[Qemu-devel] [RFC PATCH RDMA support v5: 00/12] new formal protocol design

[mrhines]

From: "Michael R. Hines" <address@hidden>

Changes since v4:

- Created a "formal" protocol for the RDMA control channel
- Dynamic, chunked page registration now implemented on *both* the server and 
client
- Created new 'capability' for page registration
- Created new 'capability' for is_zero_page() - enabled by default
  (needed to test dynamic page registration)
- Created version-check before protocol begins at connection-time 
- no more migrate_use_rdma() !

NOTE: While dynamic registration works on both sides now,
      it does *not* work with cgroups swap limits. This functionality with 
infiniband
      remains broken. (It works fine with TCP). So, in order to take full 
      advantage of this feature, a fix will have to be developed on the kernel 
side.
      Alternative proposed is use /dev/<pid>/pagemap. Patch will be submitted.

Contents:
=================================
* Compiling
* Running (please readme before running)
* RDMA Protocol Description
* Versioning
* QEMUFileRDMA Interface
* Migration of pc.ram
* Error handling
* TODO
* Performance

COMPILING:
===============================

$ ./configure --enable-rdma --target-list=x86_64-softmmu
$ make

RUNNING:
===============================

First, decide if you want dynamic page registration on the server-side.
This always happens on the primary-VM side, but is optional on the server.
Doing this allows you to support overcommit (such as cgroups or ballooning)
with a smaller footprint on the server-side without having to register the
entire VM memory footprint. 
NOTE: This significantly slows down performance (about 30% slower).

$ virsh qemu-monitor-command --hmp \
    --cmd "migrate_set_capability chunk_register_destination on" # disabled by 
default

Next, if you decided *not* to use chunked registration on the server,
it is recommended to also disable zero page detection. While this is not
strictly necessary, zero page detection also significantly slows down
performance on higher-throughput links (by about 50%), like 40 gbps infiniband 
cards:

$ virsh qemu-monitor-command --hmp \
    --cmd "migrate_set_capability check_for_zero off" # always enabled by 
default

Finally, set the migration speed to match your hardware's capabilities:

$ virsh qemu-monitor-command --hmp \
    --cmd "migrate_set_speed 40g" # or whatever is the MAX of your RDMA device

Finally, perform the actual migration:

$ virsh migrate domain rdma:xx.xx.xx.xx:port

RDMA Protocol Description:
=================================

Migration with RDMA is separated into two parts:

1. The transmission of the pages using RDMA
2. Everything else (a control channel is introduced)

"Everything else" is transmitted using a formal 
protocol now, consisting of infiniband SEND / RECV messages.

An infiniband SEND message is the standard ibverbs
message used by applications of infiniband hardware.
The only difference between a SEND message and an RDMA
message is that SEND message cause completion notifications
to be posted to the completion queue (CQ) on the 
infiniband receiver side, whereas RDMA messages (used
for pc.ram) do not (to behave like an actual DMA).
    
Messages in infiniband require two things:

1. registration of the memory that will be transmitted
2. (SEND/RECV only) work requests to be posted on both
   sides of the network before the actual transmission
   can occur.

RDMA messages much easier to deal with. Once the memory
on the receiver side is registered and pinned, we're
basically done. All that is required is for the sender
side to start dumping bytes onto the link.

SEND messages require more coordination because the
receiver must have reserved space (using a receive
work request) on the receive queue (RQ) before QEMUFileRDMA
can start using them to carry all the bytes as
a transport for migration of device state.

To begin the migration, the initial connection setup is
as follows (migration-rdma.c):

1. Receiver and Sender are started (command line or libvirt):
2. Both sides post two RQ work requests
3. Receiver does listen()
4. Sender does connect()
5. Receiver accept()
6. Check versioning and capabilities (described later)

At this point, we define a control channel on top of SEND messages
which is described by a formal protocol. Each SEND message has a 
header portion and a data portion (but together are transmitted 
as a single SEND message).

Header:
    * Length  (of the data portion)
    * Type    (what command to perform, described below)
    * Version (protocol version validated before send/recv occurs)

The 'type' field has 7 different command values:
    1. None
    2. Ready             (control-channel is available) 
    3. QEMU File         (for sending non-live device state) 
    4. RAM Blocks        (used right after connection setup)
    5. Register request  (dynamic chunk registration) 
    6. Register result   ('rkey' to be used by sender)
    7. Register finished (registration for current iteration finished)

After connection setup is completed, we have two protocol-level
functions, responsible for communicating control-channel commands
using the above list of values: 

Logically:

qemu_rdma_exchange_recv(header, expected command type)

1. We transmit a READY command to let the sender know that 
   we are *ready* to receive some data bytes on the control channel.
2. Before attempting to receive the expected command, we post another
   RQ work request to replace the one we just used up.
3. Block on a CQ event channel and wait for the SEND to arrive.
4. When the send arrives, librdmacm will unblock us.
5. Verify that the command-type and version received matches the one we 
expected.

qemu_rdma_exchange_send(header, data, optional response header & data): 

1. Block on the CQ event channel waiting for a READY command
   from the receiver to tell us that the receiver
   is *ready* for us to transmit some new bytes.
2. Optionally: if we are expecting a response from the command
   (that we have no yet transmitted), let's post an RQ
   work request to receive that data a few moments later. 
3. When the READY arrives, librdmacm will 
   unblock us and we immediately post a RQ work request
   to replace the one we just used up.
4. Now, we can actually post the work request to SEND
   the requested command type of the header we were asked for.
5. Optionally, if we are expecting a response (as before),
   we block again and wait for that response using the additional
   work request we previously posted. (This is used to carry
   'Register result' commands #6 back to the sender which
   hold the rkey need to perform RDMA.

