[escalator.git] / doc / 02-Background_and_Terminologies.rst

General Requirements Background and Terminology\r
-----------------------------------------------\r
\r
Terminologies and definitions\r
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\r
\r
NFVI\r
  The term is an abbreviation for Network Function Virtualization\r
  Infrastructure; sometimes it is also referred as data plane in this\r
  document.\r
\r
VIM\r
  The term is an abbreviation for Virtual Infrastructure Management;\r
  sometimes it is also referred as control plane in this document.\r
\r
Operator\r
  The term refers to network service providers and Virtual Network\r
  Function (VNF) providers.\r
\r
End-User\r
  The term refers to a subscriber of the Operator's services.\r
\r
Network Service\r
  The term refers to a service provided by an Operator to its\r
  End-users using a set of (virtualized) Network Functions\r
\r
Infrastructure Services\r
  The term refers to services provided by the NFV Infrastructure and the\r
  the Management & Orchestration functions to the VNFs. I.e.\r
  these are the virtual resources as perceived by the VNFs.\r
\r
Smooth Upgrade\r
  The term refers to an upgrade that results in no service outage\r
  for the end-users.\r
\r
Rolling Upgrade\r
  The term refers to an upgrade strategy that upgrades each node or\r
  a subset of nodes in a wave style rolling through the data centre. It\r
  is a popular upgrade strategy to maintain service availability.\r
\r
Parallel Universe Upgrade\r
  The term refers to an upgrade strategy that creates and deploys\r
  a new universe - a system with the new configuration - while the old\r
  system continues running. The state of the old system is transferred\r
  to the new system after sufficient testing of the new system.\r
\r
Infrastructure Resource Model\r
  The term refers to the representation of infrastructure resources,\r
  namely: the physical resources, the virtualization\r
  facility resources and the virtual resources.\r
\r
Physical Resource\r
  The term refers to a hardware pieces of the NFV infrastructure, which may\r
  also include the firmware which enables the hardware.\r
\r
Virtual Resource\r
  The term refers to a resource, which is provided as services built on top\r
  of the physical resources via the virtualization facilities; in particular,\r
  they are the resources on which VNF entities are deployed, e.g.\r
  the VMs, virtual switches, virtual routers, virtual disks etc.\r
\r
Visualization Facility\r
  The term refers to a resource that enables the creation\r
  of virtual environments on top of the physical resources, e.g.\r
  hypervisor, OpenStack, etc.\r
\r
Upgrade Campaign\r
  The term refers to a choreography that describes how the upgrade should\r
  be performed in terms of its targets (i.e. upgrade objects), the\r
  steps/actions required of upgrading each, and the coordination of these\r
  steps so that service availability can be maintained. It is an input to an\r
  upgrade tool (Escalator) to carry out the upgrade.\r
\r
Upgrade Duration\r
  The duration of an upgrade characterized by the time elapsed between its\r
  initiation and its completion. E.g. from the moment the execution of an\r
  upgrade campaign has started until it has been committed. Depending on\r
  the upgrade method and its target some parts of the system may be in a more\r
  vulnerable state.\r
\r
Outage\r
  The period of time during which a given service is not provided is referred\r
  as the outage of that given service. If a subsystem or the entire system\r
  does not provide any service, it is the outage of the given subsystem or the\r
  system. Smooth upgrade means upgrade with no outage for the user plane, i.e.\r
  no VNF should experience service outage.\r
\r
Rollback\r
  The term refers to a failure handling strategy that reverts the changes\r
  done by a potentially failed upgrade execution one by one in a reverse order.\r
  I.e. it is like undoing the changes done by the upgrade.\r
\r
Restore\r
  The term refers to a failure handling strategy that reverts the changes\r
  done by an upgrade by restoring the system from some backup data. This\r
  results in the loss of any data persisted since the backup has been taken.\r
\r
Rollforward\r
  The term refers to a failure handling strategy applied after a restore\r
  (from a backup) opertaion to recover any loss of data persisted between\r
  the time the backup has been taken and the moment it is restored. Rollforward\r
  requires that data that needs to survive the restore operation is logged at\r
