Network Working Group L. Dunbar
Internet Draft Futurewei
Intended status: Informational B. Sarikaya
Expires: May 18, 2020 Denpel Informatique
B.Khasnabish
Independent
T. Herbert
Intel
S. Dikshit
Aruba-HPE
November 18, 2019
Virtual Machine Mobility Solutions for L2 and L3 Overlay Networks
draft-ietf-nvo3-vmm-06
Abstract
This document discusses Virtual Machine (VM) mobility solutions that
are commonly used in overlay-based Data Center (DC) networks. The
objective is to describe the solutions and their impact on moving
VMs (and applications) from one rack to another connected by the
Overlay networks.
For layer 2 networks, it is based on using an NVA (Network
Virtualization Authority) - NVE (Network Virtualization Edge)
protocol to update the ARP (Address Resolution Protocol) table or
neighbor cache entries after a VM (virtual machine) moves from an
Old NVE to a New NVE. For Layer 3, it is based on migration of
address and connection after the move.
Status of this Memo
This Internet-Draft is submitted in full conformance with the
provisions of BCP 78 and BCP 79.This Internet-Draft is submitted in
full conformance with the provisions of BCP 78 and BCP 79. This
document may not be modified, and derivative works of it may not be
created, except to publish it as an RFC and to translate it into
languages other than English.
Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF), its areas, and its working groups. Note that
xxx, et al. Expires May 18, 2020 [Page 1]
Internet-Draft VM Mobility Solution November 18, 2019
other groups may also distribute working documents as Internet-
Drafts.
Internet-Drafts are draft documents valid for a maximum of six
months and may be updated, replaced, or obsoleted by other documents
at any time. It is inappropriate to use Internet-Drafts as
reference material or to cite them other than as "work in progress."
The list of current Internet-Drafts can be accessed at
http://www.ietf.org/ietf/1id-abstracts.txt
The list of Internet-Draft Shadow Directories can be accessed at
http://www.ietf.org/shadow.html
This Internet-Draft will expire on May 10, 2020.
Copyright Notice
Copyright (c) 2019 IETF Trust and the persons identified as the
document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal
Provisions Relating to IETF Documents
(http://trustee.ietf.org/license-info) in effect on the date of
publication of this document. Please review these documents
carefully, as they describe your rights and restrictions with
respect to this document. Code Components extracted from this
document must include Simplified BSD License text as described in
Section 4.e of the Trust Legal Provisions and are provided without
warranty as described in the Simplified BSD License.
Table of Contents
1. Introduction...................................................3
2. Conventions used in this document..............................4
3. Requirements...................................................5
4. Overview of the VM Mobility Solutions..........................5
4.1. VM Migration in Layer-2 Network...........................5
4.2. Task Migration in Layer-3 Network.........................7
4.2.1. Address and Connection Migration in Task Migration...8
5. Handling Packets in Flight.....................................9
Dunbar, et al. Expires May 18, 2020 [Page 2]
Internet-Draft VM Mobility Solution November 18, 2019
6. Moving Local State of VM......................................10
7. Handling of Hot, Warm and Cold VM Mobility....................10
8. VM Operation..................................................11
9. Security Considerations.......................................11
10. IANA Considerations..........................................12
11. Acknowledgments..............................................12
12. Change Log...................................................12
13. References...................................................12
13.1. Normative References....................................13
13.2. Informative References..................................14
1. Introduction
This document describes the overlay-based DC networking solutions
in support of multi-tenancy and VM mobility. Many large DCs,
especially Cloud DCs, host tasks (or workloads) for multiple
tenants. A tenant can be a department of one organization or an
organization. There is communication among tasks belonging to one
tenant and communication among tasks belonging to different
tenants or with external entities.
Server Virtualization, which is being used in almost all of
today's DCs, enables many VMs to run on a single physical computer
or server sharing the processor/memory/storage. Network
connectivity among VMs is provided by the network virtualization
edge (NVE) [RFC8014]. It is highly desirable [RFC7364] to allow
VMs to move dynamically (live, hot, or cold move) from one
server to another for dynamic load balancing or optimized workload
distribution.
