Virtualization of EPOCH IPS
April, 2016 Tatsuya Watanabe / Assistant Manager System Engineering Team SKY Perfect JSAT Corporation (SJC)
0
Agenda
1. Overview of SJC and YSCC
2. Current status of virtualization in SJC
3. Function of VMware and how SJC uses it
4. Points of concern about virtualization
Copyright © 2016 Sky Perfect JSAT Corporation. All Rights Reserved 1
** JSAT owns 5 Ku-band transponders on IS-15 Horizons-1*
(127W) GEO
JCSAT-4A (82E)
JCSAT-1B (150E)
JCSAT-2A (154E)
JCSAT-RA (Backup)
N-SAT-110 (Backup)
JCSAT-5A (132E)
JCSAT-3A (128E)
Horizons-2* (84.8E)
36,000km
Superbird-B2 (162E)
JCSAT-85** (Intelsat-15)
(85E)
Superbird-A3 (158E)
N-STAR c (136E)
Superbird-C2 (144E)
* Horizons-1 & 2 are Jointly owned by JSAT and Intelsat.
JCSAT-110R*** (110E)
JCSAT-4B (124E) 3rd Party Customer
Satellite(143E) *** JCSAT-110R is Jointly owned by JSAT and B-SAT.
3rd Party Customer Satellite
JCSAT-14
JCSAT-16
JCSAT-15
Superbird-B3
Satellite Fleet
3 Copyright © 2016 Sky Perfect JSAT Corporation. All Rights Reserved
As of 2014/12
’89 ’90 ’91 ’92 ’93 ’94 ’95 ’96 ’97 ’98 ’99 ’00 ’01 ’02 ’03 ’04 ’05 ’06 ’07 ’08 ’09 ’10 ’11 ’12 ’13 ’14
J C S A T -1
J C S A T -2
S u p e rb ird -B
S u p e rb ird -A
J C S A T -3
J C S A T -R
S u p e rb ird -A 3
J C S A T -1 B
J C S A T -4 A
S u p e rb ird -B 2
N -S T A R a
N -S T A R b
N -S A T -1 1 0
J C S A T -2 A
N -S T A R c
H o riz o n s-1
3 rd P a rty S a t-A
J C S A T -5 A
J C S A T -3 A
H o riz o n s-2
S u p e rb ird -C 2
J C S A T -R A
J C S A T -8 5
J C S A T -1 1 0 R
J C S A T -4 B
3 rd P a rty S a t-B
Lockheed Martin 社製
Boeing 社製
Orbital社製
Boeing 社製
Lockheed Martin 社製
D e -o rb it
D e -o rb it
D e -o rb it
D e -o rb it
D e -o rb it
D e -o rb it
Space System Loral 社製
三菱電機製
D e -o rb it
D e -o rb it
Lockheed Martin 社製
Orbital 社製
D e -o rb it
Boeing社製
Lockheed Martin 社製
Space Systems Loral 社製
Orbital 社製
三菱電機社製
NEC 社製
B o e in g
B o e in g
B o e in g
B o e in g
B o e in g
B o e in g
N T T S S L
N T T S S L
L M
B o e in g
O rb ita l
B o e in g
L M
S S L
S S L
B o e in g
B o e in g
S S L
M e lco
L M
L M
O r b ita l
O r b ita l
N T T
N E C
L M
Boeing
Lockheed Martin
Space Systems Loral
Orbital
Melco
NECL M
Satellite Operations Experience
4 Copyright © 2016 Sky Perfect JSAT Corporation. All Rights Reserved
Yokohama Satellite Control Center (YSCC)
Primary Satellite Control Center and Teleport Facilities
Gunma Satellite Control Station (GSCS)
Sub Ground Station
Ibaraki Network Control Center (SPE)
Backup Satellite Control Center and Teleport
Yamaguchi Network Control Center(SPW)
Sub Ground Station
SJC Ground Stations
5 Copyright © 2016 Sky Perfect JSAT Corporation. All Rights Reserved
Satellite Planning & Operations Group
Satellite Operations Division
Noriko Masuda
Satellite Control Team
Yoji Sato
System Engineering Team
Hiroaki Nagai
Satellite Engineering Team
Azuma Miyata
Orbital Dynamics Team
Hideshi Chazono
Service Operations Division
Hideo Hayashi
Network Operations Team
Teleport Operations Team-1
Teleport Operations Team-2
Satelite Facilities Management &
Administration Div.