  a location not impacted by the restore so that it can be re-applied to the\r
  system after its restoration from the backup.\r
\r
Downgrade\r
  The term refers to an upgrade in which an earlier version of the software\r
  is restored through the upgrade procedure. A system can be downgraded to any\r
  earlier version and the compatibility of the versions will determine the\r
  applicable upgrade strategies and whether service outage can be avoided.\r
  In particular any data conversion needs special attention.\r
\r
\r
\r
Upgrade Objects\r
~~~~~~~~~~~~~~~\r
\r
Physical Resource\r
^^^^^^^^^^^^^^^^^\r
\r
Most cloud infrastructures support the dynamic addition/removal of\r
hardware. Accordingly a hardware upgrade could be done by adding the new\r
piece of hardware and removing the old one. From the persepctive of smooth\r
upgrade the orchestration/scheduling of this actions is the primary concern.\r
Upgrading a physical resource may involve as well the upgrade of its firmware\r
and/or modifying its configuration data. This may require the restart of the\r
hardware.\r
\r
\r
\r
Virtual Resources\r
^^^^^^^^^^^^^^^^^\r
\r
Addition and removal of virtual resources may be initiated by the users or be\r
a result of an elasticity action. Users may also request the upgrade of their\r
virtual resources using a new VM image.\r
\r
.. Needs to be moved to requirement section: Escalator should facilitate such an\r
option and allow for a smooth upgrade.\r
\r
On the other hand changes in the infrastructure, namely, in the hardware and/or\r
the virtualization facility resources may result in the upgrade of the virtual\r
resources. For example if by some reason the hypervisor is changed and\r
the current VMs cannot be migrated to the new hypervisor - they are\r
incompatible - then the VMs need to be upgraded too. This is not\r
something the NFVI user (i.e. VNFs ) would know about. In such cases\r
smooth upgrade is essential.\r
\r
\r
Virtualization Facility Resources\r
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\r
\r
Based on the functionality they provide, virtualization facility\r
resources could be divided into computing node, networking node,\r
storage node and management node.\r
\r
The possible upgrade objects in these nodes are addressed below:\r
(Note: hardware based virtualization may be considered as virtualization\r
facility resource, but from escalator perspective, it is better to\r
consider it as part of the hardware upgrade. )\r
\r
**Computing node**\r
\r
1. OS Kernel\r
\r
2. Hypvervisor and virtual switch\r
\r
3. Other kernel modules, like driver\r
\r
4. User space software packages, like nova-compute agents and other\r
   control plane programs.\r
\r
Updating 1 and 2 will cause the loss of virtualzation functionality of\r
the compute node, which may lead to data plane services interruption\r
if the virtual resource is not redudant.\r
\r
Updating 3 might result the same.\r
\r
Updating 4 might lead to control plane services interruption if not an\r
HA deployment.\r
\r
**Networking node**\r
\r
1. OS kernel, optional, not all switches/routers allow the upgrade their\r
   OS since it is more like a firmware than a generic OS.\r
\r
2. User space software package, like neutron agents and other control\r
   plane programs\r
\r
Updating 1 if allowed will cause a node reboot and therefore leads to\r
data plane service interruption if the virtual resource is not\r
redundant.\r
\r
Updating 2 might lead to control plane services interruption if not an\r
HA deployment.\r
\r
**Storage node**\r
\r
1. OS kernel, optional, not all storage nodes allow the upgrade their OS\r
   since it is more like a firmware than a generic OS.\r
\r
2. Kernel modules\r
\r
3. User space software packages, control plane programs\r
\r
Updating 1 if allowed will cause a node reboot and therefore leads to\r
data plane services interruption if the virtual resource is not\r
redundant.\r
\r
Update 2 might result in the same.\r
\r
Updating 3 might lead to control plane services interruption if not an\r
HA deployment.\r
\r
**Management node**\r
\r
1. OS Kernel\r
\r
2. Kernel modules, like driver\r
\r
3. User space software packages, like database, message queue and\r
   control plane programs.\r
\r
Updating 1 will cause a node reboot and therefore leads to control\r
plane services interruption if not an HA deployment. Updating 2 might\r
result in the same.\r
\r
Updating 3 might lead to control plane services interruption if not an\r
HA deployment.\r
\r
\r
\r
\r
\r
Upgrade Granularity\r
~~~~~~~~~~~~~~~~~~~\r
\r
The granularity of an upgrade can be characterized from two perspective:\r
