Practical dynamic software updating for C

Abstract

Software updates typically require stopping and restarting an application, but many systems cannot afford to halt service, or would prefer not to. Dynamic software updating (DSU) addresses this difficulty by permitting programs to be updated while they run. DSU is appealing compared to other approaches for on-line upgrades because it is quite general and requires no redundant hardware. The challenge is in making DSU practical: it should be flexible, and yet safe, efficient, and easy to use.In this paper, we present a DSU implementation for C that aims to meet this challenge. We compile programs specially so that they can be dynamically patched, and generate most of a dynamic patch automatically. Our compiler performs a series of analyses that when combined with some simple runtime support ensure that an update will not violate type-safety while guaranteeing that data is kept up-to-date. We have used our system to construct and dynamically apply patches to three substantial open-source server programs-Very Secure FTP daemon, OpenSSH daemon, and GNU Zebra. In total, we dynamically patched each program with three years' worth of releases. Though the programs changed substantially, the majority of updates were easy to generate. Performance experiments show that all patches could be applied in less than 5 ms, and that the overhead on application throughput due to updating support ranged from 0 to at most 32%.

Schema evolution analysis for embedded databases

“…Many solutions exist already for dynamically updating "standard" applications written in C, C++, and Java. For example, dynamic software updating (DSU) systems such as Ginseng [6], Upstare [7], or Jvolve [8] allow on-the-fly updates to code and in-memory data. However, because of their focus on updating code and in-memory state, dynamic updating systems are insufficient for performing online upgrades to applications that require database updates.…”

How to have your cake and eat it too: Dynamic software updating with just-in-time overhead

“…Interestingly, the overhead must exist independently of the type of mechanism used, so it applies to general systems such as Upstare and Ginseng [2], [4] as well as systems that explictly try to avoid the overhead such as Ekiden [5].…”

“…Ginseng [4] transforms off-the-shelf C programs into C programs that can be updated on-the-fly via two main techniques: type wrapping and function indirection. Ginseng allows a wide range of updates to C programs, however, the indirection and type wrapping impose a permanent performance overhead on programs compiled with Ginseng.…”

“…These mechanisms introduce substantial overhead due to indirection introduced by the mechanism, the loss of optimization opportunities by the compiler, as well as the changes in cache locality due to instrumentation by DSU compiler [4], [2]. While some existing low-level mechanisms can avoid that overhead, they are architecture dependent and harder to maintain, but more importantly they are not general mechanisms for dynamic software update.…”

Online Adaptive Code Generation and Tuning

“…Dynamic code-generation and runtime loading of different versions of code-sections is a technique that has been used both in the context of dynamic software updating [15] and auto-tuning [24]. ADAPT [24] is a compiler-supported infrastructure for high-level adaptive program optimization.…”