss its receive tick at the very beginning of the next quantum. If the sync barrier can pass the data packet then the data packet may arrive only during the next quantum (or in extreme conditions even later than that) so when it arrives the receiver gem5 may pass already the receive tick. Time-stamping does help with this issue. Also, if a data packet is waiting in the host operating system socket layer when the simulation thread exits to python to complete the next sync barrier then the packet will not go into the checkpoint that may follow that sync barrier. What you mentioned as an advantage for multi-gem5 is actually a key disadvantage: buffering sync messages behind data packets can add up to the synchronization overhead and slow down simulation significantly. The purpose of sync messages is to make sure that the data packets arrive in time (in terms of simulated time) at the destination so they can be scheduled for being received at the proper computed tick. Sync messages also make sure that no data packets are in flight when a sync barrier completes before we take a checkpoint. They definitely add overhead for the simulation but they are necessary for the correctness of the simulation. The receive thread in multi-gem5 reads out packets from the socket in parallel with the simulation thread so packets normally will not be "queueing up” before a sync barrier message. There is definitely rooms for improvement in the current implementation for reducing the synchronization overhead but that is likely true for pd-gem5, too. The important thing here is that the solution must provide correctness (robustness) first. Also, multi-gem5 send huge sized messages (multiHeaderPkt) through network to perform each synchronization point, which increases synchronization overhead further. In pd-gem5, we choose to send just one character as sync message through a separate socket to reduce synchronization overhead. The TCP/IP message size is unlikely the bottleneck here. Multi-gem5 will send ~50 bytes more in a sync barrier message than pd-gem5 but that bigger sync message still fits into a single ethernet frame on the wire. The end-to-end latency overhead that is caused by 50 bytes extra payload for a small single frame TCP/IP message is likely to fall into the ‘noise’ category if one tries to measure it in a real cluster. * Packet handling. pd-gem5 uses EtherTap for data packets but changed the polling mechanism to go through the main event queue. Since this rate is actually linked with simulator progress, it cannot guarantee that the packets are serviced at regular intervals of real time. This can lead to packets queueing up which would contribute to the synchronization issues mentioned above. multi-gem5 uses plain sockets with separate receive threads and so does not have this issue. I think again you are pointing to your first concern that I’ve explained above. Packets that have queued up in EtherTap socket, will be processed and delivered to simulation environment at the beginning of next simulation quantum. As I pointed out above, packet queued up in the EtherTap socket may miss the proper quantum to get received and/or a checkpoint to be saved. Please notice that multi-gem5 introduces a new simObjects to interface simulation environment to real world which is redundant. This functionality is already there by EtherTap. Except that the EtherTap solution does not provide a correct (robust) solution for the synchronization problem. * Checkpoint accuracy. A user would like to have a checkpoint at precisely the time the 'm5 checkpoint' operation is executed so as to not miss any of the area of interest in his application. pd-gem5 requires that simulation finish the current quantum before checkpointing, so it cannot provide this. (Shortening the quantum can help, but usually the snapshot is being taken while 'fast-forwarding', i.e. simulating as fast as possible, which would motivate a longer quantum.) multi-gem5 can enter the drain cycle immediately upon receiving a checkpoint request. We find this accuracy highly desirable. It’s true that if you have a large quantum size then there would be some discrepancy between the m5_ckpt instruction tick and the actual dump tick. Based on multi-gem5 code, my understanding is that you send async checkpoint message as soon as one of the gem5 processes encounter m5_ckpt instruction. But I’m not sure how you fix the aforementioned issue, because you have to sync all gem5 processes before you start dumping checkpoint, which necessitate a global synchronization beforehand. In multi-gem5, the gem5 process who encounters the m5_ckpt instruction sends out an async checkpoint notification for the peer gem5 processes