Tetsuya Segawa
Team-1
Team-2
Team-3
Ground Systems Procurement & Maintenance
General affairs & administrations in YSCC
General affairs & administrations in SPE
YSCC Organization
6 Copyright © 2016 Sky Perfect JSAT Corporation. All Rights Reserved
Reasons to use virtualization
SJC will launch 4 satellites within 2 years. work on reducing the operational load in several ways One of the way is virtualization of EPOCH IPS
Virtualization refers to the creation of a virtual machine that acts like a real computer with an operating system(#).
8 Copyright © 2016 Sky Perfect JSAT Corporation. All Rights Reserved
Traditional Architecture Virtual Architecture
(#)Wikipedia: https://en.wikipedia.org/wiki/Virtualization
SJC decided to use VMware as a virtualization software for EPOCH IPS.
VMware has a lot of merits as follows: Cost effective
Resource Optimization, Saving Rack space, Consumed Power Functions to Reduce Operational Load
vMotion, Snapshot, cloning, HA, Backup, Converter, and so on.
Last year, SJC verified following points: Essential functions of VMware work as expected EPOCH IPS on Virtual Server is working just like on a Physical Server
Reasons to use virtualization
9
Bus
Type
Boeing
601 (5 satellites)
A2100
(5 satellites)
StarBus2 (1 satellites)
DS2000 (1 satellites)
SMACS (1 satellites)
New
Satellites (4 satellites)
Operations
software TCR/STA
EPOCH v4
OASYS
EPOCH v4
OASYS
BirdStar
ORAMS SMACS
EPOCH v4
OASYS
Coming Soon
Plan to change to EPOCH v4 & OASYS
The relation between virtualization and increase of EPOCH IPS
EPOCH IPS constitutes an increasingly large portion of the SJC. The more the percentage of EPOCH IPS is increasing, the more we’ll see
the effect of virtualization of EPOCH IPS go up.
10 Copyright © 2016 Sky Perfect JSAT Corporation. All Rights Reserved
Consider the way to change to EPOCH IPS without extra cost (because these are old so it’s difficult to secure a budget for now)
Bus Type / Operations Software
SJC has virtualized only part of EPOCH IPS servers.
SJC will continue to virtualize other servers and to investigate the appropriate method to virtualize workstation.
Servers for existing satellites in YSCC
Servers for upcoming satellites in YSCC
Servers in Sub-site
Workstations in YSCC
Workstations in Sub-site
Already virtualized
Virtualization of workstation is under investigation.
11 Copyright © 2016 Sky Perfect JSAT Corporation. All Rights Reserved
Current situation and virtualization plan in the future
Percentage of EPOCH IPS servers and workstations in each site
Physical Configuration Overview
VMware is Installed on both of 2 dell servers Virtual Machines (VM) are saved to a Shared Disk. One of 2 VMware ESXis loads and executes VM onto its memory
Dell Shared Storage for saving VMs
Dell L2SW
VMware ESXi VMware ESXi
Dell Server
EPOCH SV REDHAT
Virtual Machines (VM)
Copyright © 2016 Sky Perfect JSAT Corporation. All Rights Reserved 13
EPOCH SV REDHAT
NAS for backup VMs
Compass Win Server
Compass Win Server
1
2
3
4
5
6
7
8
9
10
2 Dell servers equipped with 2 (6 core,HT) CPUs and 192GB Memory each
2 Dell Layer-2 Switches (back)
NAS equipped with 4 7,200 RPM SATA HDDs
Dell Shared Storage equipped with 24 15,000 RPM SAS HDDs
Saving Rack Space of Physical Configuration
Lots of VM are consolidated and working on these machines. Hardware failure has a significant impact on VM, so it’s necessary to
make each Machine as a redundant configuration.
Copyright © 2016 Sky Perfect JSAT Corporation. All Rights Reserved 14
Copyright © 2016 Sky Perfect JSAT Corporation. All Rights Reserved
Basic License Type of VMware
VMware ESXi
VMware ESXi
VMware ESXi
So far, SJC is using Essentials Plus License of VMware that is reasonable but has many functional restrictions.