There are many challenges and requirements related to VM mobility
in large data centers, including dynamically attaching/detaching
VMs to/from Virtual Network Edges (VNEs). In addition, retaining
the IP addresses after a move is a key requirement [RFC7364].
Such a requirement is needed in order to maintain existing
transport connections.
In traditional Layer-3 based networks, retaining IP addresses
after a move is generally not recommended because the frequent
move will cause fragmented IP addresses, which complicates IP
address management.
In view of many VM mobility schemes that exist today, there is a
need to document comprehensive VM mobility solutions that cover
both IPv4 and IPv6. Large DC networks can be organized as one
large (a) Layer-2 network geographically distributed across
buildings/cities or (b) Layer-3 networks with large number of host
routes that cannot be aggregated as a result of frequent moves
from one location to another without changing the IP addresses.
Dunbar, et al. Expires May 18, 2020 [Page 3]
Internet-Draft VM Mobility Solution November 18, 2019
The connectivity between Layer 2 boundaries can be achieved by the
NVE functioning as Layer-3 gateway, performing routing across
bridging domain such as in Warehouse Scale Computers (WSC).
2. Conventions used in this document
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL
NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and
"OPTIONAL" in this document are to be interpreted as described in
RFC 2119 [RFC2119] and [RFC8014].
This document uses the terminology defined in [RFC7364]. In
addition, we make the following definitions:
VM: Virtual Machine
Tasks: Task is a program instantiated or running on a virtual
machine or container. Tasks in virtual machines or
containers can be migrated from one server to another.
We use task, workload and virtual machine
interchangeably in this document.
Hot VM Mobility: A given VM could be moved from one server to
another in running state.
Warm VM Mobility: In case of warm VM mobility, the VM states are
mirrored to the secondary server (or domain) at a
predefined (configurable) regular intervals. This
reduces the overheads and complexity, but this may also
lead to a situation when both servers may not contain
the exact same data (state information)
Cold VM Mobility: A given VM could be moved from one server to
another in stopped or suspended state.
Old NVE: This refers to the old NVE where packets were forwarded
to before migration.
Dunbar, et al. Expires May 18, 2020 [Page 4]
Internet-Draft VM Mobility Solution November 18, 2019
New NVE: This refers to the new NVE after migration.
Packets in flight: This refers to the packets received by the Old
NVE sent by the correspondents that have old ARP or
neighbor cache entry before VM or task migration.
Users of VMs in diskless systems or the systems that are not
using configuration files are called end user clients.
Cloud DC: Third party DCs that host applications, tasks or
workloads and owned by different organizations or
tenants.
3. Requirements
This section states VM mobility requirements on DC networks.
DC networks should support both IPv4 and IPv6 VM mobility.
VM mobility should not require changing their IP addresses after the
move.
There exist "Hot Migration" where transport service continuity is
maintained, and "Cold Migration" where the transport service needs
to be restarted, i.e., execution of the tasks is stopped on the
"Old" NVE, moved to the "New" NVE and the task is restarted.
VM mobility solutions/procedures should minimize triangular routing
except for handling packets in flight.
VM mobility solutions/procedures should not need to use tunneling
except for handling packets in flight.
4. Overview of the VM Mobility Solutions
Layer-2 and Layer-3 mobility solutions are described respectively
in the following sections.
4.1. VM Migration in Layer-2 Network
Ability to move VMs dynamically, from one server to another, makes
it possible for dynamic load balancing or workload distribution.
Dunbar, et al. Expires May 18, 2020 [Page 5]
Internet-Draft VM Mobility Solution November 18, 2019
Therefore, this scheme is highly desirable for utilization in
large scale multi-tenant DCs.