- the physical dimension and\r
- the software dimension\r
\r
\r
Physical Dimension\r
^^^^^^^^^^^^^^^^^^\r
\r
The physical dimension characterizes the number of similar upgrade objects\r
targeted by the upgrade, i.e. whether it is full / partial upgrade of a\r
data centre, cluster, zone.\r
Because of the upgrade of a data centre or a zone, it may be divided into\r
several batches. Thus there is a need for efficiency in the execution of\r
upgrades of potentially huge number of upgrade objects while still maintain\r
availability to fulfill the requirement of smooth upgrade.\r
\r
The upgrade of a cloud environment (cluster) may also\r
be partial. For example, in one cloud environment running a number of\r
VNFs, we may just try to upgrade one of them to check the stability and\r
performance, before we upgrade all of them.\r
Thus there is a need for proper organization of the artifacts associated with\r
the different upgrade objects. Also the different versions should be able\r
to coextist beyond the upgrade period.\r
\r
From this perspective special attention may be needed when upgrading\r
objects that are collaborating in a redundancy schema as in this case\r
different versions not only need to coexist but also collaborate. This\r
puts requirement on the upgrade objects primarily. If this is not possible\r
the upgrade campaign should be designed in such a way that the proper\r
isolation is ensured.\r
\r
Software Dimension\r
^^^^^^^^^^^^^^^^^^\r
\r
The software dimension of the upgrade characterizes the upgrade object\r
type targeted and the combination in which they are upgraded together.\r
\r
Even though the upgrade may\r
initially target only one type of upgrade object, e.g. the hypervisor\r
the dependency of other upgrade objects on this initial target object may\r
require their upgrade as well. I.e. the upgrades need to be combined. From this\r
perspective the main concern is compatibility of the dependent and\r
sponsor objects. To take into consideration of these dependencies\r
they need to be described together with the version compatility information.\r
Breaking dependencies is the major cause of outages during upgrades.\r
\r
In other cases it is more efficient to upgrade a combination of upgrade\r
objects than to do it one by one. One aspect of the combination is how\r
the upgrade packages can be combined, whether a new image can be created for\r
them before hand or the different packages can be installed during the upgrade\r
independently, but activated together.\r
\r
The combination of upgrade objects may span across\r
layers (e.g. software stack in the host and the VM of the VNF).\r
Thus, it may require additional coordination between the management layers.\r
\r
With respect to each upgrade object type and even stacks we can\r
distingush major and minor upgrades:\r
\r
**Major Upgrade**\r
\r
Upgrades between major releases may introducing significant changes in\r
function, configuration and data, such as the upgrade of OPNFV from\r
Arno to Brahmaputra.\r
\r
**Minor Upgrade**\r
\r
Upgrades inside one major releases which would not leads to changing\r
the structure of the platform and may not infect the schema of the\r
system data.\r
\r
Scope of Impact\r
~~~~~~~~~~~~~~~\r
\r
Considering availability and therefore smooth upgrade, one of the major\r
concerns is the predictability and control of the outcome of the different\r
upgrade operations. Ideally an upgrade can be performed without impacting any\r
entity in the system, which means none of the operations change or potentially\r
change the behaviour of any entity in the system in an uncotrolled manner.\r
Accordingly the operations of such an upgrade can be performed any time while\r
the system is running, while all the entities are online. No entity needs to be\r
taken offline to avoid such adverse effects. Hence such upgrade operations\r
are referred as online operations. The effects of the upgrade might be activated\r
next time it is used, or may require a special activation action such as a\r
restart. Note that the activation action provides more control and predictability.\r
\r
If an entity's behavior in the system may change due to the upgrade it may\r
be better to take it offline for the time of the relevant upgrade operations.\r
The main question is however considering the hosting relation of an upgrade\r
object what hosted entities are impacted. Accordingly we can identify a scope\r
which is impacted by taking the given upgrade object offline. The entities\r
that are in the scope of impact may need to be taken offline or moved out of\r
this scope i.e. migrated.\r