This License enables us to: Use basic functions such as vMotion, Snapshot, clone, HA, and VDP
This License doesn’t enable us to: Scale-out more than 3 servers Use advanced functions such as FT, DRS, Storage vMotion, App-HA,
and so on
VMware ESXi
15
vMotion enables running VM to move into another server with zero downtime
SJC confirmed that EPOCH server can finish its vMotion process without
disconnecting from DSS.
vMotion
①Stop ①Activate
DSS
Telemetry
EPOCH SV REDHAT
EPOCH SV REDHAT
VMware ESXi VMware ESXi
Copyright © 2016 Sky Perfect JSAT Corporation. All Rights Reserved 16
Copyright © 2016 Sky Perfect JSAT Corporation. All Rights Reserved 17
④Back to the previous state
Snapshot
Snapshot preserves the state of the VM at a specific point in time.
When you revert to a snapshot, you discard all changes made to the VM since you took the snapshot.
①Take a snapshot
③become Unstable state
⑤return to Stable state
②Apply Patch
Stable state
Clone
Clone is a copy of an existing virtual machine.
If you want to make a copy of a virtual machine for separate use, create a clone.
If you want to save the current state of the virtual machine, so you can revert to that state in case you make a mistake, take a snapshot.
Copyright © 2016 Sky Perfect JSAT Corporation. All Rights Reserved
EPOCH SV
REDHAT
EPOCH SV
REDHAT
VMware ESXi
①Cloning
18
Clone
Copyright © 2016 Sky Perfect JSAT Corporation. All Rights Reserved 19
VMware ESXi #2 VMware ESXi #1
1. Virtual Machine Failover
②Reboot VM on the same host
2. Host Machine Failover
VMware ESXi #2 VMware ESXi #1
③VMs start
monitor
VMware HA
Failover function when Physical or Virtual Machine fails.
①failure
②Stop all powered-on VMs
monitor
VM VM
VM ①failure
VM VM
Copyright © 2016 Sky Perfect JSAT Corporation. All Rights Reserved
VMware Data Protection (VDP)
Copyright © 2016 Sky Perfect JSAT Corporation. All Rights Reserved 20
VDP is an agent-less, image-level VM backup tool
Saving a lot of backups causes unstable state of VDP, so SJC set up
backup schedule of VDP as following Once a day for 7 days, Once a week for 4 weeks, Once a month for 12 months
Dell Shared Storage
Dell L2SW
Dell Server
VMware ESXi VMware ESXi
Backup
VDP
EPOCH SV
REDHAT
Copyright © 2016 Sky Perfect JSAT Corporation. All Rights Reserved 21
vCenter Converter
It’s hard to rebuilt already running EPOCH IPS servers as a VM.
vCenter Converter enables us to convert EPOCH IPS into VM easily.
SJC conducted load tests of the converted EPOCH server on VM, and confirmed it works fine.
SJC has already been operating the converted EPOCH Server as a VM.
VMware ESXi 5.5
EPOCH SV
REDHAT
Control Control
Target Physical Servers
Destination
vCenter Converter
EPOCH SV
REDHAT
EPOCH CL
Windows7 Copy
EPOCH CL
Windows7
Other features of VMs
VM is just a file, so it’s easy to move VM between a production environment and a test environment.
EPOCH SV
REDHAT
EPOCH SV
REDHAT
Test environment
VMware Player
EPOCH SV
REDHAT
VMware Player
VMware ESXi (Production environment)
copy
EPOCH SV
REDHAT
copy
Copyright © 2016 Sky Perfect JSAT Corporation. All Rights Reserved 22
EPOCH SV
REDHAT Cloning
24
Points of resource allocation
It’s necessary to decide resource allocation size of each VM.
Resources are such as number of CPU cores, Memory size, and HDD size
If the resource allocation size is too much, and the total of all the VM’s
actual resource consumption exceeds physical server’s resource size, all
unprotected VMs become unstable.
On the other hand, if the resource allocation size is too little, it affects the
performance.
To prevent these situations, it’s necessary for important VMs such as
EPOCH server to investigate and to reserve (a function to assure allocated
resource) the minimum requirement of resource size.