and then it starts the draining immediately (at the same tick). So the checkpoint will be taken at the exact tick form the initiator process point of view. The global synchronisation with the peer processes takes place while the initiator process is still waiting at the same tick (i.e the simulation thread is suspended). However, the receiver thread continues reading out the socket - while waiting for the global sync to complete- to make sure that in-flight data packets from peer gem5 processes are stored properly and saved into the checkpoint. By the way, we have a fix for this issue by introducing a new m5 pseudo instruction. I fail to see how a new pseudo instruction can solve the problem of completing the full quantum in pd-gem5 before a checkpoint can be taken. Could you please elaborate on that? * Implementation of network topology. pd-gem5 uses a separate gem5 process to act as a switch whereas multi-gem5 uses a standalone packet relay process. We haven't measured the overhead of pd-gem5's simulated switch yet, but we're confident that our approach is at least as fast and more scalable. There is this flexibility in pd-gem5 to simulate a switch box alongside one of the other gem5 processes. However, it might make that gem5 process the simulation bottleneck. One of the advantages of pd-gem5 over multi-gem5 is that we use gem5 to simulate a switch box, which allows us to model any network topology by instantiating several Switch simObjects and interconnect them with EhterLink in an arbitrary fashion. A standalone tcp server just can provide switch functionality (forwarding packets to destinations) and model a star network topology. Furthermore, it cannot model various network timings such as queueing delay, congestion, and routing latency. Also it has some accuracy issues that I will point out next. I agree with the complex topology argument –I already mentioned that before as an advantage for pd-gem5 from the point of view of future extensions. However, I do not agree that multi-gem5 cannot model queueing delays and congestions. For a simple crossbar switch, it can model queueing delays and congestions, but the receive queues are distributed among gem5 processes. * Broken network timing: Forwarding packets between gem5 processes using a standalone tcp server can cause reordering between packets that have different source but same destination. It causes inaccurate network timing and worse of all non-deterministic simulation. pd-gem5 resolve this by reordering packets at Switch process and then send them to their destination (it’s possible as switch is synchronized with the rest of the nodes). In multi-gem5, there is always a HeaderPkt that contains some meta information for each data packet. The meta information include the send tick and the sender rank (i.e. a unique ID of the sender gem5 process). We use those information to define a well defined ordering of packets even if packets are arriving at the same receiver from different senders. This packet ordering scheme is still being tested so the corresponding patch is not on the RB yet. * Amount of changes pd-gem5 introduce different modes in etherlink just to provide accurate timing for each component in the network subsystem (NIC, link, switch) as well as capability of modeling different network topologies (mesh, ring, fat tree, etc). To enable a simple functionality, like what multi-gem5 provides, the amount of changes in gem5 can be limited to time-stamping packets and providing synchronization through python scripts. However, multi-gem5 re-implements functionalists that are already in gem5. This argument holds only if both implementations are correct (robust). It still seems to me that pd-gem5 does not provide correctness for the synchronization/checkpointing parts. * Integrating with gem5 mainstream: pd-gem5 launch script is written in python which is suited for integration with gem5 python scripts. However multi-gem5 uses bash script. Also, all source files in pd-gem5 are already parts of gem5 mainstream. However multi-gem5 has tcp_server.cc/hh that is a standalone process and cannot be part of gem5. The multi-gem5 launch script is simply enough to rely only on the shell. It can obviously be easily re-written in python if that added any value. The tcp_server component is only a utility (like the ‘m5’ utility that is also part of gem5). Cheers, - Gabor On Fri, Jun 26, 2015 at 8:40 PM, Curtis Dunham <curtis.dun...@arm.com<mailto:curtis.dun...