\r
If the impacted entity is in a different layer managed by another manager\r
this may require coordination because taking out of service some\r
infrastructure resources for the time of their upgrade which support virtual\r
resources used by VNFs that should not experience outages. The hosted VNFs\r
may or may not allow for the hot migration of their VMs. In case of migration\r
the VMs placement policy should be considered.\r
\r
\r
\r
Upgrade duration\r
~~~~~~~~~~~~~~~~\r
\r
As the OPNFV end-users are primarily Telecom operators, the network\r
services provided by the VNFs deployed on the NFVI should meet the\r
requirement of 'Carrier Grade'.::\r
\r
  In telecommunication, a "carrier grade" or"carrier class" refers to a\r
  system, or a hardware or software component that is extremely reliable,\r
  well tested and proven in its capabilities. Carrier grade systems are\r
  tested and engineered to meet or exceed "five nines" high availability\r
  standards, and provide very fast fault recovery through redundancy\r
  (normally less than 50 milliseconds). [from wikipedia.org]\r
\r
"five nines" means working all the time in ONE YEAR except 5'15".\r
\r
::\r
\r
  We have learnt that a well prepared upgrade of OpenStack needs 10\r
  minutes. The major time slot in the outage time is used spent on\r
  synchronizing the database. [from ' Ten minutes OpenStack Upgrade? Done!\r
  ' by Symantec]\r
\r
This 10 minutes of downtime of the OpenStack services however did not impact the\r
users, i.e. the VMs running on the compute nodes. This was the outage of\r
the control plane only. On the other hand with respect to the\r
preparations this was a manually tailored upgrade specific to the\r
particular deployment and the versions of each OpenStack service.\r
\r
The project targets to achieve a more generic methodology, which however\r
requires that the upgrade objects fulfil certain requirements. Since\r
this is only possible on the long run we target first the upgrade\r
of the different VIM services from version to version.\r
\r
**Questions:**\r
\r
1. Can we manage to upgrade OPNFV in only 5 minutes?\r
 \r
.. <MT> The first question is whether we have the same carrier grade\r
   requirement on the control plane as on the user plane. I.e. how\r
   much control plane outage we can/willing to tolerate?\r
   In the above case probably if the database is only half of the size\r
   we can do the upgrade in 5 minutes, but is that good? It also means\r
   that if the database is twice as much then the outage is 20\r
   minutes.\r
   For the user plane we should go for less as with two release yearly\r
   that means 10 minutes outage per year.\r
\r
.. <Malla> 10 minutes outage per year to the users? Plus, if we take\r
   control plane into the consideration, then total outage will be\r
   more than 10 minute in whole network, right?\r
\r
.. <MT> The control plane outage does not have to cause outage to\r
   the users, but it may of course depending on the size of the system\r
   as it's more likely that there's a failure that needs to be handled\r
   by the control plane.\r
\r
2. Is it acceptable for end users ? Such as a planed service\r
   interruption will lasting more than ten minutes for software\r
   upgrade.\r
\r
.. <MT> For user plane, no it's not acceptable in case of\r
   carrier-grade. The 5' 15" downtime should include unplanned and\r
   planned downtimes.\r
   \r
.. <Malla> I go agree with Maria, it is not acceptable.\r
\r
3. Will any VNFs still working well when VIM is down?\r
\r
.. <MT> In case of OpenStack it seems yes. .:)\r
\r
The maximum duration of an upgrade\r
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\r
\r
The duration of an upgrade is related to and proportional with the\r
scale and the complexity of the OPNFV platform as well as the\r
granularity (in function and in space) of the upgrade.\r
\r
.. <Malla> Also, if is a partial upgrade like module upgrade, it depends\r
  also on the OPNFV modules and their tight connection entities as well.\r
\r
.. <MT> Since the maintenance window is shrinking and becoming non-existent\r
  the duration of the upgrade is secondary to the requirement of smooth upgrade.\r
  But probably we want to be able to put a time constraint on each upgrade\r
  during which it must complete otherwise it is considered failed and the system\r
  should be rolled back. I.e. in case of automatic execution it might not be clear\r
  if an upgrade is long or just hanging. The time constraints may be a function\r
  of the size of the system in terms of the upgrade object(s).\r
\r
