LizardFS-WhitePaper-Eng-v3.9.2-web

White paper
LizardFS v4.0 Released

1
Table of Contents
Table of Contents 1
About LizardFS 2
Architecture 3
Possible application of LizardFS 4
Scalability 4
Hardware recommendation 6
Features 7
Snapshots 7
QoS 8
Data replication 8
Replication mechanism 9
Standard replica 9
XOR replica 9
Georeplication (aka custom goals) 10
Metadata replication 11
LTO library support 12
Quotas 12
POSIX compliance 12
Trash 13
Native Windows™ client 13
Monitoring 13
Portability 14
Examples of configuration 14
Notepad 16
Contact 17

2
About LizardFS
LizardFS is a distributed, scalable, fault-tolerant and highly available file system. It
allows users to combine disk space located on many servers into a single name space which
is visible on Unix-like and Windows systems in the same way as other file systems. LizardFS
makes files secure by keeping all the data in many replicas spread over available servers. It
can be used also to build an affordable storage, because it runs without any problems on
commodity hardware.
Disk and server failures are handled transparently without any downtime or loss of data.
If storage requirements grow, it’s possible to scale an existing LizardFS installation just
by adding new servers - at any time, without any downtime. The system will automatically
move some data to newly added servers, because it continuously takes care of balancing
disk usage across all connected nodes. Removing servers is as easy as adding a new one.
Unique features like:
• support for many data centers and media types,
• fast snapshots,
• transparent trash bin,
• QoS mechanisms,
• quotas and a comprehensive set of monitoring tools make it suitable for a
range of enterprise-class applications.
RELIABILITYCOMPATIBILITY
EXTRA
FEATURES
ENTERPRISE
CLASS SOLUTIONS
COMMODITY
HARDWARE
SCALABILITY
CONSTANT
IMPROVEMENTS

3
Architecture
LizardFS keeps metadata (like file names, modification timestamps, directory trees) and
the actual data separately. Metadata is kept on metadata servers, while data is kept on ma-
chines called chunkservers. A typical installation consists of:
• At least two metadata servers, which work in master-slave mode for failure recovery.
Their role is also to manage the whole installation,so the active metadata server is often
called the master server. The role of other metadata servers is just to keep in sync with
the active master servers, so they are often called shadow master servers. Any shadow
master server is ready to take the role of the active master server at any time. A sugge-
sted configuration of a metadata server is a machine with fast CPU,at least 32 GB of RAM
and at least one drive (preferably SSD) to store several dozens of gigabytes of metadata.
• A set of chunkservers which store the data. Each file is divided into blocks called chunks
(each up to 64 MiB) which are stored on the chunkservers. A suggested configuration
of a chunkserver is a machine with large disk space available either in a JBOD or RAID
configuration, depending on requirements. CPU and RAM are not very important. You
can have as little as 2 chunkservers (a minimum to make your data resistant to any disk
failure) or as many as hundreds of them.A typical chunkserver is equipped with 8, 12, 16,
or even more hard drives.
• Clients which use the data stored on LizardFS. These machines use LizardFS mount to
access files in the installation and process them just as those on their local hard drives.
Files stored on LizardFS can be seen and simultaneously accessed by as many clients as
needed.
Figure 1: Architecture of LizardFS

4
Possible application of LizardFS
• There are many possible applications of LizardFS
• archive - with usage LTO tapes,
• storage for virtual machines (as an openstack backend or similar)
• storage for media / cctv etc
• storage for backups
• as a storage for Windows™ machine
• as a DRC (Disaster Recovery Center)
• HPC
Scalability
If storage requirements grow, an existing LizardFS installation can be scaled just by ad-
ding new chunkservers. Adding a new server is possible without any downtime and can be
performed at any time, i.e., it is transparent to clients. There are no restrictions about disk
space of newly added chunkservers (e.g., adding a 24 TiB chunkserver to an installation con-
sisting of 8 TiB chunkservers will work without any problems).
Performance of the system scales linearly with the number of disks in the system, so
adding a new chunkserver will not only increase available capacity, but also the overall
performance of the storage. The system will automatically move some data to newly added
chunkservers, because it balances disk usage across all connected nodes. Removing servers
(or taking them for longer maintenance) is as easy as adding one.
75% 75% 75% 75% 75% 75%
Example
Consider the following installation:
server total space used space
1 20 TiB 15 TiB (75%)
2 16 TiB 12 TiB (75%)
3 16 TiB 12 TiB (75%)
4 16 TiB 12 TiB (75%)
5 8 TiB 6 TiB (75%)
6 8 TiB 6 TiB (75%)
TOTAL 84 TiB 63 TiB (75%)