@arm.com>> wrote: Hello everyone, We have taken a look at how pd-gem5 compares with multi-gem5. While intending to deliver the same functionality, there are some crucial differences: * Synchronization. pd-gem5 implements this in Python (not a problem in itself; aesthetically this is nice, but...). The issue is that pd-gem5's data packets and barrier messages travel over different sockets. Since pd-gem5 could see data packets passing synchronization barriers, it could create an inconsistent checkpoint. multi-gem5's synchronization is implemented in C++ using sync events, but more importantly, the messages queue up in the same stream and so cannot have the issue just described. (Event ordering is often crucial in snapshot protocols.) Therefore we feel that multi-gem5 is a more robust solution in this respect. * Packet handling. pd-gem5 uses EtherTap for data packets but changed the polling mechanism to go through the main event queue. Since this rate is actually linked with simulator progress, it cannot guarantee that the packets are serviced at regular intervals of real time. This can lead to packets queueing up which would contribute to the synchronization issues mentioned above. multi-gem5 uses plain sockets with separate receive threads and so does not have this issue. * Checkpoint accuracy. A user would like to have a checkpoint at precisely the time the 'm5 checkpoint' operation is executed so as to not miss any of the area of interest in his application. pd-gem5 requires that simulation finish the current quantum before checkpointing, so it cannot provide this. (Shortening the quantum can help, but usually the snapshot is being taken while 'fast-forwarding', i.e. simulating as fast as possible, which would motivate a longer quantum.) multi-gem5 can enter the drain cycle immediately upon receiving a checkpoint request. We find this accuracy highly desirable. * Implementation of network topology. pd-gem5 uses a separate gem5 process to act as a switch whereas multi-gem5 uses a standalone packet relay process. We haven't measured the overhead of pd-gem5's simulated switch yet, but we're confident that our approach is at least as fast and more scalable. Thanks, Curtis ________________________________________ From: gem5-dev [gem5-dev-boun...@gem5.org<mailto:gem5-dev-boun...@gem5.org>] On Behalf Of Mohammad Alian [ al...@wisc.edu<mailto:al...@wisc.edu>] Sent: Friday, June 26, 2015 7:37 PM To: gem5 Developer List Subject: Re: [gem5-dev] pd-gem5: simulating a parallel/distributed system on multiple physical hosts Hi Anthony, I think that would be a good option, then I can add pd-gem5 functionality on top of that. Right now I've simplified your implementation. Also, I think I had found some bugs in your patch that I cannot remember now. If you decided to ship EtherSwitch patch, let me know to give you a review on that. Thanks, Mohammad On Thu, Jun 25, 2015 at 8:36 PM, Gutierrez, Anthony < anthony.gutier...@amd.com<mailto:anthony.gutier...@amd.com>> wrote: > Would it make sense for me to ship the EtherSwitch patch first, since it > has utility on its own, and then we can decide which of the "multi-gem5" > approaches is best, or if it's some combination of both? > > The only reason I never shipped it was because Steve raised an issue that > I didn't have a good alternative for, and didn't have the time to look into > one at that time. > ________________________________________ > From: gem5-dev [gem5-dev-boun...@gem5.org<mailto:gem5-dev-boun...@gem5.org>] > on behalf of Mohammad Alian [ > al...@wisc.edu<mailto:al...@wisc.edu>] > Sent: Wednesday, June 24, 2015 12:43 PM > To: gem5 Developer List > Subject: Re: [gem5-dev] pd-gem5: simulating a parallel/distributed system > on multiple physical hosts > > Hi Andreas, > > Thanks for the comment. > I think the checkpointing support in both works is the same. Here is how > checkpointing support is implemented in pd-gem5: > > Whenever one of gem5 processes encounter an m5-checkpoint pseudo > instruction, it will send a “recv-ckpt” signal to the > “barrier” process. Then the “barrier” process sends a “take-ckpt” signal to > all the simulated nodes > (including the node that encountered m5-checkpoint) at the end of the > current simulation quantum. On the reception of > “take-ckpt” signal, gem5 processes start dumping check-points. This makes > each simulated node dump a checkpoint > at the same simulated time point while ensuring there is no in-flight > packets. > > I believe this is the same as multi-gem5 patch approach for checkpoint > support (based on the commit message of http://reviews.gem5.org/r/2865/ ). > Also, we have tested our mechanism with several benchmarks and it works. As > Steve suggested, I'll look into Curtis's patch and try to review it as > well. > But as Nilay also mentioned earlier, there are some codes missing in > Curtis's patch. I prefer to first run multi-gem5 before starting to review > it. > > Thank you, > Mohammad > > On Wed, Jun 24, 2015 at 7:25 AM, Andreas Hansson < andreas.hans...