In a Layer-2 based VM migration approach, a VM that is moving to
another server does not change its IP address. But since this VM
is now under a new NVE, previously communicating NVEs will
continue sending their packets to the Old NVE. To solve this
problem, Address Resolution Protocol (ARP) cache in IPv4 [RFC0826]
or neighbor cache in IPv6 [RFC4861] in the NVEs need to be updated
promptly. All NVEs need to change their caches associating the VM
Layer-2 or Medium Access Control (MAC) address with the new NVE's
IP address as soon as the VM moves. Such a change enables all NVEs
to encapsulate the outgoing MAC frames with the current target NVE
IP address. It may take some time to refresh the ARP/ND cache when
a VM has moved to a New NVE. During this period, a tunnel is
needed for that Old NVE to forward packets destined to the VM
under the New NVE.
In case of IPv4, immediately after the move, the VM should send a
gratuitous ARP request message containing its IPv4 and Layer-2 MAC
address to its new NVE. This message's destination address is the
broadcast address. Upon receiving this message, both old and new
NVEs should update the VM's ARP entry in the central directory at
the NVA, to update its mappings to record the IPv4 address and MAC
address of the moving VM along with the new NVE IPv4 address. An
NVE-to-NVA protocol is used for this purpose [RFC8014].
Reverse ARP (RARP) which enables the host to discover its IPv4
address when it boots from a local server [RFC0903], is not used
by VMs because the VM already knows its IPv4 address. Next, we
describe a case where RARP is used.
There are some vendor deployments (e.g., diskless systems or
systems without configuration files) where the VM's user, i.e.,
end-user client asks for the same MAC address upon migration.
This can be achieved by the clients sending RARP request message
which carries the MAC address looking for an IP address
allocation. The server, in this case the new NVE, needs to
communicate with NVA, just like in the gratuitous ARP case to
ensure that the same IPv4 address is assigned to the VM. NVA uses
the MAC address as the key in the search of ARP cache to find the
IP address and informs this to the new NVE which in turn sends
RARP reply message. This completes IP address assignment to the
migrating VM.
Other NVEs communicating with this VM could have the old ARP
entry. If any VMs in those NVEs need to communicate with the VM
Dunbar, et al. Expires May 18, 2020 [Page 6]
Internet-Draft VM Mobility Solution November 18, 2019
attached to the new NVE, old ARP entries might be used. Thus, the
packets are delivered to the old NVE. The old NVE MUST tunnel
these in-flight packets to the new NVE.
When an ARP entry for those VMs times out, their corresponding
NVEs should access the NVA for an update.
IPv6 operation is slightly different:
In IPv6, after the move, the VM immediately sends an unsolicited
neighbor advertisement message containing its IPv6 address and
Layer-2 MAC address to its new NVE. This message is sent to the
IPv6 Solicited Node Multicast Address corresponding to the target
address which is the VM's IPv6 address. The NVE receiving this
message should send request to update VM's neighbor cache entry in
the central directory of the NVA. The NVA's neighbor cache entry
should include IPv6 address of the VM, MAC address of the VM and
the NVE IPv6 address. An NVE-to-NVA protocol is used for this
purpose [RFC8014].
Other NVEs communicating with this VM might still use the old
neighbor cache entry. If any VM in those NVEs need to communicate
with the VM attached to the new NVE, it could use the old neighbor
cache entry. Thus, the packets are delivered to the old NVE. The
old NVE MUST tunnel these in-flight packets to the new NVE.
When a neighbor cache entry in those VMs times out, their
corresponding NVEs should access the NVA for an update.
4.2. Task Migration in Layer-3 Network
Layer-2 based DC networks become quickly prohibitive because
ARP/neighbor caches don't scale. Scaling can be accomplished
seamlessly in Layer-3 data center networks by just giving each
virtual network an IP subnet and a default route that points to
its NVE. This means no explosion of ARP/ neighbor cache in VMs
and NVEs (just one ARP/ neighbor cache entry for the default
route) and there is no need to have Ethernet header in
encapsulation [RFC7348] which saves at least 16 bytes.
Even though the term VM and Task are used interchangeably in this
document, the term Task is used in the context of Layer-3
migration mainly to have slight emphasis on the task of moving an
entity that is instantiated on a VM or a container.