The maximum duration of a roll back when an upgrade is failed \r
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\r
\r
The duration of a roll back is short than the corresponding upgrade. It\r
depends on the duration of restore the software and configure data from\r
pre-upgrade backup / snapshot.\r
\r
.. <MT> During the upgrade process two types of failure may happen:\r
  In case we can recover from the failure by undoing the upgrade\r
  actions it is possible to roll back the already executed part of the\r
  upgrade in graceful manner introducing no more service outage than\r
  what was introduced during the upgrade. Such a graceful roll back\r
  requires typically the same amount of time as the executed portion of\r
  the upgrade and impose minimal state/data loss.\r
  \r
.. <MT> Requirement: It should be possible to roll back gracefully the\r
  failed upgrade of stateful services of the control plane.\r
  In case we cannot recover from the failure by just undoing the\r
  upgrade actions, we have to restore the upgraded entities from their\r
  backed up state. In other terms the system falls back to an earlier\r
  state, which is typically a faster recovery procedure than graceful\r
  roll back and depending on the statefulness of the entities involved it\r
  may result in significant state/data loss.\r
  \r
.. <MT> Two possible types of failures can happen during an upgrade\r
\r
.. <MT> We can recover from the failure that occurred in the upgrade process:\r
  In this case, a graceful rolling back of the executed part of the\r
  upgrade may be possible which would "undo" the executed part in a\r
  similar fashion. Thus, such a roll back introduces no more service\r
  outage during an upgrade than the executed part introduced. This\r
  process typically requires the same amount of time as the executed\r
  portion of the upgrade and impose minimal state/data loss.\r
\r
.. <MT> We cannot recover from the failure that occurred in the upgrade\r
   process: In this case, the system needs to fall back to an earlier\r
   consistent state by reloading this backed-up state. This is typically\r
   a faster recovery procedure than the graceful roll back, but can cause\r
   state/data loss. The state/data loss usually depends on the\r
   statefulness of the entities whose state is restored from the backup.\r
\r
The maximum duration of a VNF interruption (Service outage)\r
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\r
\r
Since not the entire process of a smooth upgrade will affect the VNFs,\r
the duration of the VNF interruption may be shorter than the duration\r
of the upgrade. In some cases, the VNF running without the control\r
from of the VIM is acceptable.\r
\r
.. <MT> Should require explicitly that the NFVI should be able to\r
  provide its services to the VNFs independent of the control plane?\r
\r
.. <MT> Requirement: The upgrade of the control plane must not cause\r
  interruption of the NFVI services provided to the VNFs.\r
\r
.. <MT> With respect to carrier-grade the yearly service outage of the\r
  VNF should not exceed 5' 15" regardless whether it is planned or\r
  unplanned outage. Considering the HA requirements TL-9000 requires an\r
  end-to-end service recovery time of 15 seconds based on which the ETSI\r
  GS NFV-REL 001 V1.1.1 (2015-01) document defines three service\r
  availability levels (SAL). The proposed example service recovery times\r
  for these levels are:\r
\r
.. <MT> SAL1: 5-6 seconds\r
\r
.. <MT> SAL2: 10-15 seconds\r
\r
.. <MT> SAL3: 20-25 seconds\r
\r
.. <Pva> my comment was actually that the downtime metrics of the\r
  underlying elements, components and services are small fraction of the\r
  total E2E service availability time. No-one on the E2E service path\r
  will get the whole downtime allocation (in this context it includes\r
  upgrade process related outages for the services provided by VIM etc.\r
  elements that are subject to upgrade process).\r
  \r
.. <MT> So what you are saying is that the upgrade of any entity\r
  (component, service) shouldn't cause even this much service\r
  interruption. This was the reason I brought these figures here as well\r
  that they are posing some kind of upper-upper boundary. Ideally the\r
  interruption is in the millisecond range i.e. no more than a\r
  switch-over or a live migration.\r
  \r
.. <MT> Requirement: Any interruption caused to the VNF by the upgrade\r
  of the NFVI should be in the sub-second range.\r
\r
.. <MT]> In the future we also need to consider the upgrade of the NFVI,\r
  i.e. HW, firmware, hypervisors, host OS etc.