Copyright © 2016 Sky Perfect JSAT Corporation. All Rights Reserved 25
Resource allocation size of each EPOCH IPS server in SJC
SJC conducted load tests and observed the resource utilization of each
EPOCH IPS server.
These specs are smaller than on Physical servers, yet they’re working fine.
The appropriate size depends on the use case scenario.
EPOCH server for 1 telemetry
Task Initiator
Compass Server
CPU core 2~ 2~ 2~
Memory 8GB~ 4GB~ 4GB~
HDD 100GB~ 30GB~ 30GB~
Specific resource allocation size of each EPOCH IPS server in SJC (#)
(#) The memory and HDD sizes will greatly differ according to whether or not we use the memory over commit and the thin provisioning functions
(SJC tries not to use these functions)
Copyright © 2016 Sky Perfect JSAT Corporation. All Rights Reserved 26
Failure detecting and resource monitoring
Hardware failure and resource shortage on the Physical Machine have a
significant impact on all virtual machines.
SJC built an NMS server to detect Hardware, VMware Hypervisor,
vCenter process failure and to monitor resource utilization.
SNMP Polling & VMware API
SNMP Polling & SNMP Trap
SNMP Trap
NMS Network Management System
iDrac
VMware ESXi
OS
vCenter
Dell Shared Storage
Dell L2SW
Dell Server
Copyright © 2016 Sky Perfect JSAT Corporation. All Rights Reserved 28
There was a lot of hard work involved in our first introduction of
virtualization of EPOCH IPS as follows
to consider hardware specs, redundancy design, and migration plan
to prepare the way of failure detecting and resource monitoring (#)
However, VMware functions provide the benefit of increasing the
efficiency of the system operation.
Advantages and Disadvantages of VMware
(#) Preparing fault isolation methods is indispensable. If something happened to a VM, He that is unfamiliar with virtualization may doubt VMware.
Copyright © 2016 Sky Perfect JSAT Corporation. All Rights Reserved
Thank you. Please feel free to contact me;
29
Copyright © 2016 Sky Perfect JSAT Corporation. All Rights Reserved
SJC is working on virtualizes only servers in main site for now, however it’s necessary to consider multi-site design. For example,
SJC needs sufficient number of client only in case of emergency in other site.
There is some possibility to make unused licenses that can be used for a purpose of replace TCR/STA to EPOCH IPS.
Main Site Other Site (Sub-site)
Server
REDHAT
Total optimization in the next stage
Most part of the clients are deactivate in normal time and activate in case of emergency
Client
Windows
Server
REDHAT
Client
Windows
Copy VM by using VMware or Storage replication
31
Copyright © 2016 Sky Perfect JSAT Corporation. All Rights Reserved 32
Ideal design of HA configuration
It may be possible to monitor the EPOCH server on the VM with higher
class license and custom settings of VMware.
This way is smart because it’s not necessary to prepare EPOCH Server on
the each VMware ESXi, but we have to consider the failover downtime,
reconnection with client / baseband, and get over some hurdles as
mentioned above.
Redhat
VMware ESXi #1
monitor
EPOCH Server Processes
monitor
①failure
②reboot Application
VMware ESXi #2 monitor
Copyright © 2016 Sky Perfect JSAT Corporation. All Rights Reserved 33
VMware HA in SJC
Redhat
VMware ESXi #2 VMware ESXi #1
monitor monitor
EPOCH Server Process
Not monitor
If the application's process finishes, we want VMware to reboot the
application to recover. However, due to the licensing constraints and
other problems, this is not realized for now.
SJC prepares EPOCH server on the each VMware ESXi for redundancy.
Telemetry
EPOCH Server Process
Redhat
monitor
Not monitor
BB
Problem that we have to solve
Copyright © 2016 Sky Perfect JSAT Corporation. All Rights Reserved
Redhat
VMware ESXi #2 VMware ESXi #1
monitor monitor
EPOCH Server Process
Not monitor
EPOCH Server Process
Redhat
monitor
Not monitor
NMS
monitor
SJC needs the way to confirm whether or not EPOCH IPS server process is
alive without monitoring GUI. As one of the methods, we can use “ps aux
| grep epoch”, but it’s an uncertain way.
Does anyone know any monitoring functions of the EPOCH server like the
“statusEpoch” command?
34