Dunbar, et al. Expires May 18, 2020 [Page 7]
Internet-Draft VM Mobility Solution November 18, 2019
Traditional Layer-3 based DC networks require IP address of the
task to change after moving because the pre-fixes of the IP
address usually reflect the locations. It is necessary to have an
IP based VM migration solution that can allow IP addresses staying
the same after the VMs move to different locations. The Identifier
Locator Addressing or ILA [I-D.herbert-nvo3-ila] is one of such
solutions.
Because broadcasting is not available in Layer-3 based networks,
multicast of neighbor solicitations in IPv6 would need to be
emulated.
Cold task migration, which is a common practice in many data
centers, involves the following steps:
- Stop running the task.
- Package the runtime state of the job.
- Send the runtime state of the task to the new NVE where the
task is to run.
- Instantiate the task's state on the new machine.
- Start the tasks continuing it from the point at which it was
stopped.
Address migration and connection migration in moving tasks or VMs
are addressed next.
4.2.1. Address and Connection Migration in Task Migration
Address migration is achieved as follows:
- Configure IPv4/v6 address on the target Task.
- Suspend use of the address on the old Task. This includes
handling established connections. A state may be established
to drop packets or send ICMPv4 or ICMPv6 destination
unreachable message when packets to the migrated address are
received.
- Push the new mapping to VM. Communicating VMs will learn of
the new mapping via a control plane either by participating in
a protocol for mapping propagation or by getting the new
mapping from a central database such as Domain Name System
(DNS).
Dunbar, et al. Expires May 18, 2020 [Page 8]
Internet-Draft VM Mobility Solution November 18, 2019
Connection migration involves reestablishing existing TCP
connections of the task in the new place.
The simplest course of action is to drop all TCP connections to
the VM across a migration. If the migrations are relatively rare
events in a data center, impact is relatively small when TCP
connections are automatically closed in the network stack during a
migration event. If the applications running are known to handle
this gracefully (i.e. reopen dropped connections) then this
approach may be viable.
More involved approach to connection migration entails pausing the
connection, packaging connection state and sending to target,
instantiating connection state in the peer stack, and restarting
the connection. From the time the connection is paused to the
time it is running again in the new stack, packets received for
the connection could be silently dropped. For some period of
time, the old stack will need to keep a record of the migrated
connection. If it receives a packet, it can either silently drop
the packet or forward it to the new location, as described in
Section 5.
5. Handling Packets in Flight
The Old NVE may receive packets from the VM's ongoing
communications. These packets should not be lost; they should be
sent to the New NVE to be delivered to the VM. The steps involved
in handling packets in flight are as follows:
Preparation Step: It takes some time, possibly a few seconds for
a VM to move from its Old NVE to a New NVE. During this period, a
tunnel needs to be established so that the Old NVE can forward
packets to the New NVE. Old NVE gets New NVE address from NVA in
the request to move the VM. The Old NVE can store the New NVE
address for the VM with a timer. When the timer expired, the entry
for the New NVE for the VM can be deleted.
Tunnel Establishment - IPv6: Inflight packets are tunneled to the
New NVE using the encapsulation protocol such as VXLAN in IPv6.
Tunnel Establishment - IPv4: Inflight packets are tunneled to the
New NVE using the encapsulation protocol such as VXLAN in IPv4.
Tunneling Packets - IPv6: IPv6 packets received for the migrating
VM are encapsulated in an IPv6 header at the Old NVE. New NVE
decapsulates the packet and sends IPv6 packet to the migrating VM.
Dunbar, et al. Expires May 18, 2020 [Page 9]
Internet-Draft VM Mobility Solution November 18, 2019
Tunneling Packets - IPv4: IPv4 packets received for the migrating
VM are encapsulated in an IPv4 header at the Old NVE. New NVE
decapsulates the packet and sends IPv4 packet to the migrating VM.
Stop Tunneling Packets: When the Timer for storing the New NVE
address for the VM expires. The Timer should be long enough for
all other NVEs that need to communicate with the VM to get their
NVE-VM cache entries updated.