5
1 20 TiB 15 TiB (75%)
2 16 TiB 12 TiB (75%)
3 16 TiB 12 TiB (75%)
4 16 TiB 12 TiB (75%)
5 8 TiB 6 TiB (75%)
6 8 TiB 6 TiB (75%)
7 21 TiB 0 TiB (0%)
Just after adding a new server the disk usage statistics will look like:
75% 75% 75% 75% 75% 75% +
1 20 TiB 12 TiB (60%)
2 16 TiB 9.6 TiB (60%)
3 16 TiB 9.6 TiB (60%)
4 16 TiB 9.6 TiB (60%)
5 8 TiB 4.8 TiB (60%)
6 8 TiB 4.8 TiB (60%)
7 21 TiB 12.6 TiB (60%)
Soon (administrators can decide how fast this process should be) the data will be
rebalanced so that the disk usage is almost even, e.g.:
60% 60% 60% 60%60% 60% 60%

6
Should the system administrator decide to take the 7th server out, he can mark it as a server
which is going to be removed–LizardFS will automatically copy all the data from this server
to other servers. In this example the system will be left in a state like this:
1 20 TiB 15 TiB (75%)
2 16 TiB 12 TiB (75%)
3 16 TiB 12 TiB (75%)
4 16 TiB 6 TiB (75%)
5 8 TiB 6 TiB (75%)
6 8 TiB 6 TiB (75%)
7 21 TiB 12.6 TiB (60%)
TOTAL 105 TiB 75.6 TiB (72%)
75% 75% 75% 75% 75%75% 60%%6
Hardware recommendation
LizardFS will be working on any hardware, you can use commodity hardware as well.
Minimum requirements is two dedicated nodes with bunch of disks, but to obtain proper HA
installation, you should have 3 nodes.
We recommend that each node has at least two 1Gbps network interface controllers
(NICs). Since most commodity hard disk drives have a throughput of approximately 100MB/
second, your NICs should be able to handle the traffic for the chunkservers disks on your
host.
Minimal configuration of LizardFS strongly depends on its use case. LizardFS will run on
practically any machine of reasonable configuration,but sample infrastructure for a medium
size installation could be as follows:

7
• Master / Shadow - at least 2 GHz CPU, 64bit
• RAM - depends on expected number of files (4GB should be enough for a small
installation)
• Disk - 128G, SSD would improve performance, HDD is fine
• Chunkserver - recommended 2GB RAM
• Metalogger - recommended 2GB RAM (or more)
Features
What makes LizardFS mature, enteprise solution are additional features developed on
a constantly improving core. They can transform probably the best distributed file system
in the world into Hierarchical Storage Management (HSM), help to build Disaster Recovery
Center with asynchronous replication between sites, reduce disk space required for replica-
tion, effectively manage storage pools (QoS, Quotas) and many more. If you see any other
use case for LizardFS that would require any other functionality, do let us know, we may put
it into our Road Map or develop it especially for you.
Snapshots
Copying large files and directories (eg. virtual machines) can be done efficiently with
lazy snapshot feature. When creating a snapshot, only metadata of a target file is copied,
which makes the operation really quick. Chunks of the original file and the duplicated one
are now shared until one of them is modified.
Example
Consider a directory /mnt/lizardfs/huge_vm, which contains several large files, a total of
512GB. Making a backup copy with `cp -r huge_vm huge_vm_backup` command will take a
long time and result in consuming additional 512GB of disk space.
Creating a copy with `mfsmakesnapshot huge_vm huge_vm_backup` command will, in
turn, last a negligible amount of time, and will consume no disk space whatsoever.
Now, consider modifying one file in /mnt/lizardfs/huge_vm directory, which would affect
2 of its chunks. In such case, only those chunks would be copied, consuming extra 128MB of
space, which is still negligible compared to 512GB consumed by `cp -r` variant.