@arm.com<mailto:andreas.hans...@arm.com>> > wrote: > > > Hi Steve, > > > > Apologies for the confusion. We are on the same page. My point is that we > > cannot simply take a little bit of patch A and a little bit of patch B. > > This change involves a lot of code, and we need to approach this in a > > structured fashion. My proposal is to do it bottom up, and start by > > getting the basic support in place. Since > http://reviews.gem5.org/r/2826/ > > has already been on the review board for a few months, I am merely > > suggesting that the it would be a good start to relate the newly posted > > patches to what is already there. > > > > Andreas > > > > > > > > On 24/06/2015 13:11, "gem5-dev on behalf of Steve Reinhardt" > > <gem5-dev-boun...@gem5.org<mailto:gem5-dev-boun...@gem5.org> on behalf of > > ste...@gmail.com<mailto:ste...@gmail.com>> wrote: > > > > >Hi Andreas, > > > > > >I'm a little confused by your email---you say you're fundamentally > opposed > > >to looking at both patches and picking the best features, then you point > > >out that the patches Curtis posted have the feature of better > > >checkpointing > > >support so we should pick that :). > > > > > >Obviously we can't just pick patch A from Mohammad's set and patch B > from > > >Curtis's set and expect them to work together, but I think that having > > >both > > >sets of patches available and comparing and contrasting the two > > >implementations should enable us to get to a single implementation > that's > > >the best of both. Someone will have to make the effort of integrating > the > > >better ideas from one set into the other set to create a new unified set > > >of > > >patches; (or maybe we commit one set and then integrate the best of the > > >other set as patches on top of that), but the first step is to identify > > >what "the best of both" is. Having Mohammad look at Curtis's patches, > and > > >Curtis (or someone else from ARM) closely examine Mohammad's patches > would > > >be a great start. I intend to review them both, though unfortunately my > > >time has been scarce lately---I'm hoping to squeeze that in later this > > >week. > > > > > >Once we've had a few people look at both, we can discuss the pros and > cons > > >of each, then discuss the strategy for getting the best features in. So > > >far I've heard that Mohammad's patches have a better network model but > the > > >ARM patches have better checkpointing support; that seems like a good > > >start. > > > > > >Steve > > > > > >On Wed, Jun 24, 2015 at 12:26 AM Andreas Hansson < > andreas.hans...@arm.com<mailto:andreas.hans...@arm.com> > > > > > >wrote: > > > > > >> Hi all, > > >> > > >> Great work. However, I fundamentally do not believe in the approach of > > >> ‘letting reviewers pick the best features’. There is no way we would > > >>ever > > >> get something working out if it. We need to get _one_ working solution > > >> here, and figure out how to best get there. I would propose to do it > > >> bottom up, starting with the basic multi-simulator instance support, > > >> checkpointing support, and then move on to the network between the > > >> simulator instances. > > >> > > >> Thus, I propose we go with the low-level plumbing and checkpoint > support > > >> from what Curtis has posted. I believe proper checkpointing support to > > >>be > > >> the most challenging, and from what I can tell this is far more > limited > > >>in > > >> what you just posted Mohammad. Could you perhaps review Curtis patches > > >> based on your insights, and we can try and get these patches in shape > > >>and > > >> committed asap. > > >> > > >> Once we have the baseline functionality in place, then we can start > > >> looking at the more elaborate network models. > > >> > > >> Does this sound reasonable? > > >> > > >> Thanks, > > >> > > >> Andreas > > >> > > >> On 24/06/2015 05:05, "gem5-dev on behalf of Mohammad Alian" > > >> <gem5-dev-boun...@gem5.org<mailto:gem5-dev-boun...@gem5.org> on behalf > > >> of al...@wisc.edu<mailto:al...@wisc.edu>> wrote: > > >> > > >> >Hello All, > > >> > > > >> >I have submitted a chain of patches which enables gem5 to simulate a > > >> >cluster on multiple physical hosts: > > >> > > > >> >http://reviews.gem5.org/r/2909/ > > >> >http://reviews.gem5.org/r/2910/ > > >> >http://reviews.gem5.org/r/2912/ > > >> >http://reviews.gem5.org/r/2913/ > > >> >http://reviews.gem5.org/r/2914/ <http://reviews.gem5.org/r/2914/> > > >> > > > >> >and a patch that contains run scripts for a simple experiment: > > >> >http://reviews.gem5.org/r/2915/ > > >> > > > >> >We have run several benchmarks using this infrastructure, including > NAS > > >> >parallel benchmarks (MPI) and DCBench-hadoop > > >> >(http://prof.ict.ac.cn/DCBench/), > > >> >and would be happy to share scripts/diskimages. > > >> > > > >> >We call this *pd-gem5*. *pd-gem5 *functionality is more or less the > > >>same > > >> >as > > >> >Curtis's patch for *multi-gem5.