6. Moving Local State of VM
In addition to the VM mobility related signaling (VM Mobility
Registration Request/Reply), the VM state needs to be transferred
to the New NVE. The state includes its memory and file system if
the VM cannot access the memory and the file system after moving
to the New NVE. Old NVE opens a TCP connection with New NVE over
which VM's memory state is transferred.
File system or local storage is more complicated to transfer. The
transfer should ensure consistency, i.e. the VM at the New NVE
should find the same file system it had at the Old NVE. Pre-
copying is a commonly used technique for transferring the file
system. First the whole disk image is transferred while VM
continues to run. After the VM is moved, any changes in the file
system are packaged together and sent to the New NVE Hypervisor
which reflects these changes to the file system locally at the
destination.
7. Handling of Hot, Warm and Cold VM Mobility
Both Cold and Warm VM mobility (migration), refers to the VM being
completely shut down at the old NVE before restarted at the new
NVE. Therefore, all transport services to the VM need to restart.
Upon starting at the new NVE, the VM should send an ARP or
Neighbor Discovery message. Cold VM mobility also allows the Old
NVE and all communicating NVEs to time out ARP/neighbor cache
entries of the VM. It is necessary for the NVA to push the
updated ARP/neighbor cache entry to NVEs or for NVEs to pull the
updated ARP/neighbor cache entry from NVA.
The Cold VM mobility can be facilitated by cold standby entity
receiving scheduled backup information. The cold standby entity
can be a VM or other form factors which is beyond the scope of
this document. The cold mobility option can be used for non-
critical applications and services that can tolerate interrupted
TCP connections.
Dunbar, et al. Expires May 18, 2020 [Page 10]
Internet-Draft VM Mobility Solution November 18, 2019
The Warm VM mobility refers the backup entities receive backup
information at more frequent intervals. The duration of the
interval determines the warmth of the option. The larger the
duration, the less warm (and hence cold) the Warm VM mobility
option becomes.
There is also a Hot Standby option in addition to the Hot
Mobility, where there are VMs in both primary and secondary NVEs.
They have identical information and can provide services
simultaneously as in load-share mode of operation. If the VM in
the primary NVE fails, there is no need to actively move the VM to
the secondary NVE because the VM in the secondary NVE already
contains identical information. The Hot Standby option is the
costliest mechanism, and hence this option is utilized only for
mission-critical applications and services. In Hot Standby
option, regarding TCP connections, one option is to start with and
maintain TCP connections to two different VMs at the same time.
The least loaded VM responds first and starts providing service
while the sender (origin) still continues to receive Ack from the
heavily loaded (secondary) VM and chooses not to use the service
of the secondary responding VM. If the situation (loading
condition of the primary responding VM) changes the secondary VM
may start providing service to the sender (origin).
8. VM Operation
Once a VM moves to a new NVE, the VM's IP address does not change
and the VM should be able to continue to receive packets to its
address(es).
The VM needs to send a gratuitous Address Resolution message or
unsolicited Neighbor Advertisement message upstream after each
move.
The VM lifecycle management is a complicated task, which is beyond
the scope of this document. Not only it involves monitoring server
utilization, balancing the distribution of workload, etc., but
also needs seamless management VM migration from one server to
another.
9. Security Considerations
Security threats for the data and control plane for overlay
networks are discussed in [RFC8014]. There are several issues in
a multi-tenant environment that create problems. In Layer-2 based
overlay DC networks, lack of security in VXLAN, and corruption of
VNI can lead to delivery of information to the wrong tenant.
Dunbar, et al. Expires May 18, 2020 [Page 11]
Internet-Draft VM Mobility Solution November 18, 2019
Also, ARP in IPv4 and ND in IPv6 are not secure, especially if we
accept the gratuitous versions. When these are done over a UDP
encapsulation, as in VXLAN, the problem gets worse since it is
trivial for a non-trusted entity to spoof UDP packets.
In Layer-3 based overlay data center networks, the problem of
address spoofing may arise. An NVE may have untrusted tasks
attached to it. This usually happens in situations when the VMs
(tasks) running third party applications. This requires the usage
of stronger security mechanisms.
10. IANA Considerations
This document makes no request to IANA.