8
Data replication
Files stored in LizardFS are divided into blocks, each up to 64 MiB, called chunks. Each
chunk is kept on chunkservers and administrators can choose how many copies of each
chunk the system should maintain. For example, if we keep 3 copies (we call this configu-
ration goal=3), all the data will survive a failure of any two disks or chunkservers, because
LizardFS will never keep 2 copies of the same chunk on the same server. Security levels
(number of copies) can be set independently for each file and directory tree.
If any chunk is lost due to a disk failure the system will replicate any existing copies of
such a chunk to achieve the requested number of copies.
Data is spread evenly across all the nodes. There are no servers dedicated to be mirrors
of other servers.This means that (in case of goal=3) each chunk has its three copies on three
randomly chosen servers. This unique design makes failure recovery very fast. If one disk is
lost,backup copies of its data are evenly spread over all other disks,so all disks in the instal-
lation can take part in the process of replication.The speed of recovering lost data increases
with the number of disks in the system.
QoS
LizardFS offers mechanisms which allow system administrators to set read/write ban-
dwidth limits for all the traffic generated by a given mountpoint as well as for a specific
group of processes spread over many client machines and mountpoints.This makes LizardFS
usable in situations where some restrictions are required to provide a proper quality of se-
rvice for some critical applications.
Example
Consider an installation build on top of 300 hard drives. One can expect it to serve abo-
ut 20 GB of data per second (depending on hard drive models used to build it and access
patterns) to all the connected clients (in total). There is a production system consisting of
many virtual machines using LizardFS as their storage. If a group of system administrators
has access to host machines and use it for maintenance works, there is a possibility to use
bandwidth limits to make the production system safer. After configuring host machines to
place all ssh processes (and all other spawned by them) in a cgroup named, for example,
“interactive” and defining a bandwidth limit of 100 MB/s for the “interactive” group, system
administrators won’t be able to unwittingly disrupt the production system. All their proces-
ses started manually from ssh sessions will be limited to 100 MB/s in total,even though they
might be spread over tens of machines. In the same installation there is also a possibility
to add a bandwidth limit for a set of virtual machines used by a subsystem (or a customer)
with limited performance requirements to make sure that these machines won’t disrupt
operation of other virtual machines. After setting this limit to 1 GB/s administrators can be
sure that all the traffic generated by these machines (both writes and reads) will sum up to
one gigabyte per second or less.

9
1
11
2 3 4 5 6
32
4 5
2
4 5
6
3
54
1
32
66
Replication mechanism
LizardFS implements two replication mechanisms: standard and XOR copies.
Standard replica
This kind of replication allows to set a number copies every chunk of a file should have.
Note that it is not possible to store 2 same chunks on 1 chunkserver, so standard goals sho-
uld comply with the number of chunkservers and their labels.
Example
Standard goal “3” indicates that each chunk should be placed on 3 different chunkse-
rvers. Specific location of those chunks will be determined by master server, which will try
to balance space usage evenly.
XOR replica
This kind of replication divides chunks into smaller parts according to “xor level” para-
meter. A file with xor level N is kept as N+1 chunks (N for data, 1 for parity check).
As a result, disk space needed for keeping this file is roughly (N+1)/N and it is still possi-
ble to survive failure of one chunkserver. Xor level can range from 2 to 10.

10
Example
Goal“xor3”means that each chunk of a file will be divided into 4 parts and distributed to
chunkservers. Now, if any single chunkservers fails, it is possible to reconstruct the file from
the remaining 3 xor chunks.
Georeplication (aka custom goals)
The feature called custom goals makes it possible to decide where chunks should be
placed and the topology feature allows to suggest which copy should be read by a given
client when there is more than one copy available. This makes LizardFS more useful when
working with many locations and media types (e.g., SSD and HDD drives).Administrators are
able to label chunkservers and to decide how the system spreads chunks across available
servers based on those labels.This allows for LizardFS to withstand a failure of a whole data
center (Disaster Recovery), see the example below.
Example
If there are two data centers, e.g. one in London and one in Paris, there is a possibility to
assign a label “london”to each server in the first location and “paris”to each server in the se-
cond location as well as to create a custom goal named 2_locations defined as “2_locations:
paris london”. All files created in a directory with such a goal will have their chunks kept in
two copies, one of them on a server labelled london and one of them on a server labelled
paris.This will ensure that even if a whole location is down, all the files in the working loca-
tion are still available for reading and writing.One can use the topology mechanism to make
mounts prefer reading Custom goals (like described in this example) can be used instead
of simple goals which specify only a number of copies (e.g., goal=3). There is a possibility
to set and change them at any time for any file or any directory tree and the system will
accordingly move data if required.
There are more possibilities, e.g.:
• To define a goal “safe_2_locations: paris london _“ which is interpreted as: “3 copies, at
least one in London and one in Paris”–now we can lose any two servers or a whole data
center without any downtime. The system will place the third copy in a way that evens
out disk usage on all servers to a point defined by the system administrator..

