© Copyright IBM Corporation, 2016

ownCloud Enterprise Edition on IBM

Elastic Storage Server

A performance and sizing study for large user number scenarios

A technical report

Chris Burnard

ownCloud - Senior Solutions Architect

March 2016

Intended audience ................................................................................................................................... 2

IBM Elastic Storage Server ...................................................................................................................... 2

ownCloud Enterprise Edition ................................................................................................................... 3

Proof of concept setup ............................................................................................................................. 5

Solution performance testing methods .................................................................................................... 7

Performance optimizations .................................................................................................................... 10

Apache tuning .................................................................................................................. 11

Large enterprise user account number systems ................................................................................... 13

1

them to access their files seamlessly on all devices and collaborate with colleagues, customers, and

business partners.

Enterprise IT, on the other hand, is concerned about the lack of control over corporate data, and the

security and compliance risk it presents. Recent breaches of public consumer cloud services have proven

those concerns to be justified, and the best course of action is to create a secure, easy-to-use alternative.

Together, ownCloud and IBM® are helping enterprises across industries overcome file sharing challenges,

enabling agile, scalable and highly secure file sync and share with ownCloud Enterprise Edition and IBM

deliver convenient end-user file access and ongoing IT data control. Backed by IBM storage technologies,

such as IBM Spectrum Scale™ and IBM Elastic Storage™ Server, the solution can handle an

compute infrastructure using high availability (HA) configuration with emphasis on large user number

scenarios.

Specifically, the performance of ownCloud Enterprise Edition 8.2.1 was measured running on high-end

x86 based nodes as database and application server infrastructure connected to IBM Elastic Storage

Server (ESS-GL6) system as the storage back end. The lab proof of concept solution is configured with

load balancing and maximum possible amount of authentication setup combinations. The proof of concept

was built for measuring the performance capabilities and sizing assumptions of a large number of user

scenarios of the ownCloud application and database server and the storage back end.

An important factor considered by the test team for specific system sizing is: What is the number of

intended users and their usage requirement with the system? The sizing consideration is directly related to

specifically:

There are many other considerations, but these tend to be the most important.

This technical report:

This technical report is intended for:

Spectrum Scale, is a parallel file system that is at the heart of IBM Elastic Storage Servers. IBM Spectrum

Scale scales system throughput with each new server while still providing a single name space to the

clients. This ability eliminates data silos and simplifies storage management. By consolidating storage

acquisition costs while simplifying management and improving data protection.

The capabilities of IBM Elastic Storage Server include:

ownCloud.

enables the unification of virtualization, analytics, file and object use cases into a single scale-out storage

ownCloud application and MariaDB database servers have been configured behind a HAProxy load

Figure 9: Transaction throughput versus threads illustrating the average of each test case recorded in real time for

more than 125 threads

Figure 10: Transaction throughput versus threads illustrating the average of each test case for 185 threads

Figure 9 shows the recorded transaction throughput for threads. In this chart, for approximately more than

125 threads (user simulation), the following number of transactions are recorded per second for the test

cases.

In the lab, proof of concept load test results are recorded for a load of 10,000 user threads. The system

can handle each load test case without any issues.

The upload and download test cases using HTTP PUT and GET methods are performed after creating

approximately 1 MB and 1 GB sized random files.

Apart from these test case results, separate stand-alone sync client tests also staged for more than 4500

files and mostly large size files (above 1 GB) are used during stand-alone desktop sync client tests.

The Apache JMeter uses the following formula for calculating the total server transaction throughput.

[<active threads> * 1 second (1000 ms) / <1 thread response time>]

Example: If 10 threads send requests to the ownCloud server and the ownCloud server responds after

HTTP GET and PROPFIND transactions.

performance bench mark results.

In order to estimate the required infrastructure size of a comparable system like the one used in the lab

proof of concept setup, a general assumption on the average user behavior and therefore the distribution

of their access patterns has to be taken into account. Furthermore, to give a quantitative impression within

measured performance to a count of 100,000 active users and translated the results into hardware sizing

When using ownCloud within an enterprise grade environment using the file sync and share system as a

central data hub within the daily work of each user, the desktop sync clients are of major importance. They

clients triggers PROFIND requests for every 30 seconds to check the synchronization status and to take

action, if needed. 100,000 active desktop sync clients using the default setup lead to an average number

of approximately up to 3600 PROFINDs each second. Using the benchmark results, this translates into a

request type, PROPFIND and HTTP_GET. According to lab proof of concept benchmarks, the

performance of about a quarter of the reference application server is needed additionally.

The web portal usage is assumed to be of the same order as mobile access. However, you can exchange

the PROPFIND requests with additional HTTP_GET calls of the same number, resulting in 60 (2 * 30)

HTTP_GET requests per second on the whole system. This translates to about a 0.20 seconds.

Finally, file upload and download have to be taken into account. You can anticipate average rates of one

upload and two downloads per hour. The benchmark results lead to a required number of about 0.75

nodes for downloads and 0.1 nodes for uploads. You can round that up to one node for the file transfers.

Overall, for the 100,000-user scenario (including a buffer of about 20% for the interception of peak load

times, the five hosts for the desktop sync clients, one and a half host for web and mobile access and file

the disks. The tasks that a node can perform are determined by the license type and the cluster

configuration.

the file system components. When failures are detected, IBM Elastic Storage attempts automatic recovery.

Extensive journaling and recovery capabilities help maintain metadata consistency when a node holding

locks or performing administrative services fails.

Finally, IBM Elastic Storage supports snapshots, enabling enterprises to protect the file system’s contents

against user error. Snapshots can be used as an online backup capability that allows files to be recovered

easily from common problems, such as accidental file deletion.

deliver an enterprise-class file sync and share environment for more than 100,000 users.

The test team did not reach the performance limit of the IBM Elastic Storage Server at all, whereas, it

turned out that the sizing of the application and database servers is crucial.

The IBM Elastic Storage Server used for staging the ownCloud file sync and share lab proof of concept

solution for performance and sizing information was simultaneously used for additional backup / restore

solution performance benchmark purpose. This underlines the fact that IBM Elastic Storage can be used

for hosting several industry solutions for storing critical data simultaneously.

The successful cooperation of ownCloud Inc. and IBM enables the delivery of perfectly customized

solutions for your clients. ownCloud provides enterprises with a highly scalable on-premises alternative to

consumer-grade and cloud-based applications. By installing ownCloud on-premises, fully integrated into

existing identity management, security, governance, backup, and disaster recovery tools, organizations