11. Acknowledgments
The authors are grateful to Bob Briscoe, David Black, Dave R.
Worley, Qiang Zu, and Andrew Malis for helpful comments.
12. Change Log
. submitted version -00 as a working group draft after adoption
. submitted version -01 with these changes: references are updated,
o added packets in flight definition to Section 2
. submitted version -02 with updated address.
. submitted version -03 to fix the nits.
. submitted version -04 in reference to the WG Last call comments.
. Submitted version - 05 to address IETF LC comments from TSV area.
13. References
Dunbar, et al. Expires May 18, 2020 [Page 12]
Internet-Draft VM Mobility Solution November 18, 2019
13.1. Normative References
[RFC0826] Plummer, D., "An Ethernet Address Resolution Protocol: Or
Converting Network Protocol Addresses to 48.bit Ethernet
Address for Transmission on Ethernet Hardware", STD 37,
RFC 826, DOI 10.17487/RFC0826, November 1982,
<https://www.rfc-editor.org/info/rfc826>.
[RFC0903] Finlayson, R., Mann, T., Mogul, J., and M. Theimer, "A
Reverse Address Resolution Protocol", STD 38, RFC 903,
DOI 10.17487/RFC0903, June 1984, <https://www.rfc-
editor.org/info/rfc903>.
[RFC2119] Bradner, S., "Key words for use in RFCs to Indicate
Requirement Levels", BCP 14, RFC 2119, March 1997.
[RFC2629] Rose, M., "Writing I-Ds and RFCs using XML", RFC 2629,
DOI 10.17487/RFC2629, June 1999, <https://www.rfc-
editor.org/info/rfc2629>.
[RFC4861] Narten, T., Nordmark, E., Simpson, W., and H. Soliman,
"Neighbor Discovery for IP version 6 (IPv6)", RFC 4861,
DOI 10.17487/RFC4861, September 2007, <https://www.rfc-
editor.org/info/rfc4861>.
[RFC7348] Mahalingam, M., Dutt, D., Duda, K., Agarwal, P.,
Kreeger, L., Sridhar, T., Bursell, M., and C. Wright,
"Virtual eXtensible Local Area Network (VXLAN): A
Framework for Overlaying Virtualized Layer 2 Networks over
Layer 3 Networks", RFC 7348, DOI 10.17487/RFC7348, August
2014, <https://www.rfc-editor.org/info/rfc7348>.
[RFC7364] Narten, T., Ed., Gray, E., Ed., Black, D., Fang, L.,
Kreeger, L., and M. Napierala, "Problem Statement:
Overlays for Network Virtualization", RFC 7364, DOI
10.17487/RFC7364, October 2014, <https://www.rfc-
editor.org/info/rfc7364>.
Dunbar, et al. Expires May 18, 2020 [Page 13]
Internet-Draft VM Mobility Solution November 18, 2019
[RFC8014] Black, D., Hudson, J., Kreeger, L., Lasserre, M., and T.
Narten, "An Architecture for Data-Center Network
Virtualization over Layer 3 (NVO3)", RFC 8014, DOI
10.17487/RFC8014, December 2016, <https://www.rfc-
editor.org/info/rfc8014>.
13.2. Informative References
[I-D.herbert-nvo3-ila] Herbert, T. and P. Lapukhov, "Identifier-
locator addressing for IPv6", draft-herbert-nvo3-ila-04
(work in progress), March 2017.
Dunbar, et al. Expires May 18, 2020 [Page 14]
Internet-Draft VM Mobility Solution November 18, 2019
Authors' Addresses
Linda Dunbar
Futurewei
Email: ldunbar@futurewei.com
Behcet Sarikaya
Denpel Informatique
Email: sarikaya@ieee.org
Bhumip Khasnabish
Independent
Email: vumip1@gmail.com
Tom Herbert
Intel
Email: tom@herbertland.com
Saumya Dikshit
Aruba-HPE
Bangalore, India
Email: saumya.dikshit@hpe.com
Dunbar, et al. Expires May 18, 2020 [Page 15]