11
• To define a goal “fast: ssd ssd”– if some servers are equipped only with ssd drives and
others with hdd drives (and properly labelled), this ensures that such files will be kept
in two copies only on chunkservers labelled ssd, in most cases much faster than others
• If there is only one location with a configuration from the previous example (SSD and
HDD servers) one can use a custom goal like “one_fast_copy: ssd hdd” for some files to
force access from a specified medium.
Metadata replication
Metadata in LizardFS consists of:
• names of files and directories
• POSIX attributes of files and directories, such as owner, owning group, access mask,
last modification and access timestamps, size of files, etc.
• extended attributes
• access control lists
• information about each file’s goal and trash time
• information about which chunks form each file
Metadata is stored on metadata servers. At any time, one of the metadata servers also
manages the whole installation and is often called the master server. Other metadata se-
rvers keep in sync with it and are called shadow master servers. Each running metadata
server stores all the metadata in its RAM as well as on a local hard drive. All the changes
in the metadata made by the master server are almost instantly applied by all connected
shadow master servers and also saved on their hard drives. At any time, it is possible to add
a new shadow master server or to remove an existing one.These operations are completely
transparent to the clients.
Shadow master servers provide LizardFS with high availability – failure of the master
server is automatically fixed within a couple of seconds. If there is at least one shadow ma-
ster server running and the active master server is lost, one of the shadow master servers
takes over its role immediately. All other shadow master servers, chunkservers, clients and
metaloggers will continue to work with the new master server. This process can also be
triggered manually when there is need for some maintenance works on the master server.
Moreover, you can increase security of your metadata by adding metaloggers. Each me-
talogger is a lightweight (almost no RAM nor CPU requirements) daemon which connects
to the active metadata server, continuously receives the same information that all shadow
masters receive and continuously saves it to its local hard drive. In this way an additional
copy of metadata is created and if all the metadata servers are lost (which is extremely rare,
but still possible), the whole system can be easily recovered from such a cold copy. Because
metaloggers use almost no CPU nor RAM, the cost of adding many of them is very low. In
this way one can have any number of independent and constantly updated copies of all the
metadata.

12
Example
For a file system which consists of 50 million files which sum up to 1 PiB of data, the
metadata will occupy about 13 GiB of hard drive space. A single metalogger which keeps
a couple of copies of the metadata (the most recent version as well as a couple of older
backups) will require less than 100 GiB of hard drive space. After setting up metaloggers
on 10 machines which work as chunkservers there will be 12 constantly updated copies of
the system’s metadata (including one copy kept on the active master server and one on a
shadow master server) making it virtually indestructible.
LTO library support
LizardFS offers native support for LTO libraries. Storing archival backups may consume
a lot of memory, even though such files are almost never read. This kind of data can be ef-
ficiently stored on a tape, so LizardFS offers a simple way to cooperate with back-end LTO
storage. Files can be chosen to have a backup copy on a tape by setting a tape goal.
Setting tape goal to a file makes it read-only for obvious reasons - tape storage does
not support random writes. Reading from tape storage is a timely process (may last 48h or
require manual work to insert correct tape to library), so data stored in there should be ar-
chival - meant to be read very rarely.
Method of reading a file stored on tape depends on its situation:
• If a regular copy of a file is still available, it will be used for reading
• If a file exists only on tape,it has to be restored to LizardFS first.To achieve that,one
must use lizardfs-restore-tape-copy utility: $ lizardfs-restore-tape-copy file_path
After running this command, all needed data will be read from tape storage and loaded
to file system, making the file accessible to clients.
Quotas
LizardFS support disk quota mechanism known from other POSIX file systems.It offers a
possibility to set soft and hard limits for number of files and their total size for a specific user
or a group of users. User whose hard limit is exceeded cannot write new data to LizardFS.
POSIX compliance
LizardFS offers all the features of POSIX file systems, for example:
• hierarchical structure of files and directories
• attributes of files like uid, gid, access mask, access time, modification time, change time
• support for symlinks, hard links, and other special files (unix sockets, devices)
• access control lists
• extended attributes
Semantics of access rights in LizardFS are precisely the same as in other POSIX file
systems (like ext or xfs). This makes it compatible with all applications that can use a local
hard drive as their storage.