All of the remaining command types (not including 'ready')
described above all use the aformentioned two functions to do the hard work:

1. After connection setup, RAMBlock information is exchanged using
   this protocol before the actual migration begins.
2. During runtime, once a 'chunk' becomes full of pages ready to
   be sent with RDMA, the registration commands are used to ask the
   other side to register the memory for this chunk and respond
   with the result (rkey) of the registration.
3. Also, the QEMUFile interfaces also call these functions (described below)
   when transmitting non-live state, such as devices or to send
   its own protocol information during the migration process.

Versioning
==================================

librdmacm provides the user with a 'private data' area to be exchanged
at connection-setup time before any infiniband traffic is generated.

This is a convenient place to check for protocol versioning because the
user does not need to register memory to transmit a few bytes of version
information.

This is also a convenient place to negotiate capabilities
(like dynamic page registration).

If the version is invalid, we throw an error.

If the version is new, we only negotiate the capabilities that the
requested version is able to perform and ignore the rest.

QEMUFileRDMA Interface:
==================================

QEMUFileRDMA introduces a couple of new functions:

1. qemu_rdma_get_buffer()  (QEMUFileOps rdma_read_ops)
2. qemu_rdma_put_buffer()  (QEMUFileOps rdma_write_ops)

These two functions are very short and simply used the protocol
describe above to deliver bytes without changing the upper-level
users of QEMUFile that depend on a bytstream abstraction.

Finally, how do we handoff the actual bytes to get_buffer()?

Again, because we're trying to "fake" a bytestream abstraction
using an analogy not unlike individual UDP frames, we have
to hold on to the bytes received from control-channel's SEND 
messages in memory.

Each time we receive a complete "QEMU File" control-channel 
message, the bytes from SEND are copied into a small local holding area.

Then, we return the number of bytes requested by get_buffer()
and leave the remaining bytes in the holding area until get_buffer()
comes around for another pass.

If the buffer is empty, then we follow the same steps
listed above and issue another "QEMU File" protocol command,
asking for a new SEND message to re-fill the buffer.

Migration of pc.ram:
===============================

At the beginning of the migration, (migration-rdma.c),
the sender and the receiver populate the list of RAMBlocks
to be registered with each other into a structure.
Then, using the aforementioned protocol, they exchange a
description of these blocks with each other, to be used later 
during the iteration of main memory. This description includes
a list of all the RAMBlocks, their offsets and lengths and
possibly includes pre-registered RDMA keys in case dynamic
page registration was disabled on the server-side, otherwise not.

Main memory is not migrated with the aforementioned protocol, 
but is instead migrated with normal RDMA Write operations.

Pages are migrated in "chunks" (about 1 Megabyte right now).
Chunk size is not dynamic, but it could be in a future implementation.
There's nothing to indicate that this is useful right now.

When a chunk is full (or a flush() occurs), the memory backed by 
the chunk is registered with librdmacm and pinned in memory on 
both sides using the aforementioned protocol.

After pinning, an RDMA Write is generated and tramsmitted
for the entire chunk.

Chunks are also transmitted in batches: This means that we
do not request that the hardware signal the completion queue
for the completion of *every* chunk. The current batch size
is about 64 chunks (corresponding to 64 MB of memory).
Only the last chunk in a batch must be signaled.
This helps keep everything as asynchronous as possible
and helps keep the hardware busy performing RDMA operations.

Error-handling:
===============================

Infiniband has what is called a "Reliable, Connected"
link (one of 4 choices). This is the mode in which
we use for RDMA migration.

If a *single* message fails,
the decision is to abort the migration entirely and
cleanup all the RDMA descriptors and unregister all
the memory.

After cleanup, the Virtual Machine is returned to normal
operation the same way that would happen if the TCP
socket is broken during a non-RDMA based migration.

TODO:
=================================
1. Currently, cgroups swap limits for *both* TCP and RDMA
   on the sender-side is broken. This is more poignant for
   RDMA because RDMA requires memory registration.
   Fixing this requires infiniband page registrations to be
   zero-page aware, and this does not yet work properly.
2. Currently overcommit for the the *receiver* side of
   TCP works, but not for RDMA. While dynamic page registration
   *does* work, it is only useful if the is_zero_page() capability
   is remained enabled (which it is by default).
   However, leaving this capability turned on *significantly* slows
   down the RDMA throughput, particularly on hardware capable
   of transmitting faster than 10 gbps (such as 40gbps links).
3. Use of the recent /dev/<pid>/pagemap would likely solve some
   of these problems.
4. Also, some form of balloon-device usage tracking would also
   help aleviate some of these issues.

PERFORMANCE
===================

Using a 40gbps infinband link performing a worst-case stress test:

RDMA Throughput With $ stress --vm-bytes 1024M --vm 1 --vm-keep
Approximately 30 gpbs (little better than the paper)
1. Average worst-case throughput 
TCP Throughput With $ stress --vm-bytes 1024M --vm 1 --vm-keep
2. Approximately 8 gpbs (using IPOIB IP over Infiniband)

Average downtime (stop time) ranges between 28 and 33 milliseconds.

An *exhaustive* paper (2010) shows additional performance details
linked on the QEMU wiki:

http://wiki.qemu.org/Features/RDMALiveMigration