* However, I feel *pd-gem5 *network > > >>model > > >> >is > > >> >more thorough; it also enables modeling different network topologies. > > >> >Having both set of changes together let reviewers to pick best > features > > >> >from both works. > > >> > > > >> >Thank you, > > >> >Mohammad Alian > > >> >_______________________________________________ > > >> >gem5-dev mailing list > > >> >gem5-dev@gem5.org<mailto:gem5-dev@gem5.org> > > >> >http://m5sim.org/mailman/listinfo/gem5-dev > > >> > > >> > > >> -- IMPORTANT NOTICE: The contents of this email and any attachments > are > > >> confidential and may also be privileged. If you are not the intended > > >> recipient, please notify the sender immediately and do not disclose > the > > >> contents to any other person, use it for any purpose, or store or copy > > >>the > > >> information in any medium. Thank you. > > >> > > >> ARM Limited, Registered office 110 Fulbourn Road, Cambridge CB1 9NJ, > > >> Registered in England & Wales, Company No: 2557590 > > >> ARM Holdings plc, Registered office 110 Fulbourn Road, Cambridge CB1 > > >>9NJ, > > >> Registered in England & Wales, Company No: 2548782 > > >> _______________________________________________ > > >> gem5-dev mailing list > > >> gem5-dev@gem5.org<mailto:gem5-dev@gem5.org> > > >> http://m5sim.org/mailman/listinfo/gem5-dev > > >> > > >_______________________________________________ > > >gem5-dev mailing list > > >gem5-dev@gem5.org<mailto:gem5-dev@gem5.org> > > >http://m5sim.org/mailman/listinfo/gem5-dev > > > > > > -- IMPORTANT NOTICE: The contents of this email and any attachments are > > confidential and may also be privileged. If you are not the intended > > recipient, please notify the sender immediately and do not disclose the > > contents to any other person, use it for any purpose, or store or copy > the > > information in any medium. Thank you. > > > > ARM Limited, Registered office 110 Fulbourn Road, Cambridge CB1 9NJ, > > Registered in England & Wales, Company No: 2557590 > > ARM Holdings plc, Registered office 110 Fulbourn Road, Cambridge CB1 9NJ, > > Registered in England & Wales, Company No: 2548782 > > _______________________________________________ > > gem5-dev mailing list > > gem5-dev@gem5.org<mailto:gem5-dev@gem5.org> > > http://m5sim.org/mailman/listinfo/gem5-dev > > > _______________________________________________ > gem5-dev mailing list > gem5-dev@gem5.org<mailto:gem5-dev@gem5.org> > http://m5sim.org/mailman/listinfo/gem5-dev > _______________________________________________ > gem5-dev mailing list > gem5-dev@gem5.org<mailto:gem5-dev@gem5.org> > http://m5sim.org/mailman/listinfo/gem5-dev > _______________________________________________ gem5-dev mailing list gem5-dev@gem5.org<mailto:gem5-dev@gem5.org> http://m5sim.org/mailman/listinfo/gem5-dev -- IMPORTANT NOTICE: The contents of this email and any attachments are confidential and may also be privileged. If you are not the intended recipient, please notify the sender immediately and do not disclose the contents to any other person, use it for any purpose, or store or copy the information in any medium. Thank you. ARM Limited, Registered office 110 Fulbourn Road, Cambridge CB1 9NJ, Registered in England & Wales, Company No: 2557590 ARM Holdings plc, Registered office 110 Fulbourn Road, Cambridge CB1 9NJ, Registered in England & Wales, Company No: 2548782 _______________________________________________ gem5-dev mailing list gem5-dev@gem5.org<mailto:gem5-dev@gem5.org> http://m5sim.org/mailman/listinfo/gem5-dev _______________________________________________ gem5-dev mailing list gem5-dev@gem5.org<mailto:gem5-dev@gem5.org> http://m5sim.org/mailman/listinfo/gem5-dev -- IMPORTANT NOTICE: The contents of this email and any attachments are confidential and may also be privileged. If you are not the intended recipient, please notify the sender immediately and do not disclose the contents to any other person, use it for any purpose, or store or copy the information in any medium. Thank you. ARM Limited, Registered office 110 Fulbourn Road, Cambridge CB1 9NJ, Registered in England & Wales, Company No: 2557590 ARM Holdings plc, Registered office 110 Fulbourn Road, Cambridge CB1 9NJ, Registered in England & Wales, Company No: 2548782 _______________________________________________ gem5-dev mailing list gem5-dev@gem5.org http://m5sim.org/mailman/listinfo/gem5-dev