13
Trash
Another feature of LizardFS is a transparent and fully automatic
trash bin.After removing any file,it is moved to a special place,called
trash,which is visible only to the system administrator.Any file which
is in the trash can be either undeleted or removed permanently.This
makes data stored on LizardFS even more secure than data stored
on a hardware RAID.The trash is automatically cleaned up: each file
is removed permanently from the trash by the system after some
period of time (called trash time) and at any moment, this period
can be adjusted separately for any file or directory tree in the system.
Example
If trash time for the whole installation is set to two hours, every file will be available for
undeleting within two hours of removing it. This should be enough for accidents caused by
careless users. Additionally trash time can be adjusted for any file or any directory tree, for
example:
• set to 0 for a directory with temporary files, so they would disappear immediately
after removing and free up space for new files
• set to 1 month for very important data
• set to 0 for some file before removing it when we need to immediately recover
some free space and we know that this file is no longer needed
Native Windows™ client
There is a native Windows Client available for LizardFS. This allows users to build a so-
ftware solution free from single point of failures like SMB servers which are typically used
to export Unix file systems to Windows applications. LizardFS Windows Client has a friendly
graphical user interface and is suitable for both desktop and server use. It provides access
to files stored on LizardFS via a virtual drive, for example L:, which makes it compatible with
almost all existing Windows™ applications.
Monitoring
LizardFS offers two kinds of monitoring interfaces. There is a command line tool useful
for automated monitoring systems like Nagios,Zabbix,Icinga,and similar which are typically
used for proactive monitoring. It offers a comprehensive set of information about connected
servers and their disks, client machines, storage space, errors, chunk copies lost due to disk
failures, and much more. Moreover, there is a graphical web-based monitoring interface
available for system administrators. It makes it easy to track almost all aspects of a running
system, including tens of charts which show how the way in which LizardFS installation is
used changes over time.

14
Portability
LizardFS packages are available for most operating systems used in production, inclu-
ding currently supported releases of:
• Ubuntu LTS
• Debian
• Red Hat Enterprise Linux
• CentOS
For these operating systems installation is very easy and is performed by the package
manager which is a part of these Linux distributions. The software is highly portable and
besides of the official package repositories, there is a possibility to use a tarball to build
LizardFS for almost any operating system from the Unix family.
Additionally there is a client driver available for Windows machines and it can be instal-
led using a simple installation wizard.
Examples of configuration
Example 1
• 2 metadata servers, each equipped with 64 GB of RAM and a RAID-1 storage build
on top of two 128 GB SSD drives
• 20 chunkservers, each equipped with a RAID-6 controller with 8 SAS drives,
3 TiB each
• 10 Gbit network
• replication level 2
• a metalogger running on 3 out of 20 chunkservers
Such a configuration offers a possibility to store 180 TiB of data in goal=2. It offers a
high performance as well as a high durability. There will be no downtime in case of:
• losing any 5 disks
• losing any single server (including the master server)
No data will be lost in case of:
• losing any single chunkserver
• losing both metadata servers

15
Example 2
• 2 metadata servers, each equipped with 64 GB of RAM and a RAID-1 storage build
on top of two 128 GB SSD drives
• 30 chunkservers, each equipped with 12 hard drives (in JBOD configuration),
3 TiB each
• 1 Gbit network
• replication level 3
• a metalogger running on 3 out of 30 chunkservers
This configuration offers a possibility to store 360 TiB of data in goal=3. There will be
no downtime in case of:
• losing any 2 chunkservers or any 2 disks
• losing any single metadata server
No data will be lost in case of:
• losing any 2 chunkservers or any 2 disks
• losing both metadata servers
Example 3
• 2 data centers, primary and secondary
• 3 metadata servers, two in the primary data center one in the backup data center,
each equipped with 64 GB of RAM and a RAID-1 storage build on top of two 128
GB SSD drives
• 15 chunkservers in each of 2 data centers, each equipped with a RAID-6 controller
with 16 SAS
• drives, 3 TiB each
• 10 Gbit link between data centers
• replication level which enforces one copy to be kept in each data center
• 2 metaloggers (on selected chunkserver machines) in each data center
This configuration offers a possibility to store 630 TiB of data. There will be no downti-
me and no data will be lost in case of:
• losing any single server (including the master server)
• losing the connection between data centers
• losing all servers in the secondary data center (e.g., as a result of a power failure,
major fire or a flooding)
• losing all chunkservers in the primary data center
No data will be lost in case of losing all servers in the primary location, but there would
be a need to start the master server manually (no quorum in the secondary data center pre-
vents this to happen automatically), thus some downtime may occur.

16
Notepad
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

LizardFS Contact:
info@lizardfs.com
Poland tel. +48 22 100 32 48
USA tel. +1 (646) 655 0693
visit our website:
www.lizardfs.com
Contact
Company contact:
Skytechnology sp. z o.o.
ul. Głogowa 21
02-639 Warsaw
NIP: 5213633370
KRS: 0000426381
tel. +48 22 100 32 48
fax. +48 22 250 29 13
info@skytechnology.pl
visit our website:
www.skytechnology.pl

LizardFS-WhitePaper-Eng-v3.9.2-web

More Related Content

What's hot (20)

Viewers also liked (6)

Similar to LizardFS-WhitePaper-Eng-v3.9.2-web (20)

LizardFS-WhitePaper-Eng-v3.9.2-web